{"paper_id":"4bd1de01-0d59-4bf5-9eaf-250ee85d3ac4","body_text":"Machine Learning-Based Modeling of Land Surface Temperature in Lagos, Nigeria: Integrating Canopy Structure, Built Environment, and Surface Reflectance Variables | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Machine Learning-Based Modeling of Land Surface Temperature in Lagos, Nigeria: Integrating Canopy Structure, Built Environment, and Surface Reflectance Variables Peter Edemewe Ugege, Ayodotun Olufemi Bobadoye, Peace Amarachi Ukoha, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8034609/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract Urban heat is an escalating environmental challenge in tropical megacities, where rapid urbanization and declining vegetation cover intensify surface warming. This study applies a machine learning approach to predict and map Land Surface Temperature (LST) across Lagos, Nigeria, by integrating multisource remote sensing variables within a Random Forest (RF) framework. Three models of increasing complexity were developed using combinations of vegetation, structural, and spectral predictors derived from Landsat 8, GlobeFCH canopy height, ESA WorldCover, and SRTM data. Model 1, using vegetation variables (NDVI and canopy height), achieved an R² of 0.51, while Model 2, incorporating built-up and elevation variables, improved performance to R² = 0.66. The final model (Model 3), combining NDVI, canopy height, built-up percentage, elevation, NDBI, LULC, and albedo, achieved the best accuracy (R² = 0.74; RMSE = 1.77°C; MAE = 1.21°C). Partial dependence analysis revealed that NDVI and canopy height exert strong cooling effects, whereas NDBI and albedo were positively associated with surface warming. Spatial predictions highlighted pronounced thermal gradients, with high LST values concentrated in industrial and densely built-up areas like Ikeja, Apapa, Lagos Island; and cooler conditions in vegetated and coastal zones. These findings underscore the role of vegetation structure in mitigating urban heat and provide actionable spatial insights for urban greening and climate adaptation planning. The reproducible workflow demonstrates the potential of machine learning and Earth observation data for urban climate monitoring in data-limited tropical regions. Land Surface Temperature Random Forest Urban Heat Island NDVI Canopy Height Lagos Remote Sensing Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 1. Introduction Urbanization is intensifying global heat exposure, especially in tropical megacities such as Lagos, Nigeria, where dense built environments, impervious surfaces, and reduced vegetation exacerbate the urban heat island (UHI) effect (Obe et al., 2024 ). Elevated surface temperatures contribute to increased energy demand, health risks (e.g., heat stress, respiratory problems), and challenges for vulnerable populations, particularly in informal settlements(Abou Samra, 2023 ; Tanoori et al., 2024 ). Vegetation plays a critical moderating role through evapotranspiration and shade; recent advances in remote sensing have enabled more detailed mapping of canopy structure and vegetation indices such as NDVI at increasingly fine spatial and temporal scales (Lang et al., 2023 ; Nie et al., 2023 ). Machine learning (ML), particularly ensemble methods such as Random Forest and Gradient Boosting, has become a powerful tool to relate remotely sensed predictors (e.g., canopy height, spectral indices, built-up area) to land surface temperature (LST), allowing spatially explicit predictions and scenario analysis (Tanoori et al., 2024 ). Global datasets such as the GlobeFCH (Global Forest Canopy Height 2020, 30 m) offer vegetation structural variables over large spatial extents, and high fidelity canopy products at 10 m resolution are now available to support such analyses (Lang et al., 2023 ; Nie et al., 2023 ). In the context of Lagos, prior work has documented increasing LST trends and land cover changes (Emekwuru & Ejohwomu, 2023 ; Obe et al., 2024 ). However, few studies have performed spatially explicit ML-based LST modelling that integrates canopy height, vegetation indices, and other environmental predictors at high resolution to identify thermal hotspots and prioritize urban greening zones. This research bridges that gap by developing and rigorously comparing three distinct Random Forest models designed to estimate Land Surface Temperature (LST) at a fine spatial resolution (~ 100 m) throughout Lagos. The models are structured to progressively test the predictive power of various biophysical and surface parameters. Model 1 employs an initial set of vegetation-related predictors (NDVI and canopy height). Model 2 introduces key urban and elevation characteristics by adding built-up proportion and elevation. The full-featured Model 3 incorporates a complete set of surface descriptors, including NDBI, Land-Use/Land-Cover (LULC), and albedo. The objectives are to (1) evaluate the predictive performance of each model, (2) analyze spatial variations in predicted LST, and (3) assess the relative importance of vegetation, built-up density, and surface properties in shaping urban heat patterns. We hypothesize that the inclusion of built‐up and spectral surface variables will significantly enhance model accuracy, and that NDVI, NDBI, and built‐up percentage will emerge as dominant predictors of LST variability. The results provide spatially explicit evidence for urban heat mitigation and greening strategies in Lagos. 2. Methodology 2.1 Study Area Lagos State, Nigeria’s largest metropolitan area, is a rapidly expanding coastal megacity located along the Gulf of Guinea between 6.35°–6.75° N and 3.0°–3.6° E. The city’s low-lying coastal plain, complex lagoon network, and limited elevation (mostly < 50 m above sea level) make it particularly prone to heat accumulation (Omosanya et al., 2025 ). The climate is tropical monsoon (Köppen Am), characterized by high humidity and two main seasons: a wet season (April–October) and a dry season (November–March). Annual rainfall exceeds 1,500 mm, while mean daily temperatures range from 26°C to 34°C (Akinwumi et al., 2023 ; Emekwuru & Ejohwomu, 2023 ; Njoku et al., 2023 ; Ogunjo et al., 2021 ). Figure 1 presents the geographic setting of Lagos State, highlighting major roads, urban centers, waterbodies, and coastal features that characterize its low-lying terrain and urban sprawl. Lagos’s rapid urbanization has led to extensive conversion of vegetated land to impervious surfaces, with built-up areas increasing from ~ 23.6% in 2000 to 47.2% in 2022 (Emekwuru & Ejohwomu, 2023 ). This has intensified the UHI effect, producing surface temperature differences of up to ~ 5°C between urban cores and peri-urban zones(Obiefuna et al., 2021 ). Vegetation loss and canopy degradation have further reduced urban cooling capacity (Ayanlade, 2017 ; Sulaiman et al., 2025 ). Recent advances such as GEDI LiDAR and Sentinel-2 data now enable fine‐scale mapping of canopy height and structure (Tsao et al., 2023 ), providing new opportunities to quantify vegetation–temperature interactions. Given these conditions, Lagos serves as an ideal case study for investigating how vegetation, surface composition, and built-up intensity jointly influence LST in tropical urban environments. 2.2 Data Sources This study integrated multi-source remote sensing datasets to model and map spatial patterns of land surface temperature (LST) across Lagos. The datasets included canopy height, vegetation indices, built-up area, elevation, surface reflectance, and derived spectral indices. The dependent and predictor variables were derived from multiple Earth observation sources, including Landsat 8, GlobeFCH 2020, ESA WorldCover, and SRTM DEM. LST was derived from Landsat 8 TIRS Band 10 (Collection 2, Tier 1) after atmospheric correction and conversion from Kelvin to Celsius following USGS guidelines(Barsi et al., 2014 ). Vegetation structure and greenness were represented by canopy height GlobeFCH 2020 (Nie et al., 2023 ) and the Normalized Difference Vegetation Index (NDVI = (NIR – RED)/(NIR + RED)), which captures the cooling influence of vegetated surfaces. Built-up density was quantified using the built-up percentage (Built_pct) and the Normalized Difference Built-up Index (NDBI = (SWIR – NIR)/(SWIR + NIR)), both indicators of impervious surface extent(Huang et al., 2021 ). Elevation (from SRTM DEM) and surface albedo (computed from Landsat reflectance bands following (Liang, 2001 )) were included to account for topographic and radiative influences on surface temperature. Land-use/land-cover (LULC) data were obtained from ESA WorldCover 2020 (10 m) and resampled to 100 m for modelling consistency.(Zhang et al., 2022 ). A summary of all variables, their descriptions, and data sources is provided in Table 1 . Table 1 Description and data sources of predictor variables used in the Random Forest models. Variable Description Units / Index Source Resolution LST (Target) Land Surface Temperature (dependent variable) °C Landsat 8 TIRS (USGS) 100 m NDVI Normalized Difference Vegetation Index Unitless Landsat 8 OLI 100 m Canopy Height Forest canopy height Meters GlobeFCH 2020 30 m resampled Built_pct Built-up percentage per pixel % ESA WorldCover (2020) 10 m resampled Elevation Digital elevation model Meters SRTM DEM (NASA JPL, 2013) 30 m resampled NDBI Normalized Difference Built-up Index Unitless Landsat 8 OLI 100 m LULC Land-use / land-cover class Class (1–5) ESA WorldCover (2020) 10 m resampled Albedo Broadband surface albedo Unitless Computed from Landsat 8 OLI 100 m 2.3 Data Pre-processing and Integration Pre-processing was conducted using Google Earth Engine (GEE) and Python libraries ( rasterio, numpy, geopandas, pandas ). The Lagos administrative boundary was used to mask all datasets. Raster values for each predictor layer (NDVI, canopy height, Built_pct, elevation, NDBI, LULC, albedo) were extracted and stacked into a multiband GeoTIFF. Missing values were assessed: predictors with minor gaps; like NDBI, albedo were imputed using column‐wise mean substitution. A stratified random sample of approximately 10,000 points was extracted from valid pixels for modelling. The overall methodological workflow is summarized in Fig. 2 , which outlines the data sources, preprocessing procedures, model training, and spatial prediction steps used in the Random Forest framework. 2.4 Model Development and Refinement A three-tier Random Forest (RF) regression framework was implemented in Python using the scikit-learn library to model land surface temperature (LST) across Lagos. Each model incorporated a progressively richer set of predictors to evaluate the relative contribution of vegetation, built-up intensity, and surface properties. The three models were defined as follows: Model 1 included NDVI and canopy height; Model 2 included NDVI, canopy height, Built_pct, and elevation; while Model 3 incorporated NDVI, canopy height, Built_pct, elevation, NDBI, LULC, and albedo. The dataset was divided into an 80:20 train–test split to ensure robust evaluation and avoid overfitting. Hyperparameters, including the number of trees ( n_estimators = 500), maximum tree depth, and minimum samples per split, were optimized using grid search and 10-fold cross-validation. The parameter n_jobs = − 1 enabled full parallel computation using all available CPU cores. Model performance was assessed using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). Feature importance scores were derived from the RF’s internal Gini importance metric to quantify the relative contribution of each predictor to LST prediction. The configuration of the three Random Forest models, including predictor composition and key hyperparameters, is summarized in Table 2 . Table 2 Random Forest model configurations and predictors used for land surface temperature (LST) estimation in Lagos. Model ID Predictors Used Number of Trees (n_estimators) Max Depth Cross-validation n_jobs Model 1 NDVI, Canopy Height 500 Auto 10-fold -1 Model 2 NDVI, Canopy Height, Built_pct, Elevation 500 Auto 10-fold -1 Model 3 NDVI, Canopy Height, Built_pct, Elevation, NDBI, LULC, Albedo 500 Auto 10-fold -1 2.5 Mapping and Output Generation Predicted LST surfaces for each model were mapped in Python using matplotlib and rasterio, and further refined in QGIS 3.34. The outputs included predicted LST maps, residual maps (Observed – Predicted), and feature importance charts. All raster datasets were projected to WGS 84 (EPSG:4326) to ensure spatial compatibility and were exported in both GeoTIFF and PNG formats. Results from Models 1 and 2 were archived as supplementary figures, while Model 3 outputs were used for the main analysis. 2.6 Validation and Reproducibility Model validation combined statistical metrics (R², RMSE, and MAE), residual mapping to assess spatial bias, and scatterplots comparing observed and predicted LST values. The complete Python-based Random Forest workflow is provided in the supplementary notebook (File S3), while preliminary GEE preprocessing steps are described in the Methods section 2.7 Model Interpretability To gain deeper insights into how individual predictors influence modelled Land Surface Temperature (LST), Partial Dependence Plots (PDPs) were generated for the best-performing model (Model 3). PDPs illustrate the marginal effect of one or more predictors on the model’s predicted outcome while averaging over the influence of other variables (He et al., 2023 ). This approach provides an interpretable visualization of nonlinear or threshold relationships between predictors and LST that may not be evident from feature importance rankings alone. Four primary predictors were selected for interpretability analysis: NDVI, Built-up Percentage (Built_pct), Normalized Difference Built-up Index (NDBI), and Albedo because they represent distinct biophysical and surface-energy dimensions of the urban environment. NDVI was expected to show a negative relationship with LST, reflecting vegetation-induced cooling, while Built_pct and NDBI were hypothesized to increase LST due to higher impervious surface density. Albedo was examined for its moderating influence on surface reflectance and heat absorption. Partial Dependence Plots (PDPs) were computed using the PartialDependenceDisplay function from scikit-learn and visualized in Matplotlib with a grid resolution of 100 to ensure smooth response curves. For each selected predictor, the average predicted LST across all observations was plotted against the predictor’s value range. The resulting plots provided a mechanistic understanding of the model’s behavior and were later referenced in the Discussion to interpret urban heat dynamics in Lagos. 3. Results 3.1 Model Performance Evaluation The three Random Forest (RF) models exhibited progressive improvements in predicting land surface temperature (LST) as additional predictors were incorporated (Table 3 ). Model 1, which used only vegetation-based predictors (NDVI and canopy height), achieved an R² of 0.51 and RMSE of 2.44°C, indicating moderate explanatory power primarily driven by vegetation greenness and canopy structure. Introducing built-up and topographic variables in Model 2 increased predictive accuracy (R² = 0.66, RMSE = 2.04°C, MAE = 1.43°C), suggesting that structural and elevation factors considerably influence LST variability across Lagos. The full-feature Model 3, which integrated additional predictors including NDBI, LULC, and surface albedo, provided the highest performance (R² = 0.74, RMSE = 1.77°C, MAE = 1.21°C). This model was therefore selected for spatial prediction and further analysis. The sequential improvement across models highlights the value of integrating spectral and structural variables for explaining the thermal heterogeneity of Lagos’ urban landscape. Table 3 Summary of Random Forest model performance for LST prediction in Lagos. Model Predictor Variables R² RMSE (°C) MAE (°C) Interpretation Model 1 NDVI, Canopy Height 0.5113 2.44 1.75 Vegetation-only predictors provide moderate explanatory power — NDVI and canopy height account for about 51% of LST variance. Model 2 NDVI, Canopy Height, Built_pct, Elevation 0.6582 2.04 1.43 Addition of built-up and topographic variables improves prediction by ~ 15%, reflecting structural and terrain influence. Model 3 NDVI, Canopy Height, Built_pct, Elevation, NDBI, LULC, Albedo 0.7438 1.77 1.21 Comprehensive model incorporating spectral and surface properties yields the best performance; ~74% of LST variance explained. The performance metrics obtained for all three models indicate a robust and credible predictive framework for LST estimation across Lagos. The final Random Forest model (Model 3) achieved an R² of 0.74, RMSE of 1.77°C, and MAE of 1.21°C, demonstrating strong predictive accuracy in a complex tropical urban environment. These results are consistent with similar Random Forest-based LST models reported in recent studies by (Z. Li et al., 2023 ; Tanoori et al., 2024 ; Ünsal et al., 2023 ), where R² values typically range between 0.65 and 0.78 for heterogeneous urban landscapes. Given the inherent variability of land surface conditions and sensor-related uncertainties in remote sensing data, this level of performance reflects a well-generalized and reliable model capable of accurately reproducing spatial temperature patterns across the Lagos metropolis. 3.2 Correlation Analysis of Predictor Variables A correlation heatmap was generated to examine relationships among all predictor variables and their association with land surface temperature (lst_l8) (Fig. 3 ). NDVI showed a strong negative correlation with lst_l8 (r ≈ − 0.72), confirming its significant cooling influence through vegetation cover and evapotranspiration. In contrast, Built_pct and NDBI exhibited strong positive correlations (r ≈ + 0.64 and + 0.68, respectively), highlighting the warming effect of impervious surfaces. Albedo and Elevation demonstrated weak-to-moderate negative correlations with lst_l8, suggesting minor cooling effects linked to reflective and elevated surfaces. These relationships emphasize the opposing roles of vegetation and built-up surfaces in shaping the city’s thermal environment and provide a foundation for understanding variable interactions before modeling. Importantly, the moderate intercorrelations among predictors like between Built_pct and NDBI, underscore the need for algorithms like Random Forest, which can handle multicollinearity and complex nonlinear effects. 3.3 Variable Importance Feature importance analysis revealed distinct shifts in predictor influence as model complexity increased (Fig. 4 ). In Model 1, NDVI dominated, accounting for more than 80% of the total predictive power, indicating that vegetation greenness alone strongly influences surface cooling in the absence of built-environment variables. In Model 2, Built_pct became the most influential predictor (~ 40%), followed by NDVI (~ 30%) and Elevation (~ 20%). The growing importance of Built_pct underscores the role of impervious surfaces in controlling LST variability. Model 3 exhibited a more balanced influence distribution. Built_pct, NDVI, NDBI, and LULC contributed comparably to LST prediction, while Albedo and Elevation had moderate effects. Canopy height showed the lowest relative importance across all models. These results highlight that both vegetation cover and built-up density jointly determine spatial variations in surface temperature across Lagos. 3.4 Spatial Pattern of Predicted Land Surface Temperature (LST) The spatial distribution of predicted LST from the best-performing Random Forest model (Model 3) is presented in Fig. 5 , revealing clear spatial heterogeneity in surface temperature across Lagos. Higher LST values (approximately 40–45°C) were concentrated in densely built-up and industrial zones such as Ikeja, Apapa, Lagos Island, Surulere, and Agege, while lower temperatures (around 20–30°C) were observed in vegetated and coastal areas of Lekki, Epe, and Badagry. Intermediate temperature zones corresponded to mixed-use landscapes. Compared with Models 1 and 2, Model 3 produced more coherent and realistic LST patterns. The inclusion of NDBI, LULC, and Albedo enhanced the model’s ability to distinguish between impervious and vegetated surfaces, resulting in a more accurate depiction of urban heat island (UHI) patterns. 3.5 Land Surface Temperature Variation Across Land-Use Classes To further interpret LST variability, predicted lst_l8 values were analyzed across five major land-use/land-cover (LULC) classes (Fig. 6 ). Distinct temperature gradients were evident: built-up and bare/transitional areas recorded the highest median LST (~ 33–35°C), reflecting the dominance of impervious materials and low vegetation cover. Vegetation and wetland classes exhibited markedly lower median temperatures (~ 28–30°C), consistent with the cooling effects of canopy cover and evapotranspiration. Water bodies were the coolest overall but displayed moderate variability, likely due to shallow lagoon heating and mixed-pixel effects near shorelines. These contrasts provide clear empirical evidence of the urban heat island effect in Lagos and validate the model’s capacity to capture thermal differences among land-cover types. The findings emphasize that expanding vegetative cover can significantly mitigate surface heating in densely urbanized districts. 3.6 Residual Analysis Residual maps (Observed – Predicted) were used to assess spatial bias and uncertainty in model performance (Fig. 7 ). In Model 1, residuals exhibited strong clustering, with underestimation (positive residuals) in dense urban areas and overestimation (negative residuals) in vegetated zones. Model 2 reduced this bias but still showed localized under-prediction in built-up zones. Model 3 demonstrated the lowest and most spatially uniform residuals, with errors typically within ± 5°C, indicating enhanced predictive stability and minimal bias. Residual maps for Models 1 and 2 are provided in the Supplementary Materials (Figures S3 –S4). 3.7 Partial Dependence Analysis Partial Dependence Plots (PDPs) were used to examine the marginal effects of key predictors on LST (Fig. 8 ). NDVI displayed a strong negative association with LST, confirming that increasing vegetation greenness reduces surface temperature through evapotranspiration and shading. Built_pct and NDBI exhibited positive nonlinear relationships, with rapid temperature increases at moderate urban densities that stabilized at higher built-up fractions. Albedo had a mild, nearly linear positive relationship with LST, suggesting that reflective materials in Lagos do not significantly mitigate heating where vegetation is absent. Collectively, the PDPs affirm that vegetation cover is the most effective biophysical control on surface temperature, while built-up intensity and surface composition drive localized heat accumulation. 3.8 Predicted vs. Observed Validation The scatter plot comparing observed and predicted LST values for Model 3 (Fig. 9 ) shows a strong linear relationship, with most points clustering along the 1:1 reference line. The Random Forest model achieved R² = 0.74, RMSE = 1.77°C, and MAE = 1.21°C, confirming robust predictive performance consistent with the official evaluation metrics reported in Table 3 . The model slightly underestimates extreme high-temperature values in dense urban cores and marginally overestimates lower values in vegetated and coastal areas, reflecting the spatial heterogeneity typical of tropical megacities. However, these deviations are minor and do not affect the overall model accuracy or spatial coherence. The consistency between independent validation and model evaluation underscores the reliability and generalization capacity of the Random Forest approach. Overall, the model provides a stable and accurate spatial representation of land surface temperature in Lagos suitable for urban heat analysis and policy planning. 3.7 Summary of Findings The Random Forest models effectively captured the spatial variability of land surface temperature (LST) across Lagos, with Model 3, integrating NDVI, canopy height, built-up percentage, elevation, NDBI, LULC, and albedo, achieving the highest predictive accuracy (R² = 0.74; RMSE = 1.77°C; MAE = 1.21°C). Feature importance results indicated that Built_pct, NDVI, and NDBI were the most influential predictors, while Albedo, Elevation, and Canopy Height showed moderate contributions. Spatial predictions revealed clear temperature contrasts between densely built-up districts such as Ikeja, Apapa, Lagos Island, Surulere, and Agege, and the cooler, vegetated or coastal areas of Lekki, Epe, and Badagry. Residual and validation analyses confirmed that Model 3 produced the most spatially coherent and unbiased predictions, demonstrating that the LST distribution in Lagos is shaped by the combined influence of vegetation cover, urban density, and surface reflectance characteristics. 4. Discussion 4.1 Urban heat in tropical megacities The results of our study reinforce the understanding that tropical megacities such as Lagos experience pronounced surface heating due to rapid urbanization, dense built environments, and loss of vegetation. Prior research demonstrated that urban growth in tropical climates intensifies land-surface and air temperature differentials more rapidly than in many temperate cities (Marcotullio et al., 2021 ). In Lagos in particular, long‐term analysis of the UHI (urban heat island) indicates that land‐surface and near‐surface warming has been accelerating alongside built‐up expansion and vegetation decline (Bassett et al., 2020 ). The relatively high coefficient of determination (R² = 0.74) achieved by our best-performing model demonstrates that a combined set of canopy, spectral, and built-cover variables can explain a substantial portion of the spatial variability in LST across Lagos, reflecting the dominant influence of urban form and land cover in tropical heat dynamics. Additionally, recent global studies of tropical city heat exposure highlight that even a 1°C increase in surface temperature can significantly raise human heat‐stress risks and energy demand in low-income settlements (Ramsay et al., 2021a ; Zhou et al., 2022 ). 4.2 Role of vegetation and canopy structure One of our key findings is the strong negative relationship between NDVI and canopy height and predicted LST, confirming that vegetation and canopy structure are critical for urban cooling. This aligns with studies showing that higher vegetation greenness and forest-canopy height reduce surface temperatures through shading and evapotranspiration (Alonzo et al., 2025 ; Bedra & Li, 2025 ; Lima Alves & Lopes, 2017 ; Pfeifer et al., 2019 ). Our partial dependence plot for NDVI showed a continuous cooling effect across the range of values, with no clear saturation observed. The boxplot results, which show lower median predicted LST values for vegetation and wetland land-cover classes (~ 28–30°C) compared to built-up zones (~ 33–35°C), further demonstrate how vegetation loss in Lagos increases vulnerability to elevated surface temperature. These findings support the argument that urban greening and canopy preservation should be top priority in tropical cities. Emerging research also suggests that mixed-layer vegetation (trees + shrubs + grasses) may deliver enhanced cooling when integrated into urban form (Gallay et al., 2023 ; Tan et al., 2022 ) 4.3 Built environment, imperviousness and surface reflectance In contrast, our results show that built-up percentage (Built_pct) and the Normalized Difference Built‐up Index (NDBI) are strong positive predictors of LST. These findings mirror the notion that impervious surfaces absorb and re‐emit more radiation, reduce evaporative cooling, and trap heat in urban fabric (Lima Alves & Lopes, 2017 ). The marginal effect curves for Built_pct and NDBI show steep increases in LST for moderate to high values of urban cover, indicating non‐linear sensitivities: once impervious cover passes a threshold, surface heating may escalate rapidly. These results have practical importance for urban design: limiting impervious cover, ensuring pervious/vegetated surfaces, and using higher‐reflectance materials may reduce surface heating. Interestingly, the positive but moderate effect of Albedo suggests that while surface reflectance has some mitigation potential, in the Lagos context it appears secondary to vegetation and imperviousness. Similar studies suggest that reflectance and albedo changes must be coupled with vegetation interventions to yield meaningful cooling (Bedra & Li, 2025 ; Luo et al., 2023 ). 4.4 Spatial-temporal patterns and implications for urban planning Our spatial maps derived from Model 3 clearly delineated high-temperature “hotspots” in dense built-up districts (e.g., Ikeja, Surulere, Lagos Island) and cooler zones in vegetated and coastal districts like Lekki, Epe and Badagry. This pattern aligns with previously observed UHI gradients in Lagos and other tropical cities (Bassett et al., 2020 ). The residual analysis; showing that our model achieved enhanced spatial coherence and lower bias compared to simpler models, provides confidence that the predictions are reliable for planning purposes. For urban planners and policy-makers in Lagos, these results mean that targeted interventions should focus on built-up hotspots: greening, canopy enhancement, and surface material redesign. Equally, preserving existing vegetated and coastal zones is critical to maintaining cooling capacity. Because surface temperature correlates strongly with energy demand, outdoor thermal comfort, and heat stress; especially under high humidity in tropical climates where every degree of surface temperature reduction may translate into meaningful public health and energy savings (Ramsay et al., 2021b ; Santamouris, 2020 ; Ziter et al., 2019 ) 4.5 Limitations and future research Despite strong model performance, there are important caveats. First, our dependent variable is land surface temperature (lst_l8) rather than air temperature; while LST is a widely accepted proxy for surface heating, human-experienced thermal impact may differ. Second, the model uses a single raster stack at 100 m resolution and dry-season imagery, meaning that seasonal or diurnal variations may not be fully captured. Future work should incorporate multi-temporal data (wet and dry seasons), diurnal cycles, and time-series of heating patterns (Naserikia et al., 2023 ; Winckler et al., 2019 ). Third, although vegetation indices and canopy height were included, micro-scale urban morphological variables like building height, street canyon geometry and ventilation were not explicitly integrated; the inclusion of such features could further refine predictions and urban niche-scale gradients (Y. Li et al., 2020 ; Peng & Huang, 2022 ; Yin et al., 2022 ). Finally, the observed moderate effect of albedo suggests that additional field-based data on material reflectance and thermal emissivity may enhance the understanding of surface energy dynamics (Solanki et al., 2025 ) 4.6 Policy relevance and actionable recommendations From a policy perspective, our evidence supports several actionable planning strategies for Lagos and comparable tropical urban centers: Prioritize tree-canopy planting and preservation in dense built-up zones, especially informal settlements and industrial estates, to maximize cooling benefits. Limit the expansion of impervious surfaces and promote permeable pavements and green-infrastructure corridors. Integrate thermal-model outputs (as demonstrated here) into urban-greening policy to prioritize intervention zones based on high-predicted LST hotspots. Use LST maps and predictor-importance rankings to inform municipal bylaws on land-cover change, building-material standards, and urban-planning codes. In summary, surface-temperature mitigation in tropical cities must prioritize structural redesign of the built environment and vegetation interventions, rather than relying solely on reflective-surface or albedo enhancements. By linking remotely-sensed datasets with machine-learning models, this study provides a transferrable framework for urban-climate adaptation in data-limited tropical environments. 5. Conclusion This study applied a Random Forest regression framework to model the spatial distribution of land surface temperature (LST) across Lagos, Nigeria, using high-resolution remote sensing data and biophysical predictors. Among the three models developed, the most comprehensive configuration (Model 3); integrating NDVI, canopy height, built-up percentage, elevation, NDBI, LULC, and albedo, achieved the highest predictive accuracy (R² = 0.74, RMSE = 1.77°C, MAE = 1.21°C). These results demonstrate that machine-learning approaches can effectively capture the spatial heterogeneity of surface heating in complex tropical urban systems. Vegetation greenness (NDVI) and canopy structure emerged as dominant cooling variables, underscoring the crucial role of urban greenery in mitigating surface heat. In contrast, built-up density and impervious surfaces significantly amplified LST, while albedo exerted only a secondary moderating effect. The resulting LST maps revealed pronounced urban heat island (UHI) patterns, with the highest temperatures concentrated in industrial and densely populated zones (Ikeja, Apapa, Lagos Island, Surulere) and cooler areas aligned with vegetated and coastal regions (Lekki, Epe, Badagry). These findings have strong implications for sustainable urban planning and climate adaptation in tropical megacities. Strategic canopy restoration, integration of green infrastructure, and limitation of impervious surface expansion can yield substantial cooling benefits and enhance urban resilience. The Random Forest modeling framework presented here provides a transferable, data-driven approach for mapping urban thermal environments in other data-limited tropical cities. Future studies should extend this framework to multi-temporal and multi-seasonal analyses, incorporating diurnal variation, urban morphological parameters, and field-measured air temperature data to improve model generalization and deepen understanding of surface-energy interactions. Ultimately, integrating remote sensing, machine learning, and urban design insights can guide evidence-based policies aimed at reducing heat vulnerability and improving livability in rapidly urbanizing regions such as Lagos. Declarations Supplementary Materials The supplementary materials accompanying this article include additional figures, datasets, and code resources supporting the main analysis. Funding The authors received no financial support for the research, authorship, and/or publication of this article. Author Contribution Concept and design was done by A.B.C.Acquisition, analysis, or interpretation of data was by A.B.C.D.Model Development was by A.A.C. Drafted the manuscriptCritical review of the manuscript for important intellectual content was by A.B.C.D. Data Availability All data supporting the findings of this study are available within the paper and its Supplementary Information References Abou Samra, R. M. (2023). “Investigating and mapping day-night urban heat island and its driving factors using Sentinel/MODIS data and Google Earth Engine. Case study: Greater Cairo, Egypt.” Urban Climate , 52 , 101729. https://doi.org/10.1016/j.uclim.2023.101729 Akinwumi, S. A., Ayo-Akanbi, O. A., Omotosho, T. V., & Mastorakis, Nikos. E. (2023). Monthly and Seasonal Variation of Cloud Cover, Humidity and Rainfall in Lagos, Nigeria. WSEAS TRANSACTIONS ON ENVIRONMENT AND DEVELOPMENT , 19 , 1371–1379. https://doi.org/10.37394/232015.2023.19.123 Alonzo, M., Ibsen, P. C., & Locke, D. H. (2025). Urban Trees and Cooling: A Review of the Recent Literature (2018 to 2024). Arboriculture & Urban Forestry , 51 (5), 420–444. https://doi.org/10.48044/jauf.2025.023 Ayanlade, A. (2017). Variations in urban surface temperature: An assessment of land use change impacts over Lagos metropolis. Weather , 72 (10), 315–319. https://doi.org/10.1002/wea.2925 Barsi, J., Schott, J., Hook, S., Raqueno, N., Markham, B., & Radocinski, R. (2014). Landsat-8 Thermal Infrared Sensor (TIRS) Vicarious Radiometric Calibration. Remote Sensing , 6 (11), 11607–11626. https://doi.org/10.3390/rs61111607 Bassett, R., Young, P. J., Blair, G. S., Samreen, F., & Simm, W. (2020). The Megacity Lagos and Three Decades of Urban Heat Island Growth. Journal of Applied Meteorology and Climatology , 59 (12), 2041–2055. https://doi.org/10.1175/JAMC-D-20-0059.1 Bedra, K. B., & Li, J. (2025). Relevance of Ground and Wall Albedo for Outdoor Thermal Comfort in Tropical Savanna Climates: Evidence from Parametric Simulations. Sustainability , 17 (14), 6303. https://doi.org/10.3390/su17146303 Emekwuru, N., & Ejohwomu, O. (2023). Temperature, Humidity and Air Pollution Relationships during a Period of Rainy and Dry Seasons in Lagos, West Africa. Climate , 11 (5), 113. https://doi.org/10.3390/cli11050113 Gallay, I., Olah, B., Murtinová, V., & Gallayová, Z. (2023). Quantification of the Cooling Effect and Cooling Distance of Urban Green Spaces Based on Their Vegetation Structure and Size as a Basis for Management Tools for Mitigating Urban Climate. Sustainability , 15 (4), 3705. https://doi.org/10.3390/su15043705 He, Z., Yang, Y., Fang, R., Zhou, S., Zhao, W., Bai, Y., Li, J., & Wang, B. (2023). Integration of shapley additive explanations with random forest model for quantitative precipitation estimation of mesoscale convective systems. Frontiers in Environmental Science , 10 , 1057081. https://doi.org/10.3389/fenvs.2022.1057081 Huang, S., Tang, L., Hupy, J. P., Wang, Y., & Shao, G. (2021). A commentary review on the use of normalized difference vegetation index (NDVI) in the era of popular remote sensing. Journal of Forestry Research , 32 (1), 1–6. https://doi.org/10.1007/s11676-020-01155-1 Lang, N., Jetz, W., Schindler, K., & Wegner, J. D. (2023). A high-resolution canopy height model of the Earth. Nature Ecology & Evolution , 7 (11), 1778–1789. https://doi.org/10.1038/s41559-023-02206-6 Li, Y., Schubert, S., Kropp, J. P., & Rybski, D. (2020). On the influence of density and morphology on the Urban Heat Island intensity. Nature Communications , 11 (1), 2647. https://doi.org/10.1038/s41467-020-16461-9 Li, Z., Wu, H., Duan, S., Zhao, W., Ren, H., Liu, X., Leng, P., Tang, R., Ye, X., Zhu, J., Sun, Y., Si, M., Liu, M., Li, J., Zhang, X., Shang, G., Tang, B., Yan, G., & Zhou, C. (2023). Satellite Remote Sensing of Global Land Surface Temperature: Definition, Methods, Products, and Applications. Reviews of Geophysics , 61 (1), e2022RG000777. https://doi.org/10.1029/2022RG000777 Liang, S. (2001). Narrowband to broadband conversions of land surface albedo I. Remote Sensing of Environment , 76 (2), 213–238. https://doi.org/10.1016/S0034-4257(00)00205-4 Lima Alves, E., & Lopes, A. (2017). The Urban Heat Island Effect and the Role of Vegetation to Address the Negative Impacts of Local Climate Changes in a Small Brazilian City. Atmosphere , 8 (2), 18. https://doi.org/10.3390/atmos8020018 Luo, Y., Yang, J., Shi, Q., Xu, Y., Menenti, M., & Wong, M. S. (2023). Seasonal Cooling Effect of Vegetation and Albedo Applied to the LCZ Classification of Three Chinese Megacities. Remote Sensing , 15 (23), 5478. https://doi.org/10.3390/rs15235478 Marcotullio, P. J., Keßler, C., Quintero Gonzalez, R., & Schmeltz, M. (2021). Urban Growth and Heat in Tropical Climates. Frontiers in Ecology and Evolution , 9 , 616626. https://doi.org/10.3389/fevo.2021.616626 Naserikia, M., Hart, M. A., Nazarian, N., Bechtel, B., Lipson, M., & Nice, K. A. (2023). Land surface and air temperature dynamics: The role of urban form and seasonality. Science of The Total Environment , 905 , 167306. https://doi.org/10.1016/j.scitotenv.2023.167306 Nie, S., Zhu, X., Wang, C., & Xi, X. (2023). The global 30-m forest canopy height map for 2020 (Version v1) [Dataset]. Zenodo. https://doi.org/10.5281/ZENODO.7643402 Njoku, C. A., Ojeh, V. N., Naabil, E., Ekwe, M. C., Joshua, M. M., Elijah, E., & Malgwi, M. W. (2023). Analysis of Rainfall Trend and Variability in Lagos, Southwestern Nigeria. In J. C. Egbueri, J. O. Ighalo, & C. B. Pande (Eds.), Climate Change Impacts on Nigeria (pp. 115–134). Springer International Publishing. https://doi.org/10.1007/978-3-031-21007-5_7 Obe, O. B., Morakinyo, T. E., & Mills, G. (2024). An assessment of WRF-urban schemes in simulating local meteorology for heat stress analysis in a tropical sub-Saharan African city, Lagos, Nigeria. International Journal of Biometeorology , 68 (5), 811–828. https://doi.org/10.1007/s00484-024-02627-3 Obiefuna, J. N., Okolie, C. J., Nwilo, P. C., Daramola, O. E., & Isiofia, L. C. (2021). Potential Influence of Urban Sprawl and Changing Land Surface Temperature on Outdoor Thermal Comfort in Lagos State, Nigeria. Quaestiones Geographicae , 40 (1), 5–23. https://doi.org/10.2478/quageo-2021-0001 Ogunjo, S. T., Akinsusi, J. O., & Fuwape, I. A. (2021). Trends in extreme temperature indices over Lagos, Nigeria. IOP Conference Series: Earth and Environmental Science , 655 (1), 012003. https://doi.org/10.1088/1755-1315/655/1/012003 Omosanya, K. O. L., Oseni, R., Lawal, M. A., Bankole, A., & Mosuro, G. O. (2025). Evolving dynamics of lagoons in a megacity: Insights on climate and urbanization over four decades. Urban Climate , 61 , 102439. https://doi.org/10.1016/j.uclim.2025.102439 Peng, M., & Huang, H. (2022). The Synergistic Effect of Urban Canyon Geometries and Greenery on Outdoor Thermal Comfort in Humid Subtropical Climates. Frontiers in Environmental Science , 10 , 851810. https://doi.org/10.3389/fenvs.2022.851810 Pfeifer, M., Boyle, M. J. W., Dunning, S., & Olivier, P. I. (2019). Forest floor temperature and greenness link significantly to canopy attributes in South Africa’s fragmented coastal forests. PeerJ , 7 , e6190. https://doi.org/10.7717/peerj.6190 Ramsay, E. E., Fleming, G. M., Faber, P. A., Barker, S. F., Sweeney, R., Taruc, R. R., Chown, S. L., & Duffy, G. A. (2021a). Chronic heat stress in tropical urban informal settlements. iScience , 24 (11), 103248. https://doi.org/10.1016/j.isci.2021.103248 Ramsay, E. E., Fleming, G. M., Faber, P. A., Barker, S. F., Sweeney, R., Taruc, R. R., Chown, S. L., & Duffy, G. A. (2021b). Chronic heat stress in tropical urban informal settlements. iScience , 24 (11), 103248. https://doi.org/10.1016/j.isci.2021.103248 Santamouris, M. (2020). Recent progress on urban overheating and heat island research. Integrated assessment of the energy, environmental, vulnerability and health impact. Synergies with the global climate change. Energy and Buildings , 207 , 109482. https://doi.org/10.1016/j.enbuild.2019.109482 Solanki, P., Amirtham, L. R., & Deb, C. (2025). Optimizing sky view factor and vegetation to mitigate urban heat in hot-humid climates. Discover Cities , 2 (1), 79. https://doi.org/10.1007/s44327-025-00108-9 Sulaiman, K., Mohd Haniffah, M. R., & Shahid, S. (2025). Spatiotemporal Analysis of Urban Heat Island Effects in Nigerian Cities Using MODIS Data. Journal of Advanced Geospatial Science & Technology , 5 (2), 168–229. https://doi.org/10.11113/jagst.v5n2.112 Tan, X., Liao, J., Bedra, K. B., & Li, J. (2022). Evaluating the 3D cooling performances of different vegetation combinations in the urban area. Journal of Asian Architecture and Building Engineering , 21 (3), 1124–1136. https://doi.org/10.1080/13467581.2021.1903905 Tanoori, G., Soltani, A., & Modiri, A. (2024). Machine Learning for Urban Heat Island (UHI) Analysis: Predicting Land Surface Temperature (LST) in Urban Environments. Urban Climate , 55 , 101962. https://doi.org/10.1016/j.uclim.2024.101962 Tsao, A., Nzewi, I., Jayeoba, A., Ayogu, U., & Lobell, D. B. (2023). Canopy Height Mapping for Plantations in Nigeria Using GEDI, Landsat, and Sentinel-2. Remote Sensing , 15 (21), 5162. https://doi.org/10.3390/rs15215162 Ünsal, Ö., Lotfata, A., & Avcı, S. (2023). Exploring the Relationships between Land Surface Temperature and Its Influencing Determinants Using Local Spatial Modeling. Sustainability , 15 (15), 11594. https://doi.org/10.3390/su151511594 Winckler, J., Reick, C. H., Luyssaert, S., Cescatti, A., Stoy, P. C., Lejeune, Q., Raddatz, T., Chlond, A., Heidkamp, M., & Pongratz, J. (2019). Different response of surface temperature and air temperature to deforestation in climate models. Earth System Dynamics , 10 (3), 473–484. https://doi.org/10.5194/esd-10-473-2019 Yin, S., Liu, J., & Han, Z. (2022). Relationship between urban morphology and land surface temperature—A case study of Nanjing City. PLOS ONE , 17 (2), e0260205. https://doi.org/10.1371/journal.pone.0260205 Zhang, N., Nex, F., Kerle, N., & Vosselman, G. (2022). LISU: Low-light indoor scene understanding with joint learning of reflectance restoration. ISPRS Journal of Photogrammetry and Remote Sensing , 183 , 470–481. https://doi.org/10.1016/j.isprsjprs.2021.11.010 Zhou, S., Liu, D., Zhu, M., Tang, W., Chi, Q., Ye, S., Xu, S., & Cui, Y. (2022). Temporal and Spatial Variation of Land Surface Temperature and Its Driving Factors in Zhengzhou City in China from 2005 to 2020. Remote Sensing , 14 (17), 4281. https://doi.org/10.3390/rs14174281 Ziter, C. D., Pedersen, E. J., Kucharik, C. J., & Turner, M. G. (2019). Scale-dependent interactions between tree canopy cover and impervious surfaces reduce daytime urban heat during summer. Proceedings of the National Academy of Sciences , 116 (15), 7575–7580. https://doi.org/10.1073/pnas.1817561116 Additional Declarations No competing interests reported. Supplementary Files FigureS1.png Figure S1: Correlation heatmap of predictor variables. FigureS2.png Figure S2: Predicted LST map from Random Forest Model 1. FigureS3.png Figure S3: Residual map of Random Forest Model 1. FigureS4.png Figure S4: Residual map of Random Forest Model 2. FileS1.tif File S1: Predicted Land Surface Temperature (LST) raster from Model 3 FileS2.csv File S2: Correlation matrix of predictor variables FileS3.ipynb File S3: Jupyter Notebook containing all analytical workflows, including data integration, Random Forest modeling, and validation steps ( biophysical_model_for_lagos_canopy-temperature_CLEANED.ipynb ). FileS4.tif File S4: Stacked predictor raster dataset used for modeling ( Lagos_full_stack_100m.tif ). The Google Earth Engine (GEE) platform was used for initial image preprocessing and data export; these scripts are not included but are fully described in Sections 2.2–2.3. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 19 Dec, 2025 Editor assigned by journal 08 Nov, 2025 Submission checks completed at journal 08 Nov, 2025 First submitted to journal 05 Nov, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {\"props\":{\"pageProps\":{\"initialData\":{\"identity\":\"rs-8034609\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":false,\"archivedVersions\":[],\"articleType\":\"Research Article\",\"associatedPublications\":[],\"authors\":[{\"id\":563722066,\"identity\":\"f71e0898-7c09-42d8-8dd5-d5c02a840d25\",\"order_by\":0,\"name\":\"Peter Edemewe Ugege\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0ElEQVRIiWNgGAWjYPACG34GZhK1pEk2kKrlsGQD0WrNJdIvPvxScV7CvJ33AOOXisMM5vwL8GuxnJFTbCxz5raEzGG+BGaZM4eBIg/wazG4kZMmLdl2u06CmceAWbItDShygKCW9N+S/85JkKIl/Rjjx4YDYC2MH9tsGAzONxDwS88bZmmGY8lgLYcZztjwGNzAr4PBnD394ccfNXYSEvxnDB/+qJCQMzhPyGEMQC/wQDmHgQweBokEQlrYHzD+gHIgDH4CtoyCUTAKRsGIAwAAtEEhFelLhAAAAABJRU5ErkJggg==\",\"orcid\":\"\",\"institution\":\"Forestry Research Institute of Nigeria\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Peter\",\"middleName\":\"Edemewe\",\"lastName\":\"Ugege\",\"suffix\":\"\"},{\"id\":563722067,\"identity\":\"99feea97-577f-4664-99c1-6f70c41137b4\",\"order_by\":1,\"name\":\"Ayodotun Olufemi Bobadoye\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Forestry Research Institute of Nigeria\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Ayodotun\",\"middleName\":\"Olufemi\",\"lastName\":\"Bobadoye\",\"suffix\":\"\"},{\"id\":563722070,\"identity\":\"8b77af32-53f1-493f-b225-38f67401165a\",\"order_by\":2,\"name\":\"Peace Amarachi Ukoha\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Forestry Research Institute of Nigeria\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Peace\",\"middleName\":\"Amarachi\",\"lastName\":\"Ukoha\",\"suffix\":\"\"},{\"id\":563722071,\"identity\":\"f02247b6-125c-41d1-97b9-b955b903ff42\",\"order_by\":3,\"name\":\"John Oseyomon Aigbokhan\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Forestry Research Institute of Nigeria\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"John\",\"middleName\":\"Oseyomon\",\"lastName\":\"Aigbokhan\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2025-11-05 06:08:09\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-8034609/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-8034609/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":99309866,\"identity\":\"1b74c4b2-2d41-4c7f-9544-922ffa676974\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:19\",\"extension\":\"docx\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":78974,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"anonymzieMachineLearningAnalysisofLSTinLagos.docx\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/ef7101026f8eda4e17616236.docx\"},{\"id\":98899050,\"identity\":\"08989414-4c5a-4158-aa06-ff0f91ace04b\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:08\",\"extension\":\"docx\",\"order_by\":1,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":14763,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"FigureCaptions.docx\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/15d834459068ede116cfa591.docx\"},{\"id\":98899056,\"identity\":\"ac1a83de-1cdd-4fb4-b735-329856c3f229\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:08\",\"extension\":\"jpg\",\"order_by\":2,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":5601108,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure1.jpg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/2409f3a75c538c5413caddbe.jpg\"},{\"id\":99309921,\"identity\":\"8dc21614-3a0f-4f66-8800-55a9daca86ab\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:24\",\"extension\":\"png\",\"order_by\":3,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":216566,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/61f6f29d103a754e873506eb.png\"},{\"id\":99309701,\"identity\":\"b4bd8a43-57c9-4d05-9388-ed64cfa52ab8\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:00\",\"extension\":\"png\",\"order_by\":4,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":275569,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/0f1fa318401083db48d4cf30.png\"},{\"id\":98899060,\"identity\":\"e36f2d93-6ef3-4b5b-8fb1-a9e03620df8a\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:08\",\"extension\":\"png\",\"order_by\":5,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":66071,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/d47bf87faa7aa7babb7fd3b2.png\"},{\"id\":99309815,\"identity\":\"9b1f9948-2501-4668-9a7d-8b196092c22a\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:12\",\"extension\":\"png\",\"order_by\":6,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":1083549,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/cc0b8739d72a848aa4f529c1.png\"},{\"id\":98899074,\"identity\":\"c2ea010b-d475-449e-9830-cff064391c4f\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"png\",\"order_by\":7,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":165097,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure6.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/d20665f8a4a5d00f4fa6945d.png\"},{\"id\":99310002,\"identity\":\"59f398e8-0061-4f9b-8aeb-f7e4f34085e9\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:36\",\"extension\":\"png\",\"order_by\":8,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":1300612,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure7.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/1681ba06e4b4a07012a9b84b.png\"},{\"id\":99310004,\"identity\":\"2d28fb7e-60f5-4ee2-8e72-c24097749d3c\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:37\",\"extension\":\"png\",\"order_by\":9,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":260795,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure8.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/96b1ab49ba2431c3b4b511da.png\"},{\"id\":98899104,\"identity\":\"14731103-768c-4578-86f0-84225fd1260e\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:10\",\"extension\":\"png\",\"order_by\":10,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":331860,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure9.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/c1996f93f33beaa34f4db73e.png\"},{\"id\":98899097,\"identity\":\"9500e3aa-21e0-4455-af38-07c65cdecba3\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"json\",\"order_by\":11,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":6069,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"e1814ff6c5a34f51897bbcf0ada1b3b3.json\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/c3c7e69c2a498833043f81c8.json\"},{\"id\":99309846,\"identity\":\"c90003a4-f921-4be3-bd80-d8118b44e30a\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:17\",\"extension\":\"png\",\"order_by\":12,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":46712,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"FigureS1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/0bd12180306ebbbf73c92aa6.png\"},{\"id\":98899075,\"identity\":\"37f27f7c-3291-44a0-9526-8a6754f05306\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"png\",\"order_by\":13,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":53865,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"FigureS2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/a0e2093bf12faf2725193980.png\"},{\"id\":98899072,\"identity\":\"28be287b-d304-4a37-adf0-ce96b990ee2f\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:08\",\"extension\":\"png\",\"order_by\":14,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":442954,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"FigureS3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/3462bc09a0c48c3638afe50a.png\"},{\"id\":98899090,\"identity\":\"5b138147-789f-4847-b7c3-86bf832adfa0\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"png\",\"order_by\":15,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":437177,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"FigureS4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/3c03c54627853c066fc40931.png\"},{\"id\":99309658,\"identity\":\"e797d7b1-4885-46cd-a690-eb942a7663e8\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:10:55\",\"extension\":\"tif\",\"order_by\":16,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":2738794,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"FileS1.tif\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/2755771169c3a62029110474.tif\"},{\"id\":98899076,\"identity\":\"8c0efb6c-9bae-4afe-a8a1-a36670cd4b96\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"csv\",\"order_by\":17,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":1976,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"FileS2.csv\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/2354b3b63d652113eb012c30.csv\"},{\"id\":99310003,\"identity\":\"89fdb91c-30a2-4110-8912-52415b4fcdd6\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:37\",\"extension\":\"ipynb\",\"order_by\":18,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":399620,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"FileS3.ipynb\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/02b164dc2c92622af678ba55.ipynb\"},{\"id\":98899101,\"identity\":\"5a56bd15-4638-4917-a12d-45d2762682d6\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"tif\",\"order_by\":19,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":27365268,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"FileS4.tif\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/bbb2a644a152d685b2b17cb7.tif\"},{\"id\":99309791,\"identity\":\"b8ea145b-2121-49ba-90fd-5f3f87c82d5c\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:11\",\"extension\":\"xml\",\"order_by\":20,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":124151,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"e1814ff6c5a34f51897bbcf0ada1b3b31enriched.xml\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/109e1d51e9043c215e4878f5.xml\"},{\"id\":99309795,\"identity\":\"71913324-03e5-406c-b461-0af68a1d27a8\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:12\",\"extension\":\"jpg\",\"order_by\":21,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":5601108,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure1.jpg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/f0ad30c6f9413561517123ec.jpg\"},{\"id\":99309509,\"identity\":\"1a2bb92c-d0da-498b-a259-7bdca936deb4\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:10:34\",\"extension\":\"png\",\"order_by\":22,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":216566,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/f3e0923657f117d92be59b97.png\"},{\"id\":98899079,\"identity\":\"2f627979-60fa-4043-ae79-32ff327cd3fe\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"png\",\"order_by\":23,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":275569,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/eda028e35f456e3ea6edeb21.png\"},{\"id\":99309677,\"identity\":\"e9c869e6-8eed-412e-9f5f-78fa4bb87d46\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:10:56\",\"extension\":\"png\",\"order_by\":24,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":66071,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/6fbaaa4e64aa5607286bbeba.png\"},{\"id\":98899093,\"identity\":\"889d470f-e14b-4ac4-9130-0f95479fb01b\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"png\",\"order_by\":25,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":1083549,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/f519b43f99f6f8b8c38fda28.png\"},{\"id\":98899083,\"identity\":\"8ff5af86-04e5-4dc8-a95e-6e0dbb7e50d8\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"png\",\"order_by\":26,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":165097,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure6.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/b6df1f16013f2f2a132b1587.png\"},{\"id\":98899081,\"identity\":\"847c0049-ba37-4469-9c79-4980ef041c1a\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"png\",\"order_by\":27,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":1300612,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure7.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/69b2430dd63fe8e1c60465af.png\"},{\"id\":99309817,\"identity\":\"c234a6b2-a6ee-4310-b2fe-2ee873cb1e74\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:13\",\"extension\":\"png\",\"order_by\":28,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":260795,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure8.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/51e20725a6f6894b86ceec63.png\"},{\"id\":99309642,\"identity\":\"7f173104-4d32-40a8-a259-117f4d98ae5a\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:10:51\",\"extension\":\"png\",\"order_by\":29,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":331860,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Figure9.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/6dcd2789fb60648d6e3c6366.png\"},{\"id\":99309778,\"identity\":\"b7bed223-f486-43ff-b5f0-eed3fecb8635\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:08\",\"extension\":\"png\",\"order_by\":30,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":607863,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"OnlineFigure1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/f0e6b30d48359bdc01a46047.png\"},{\"id\":99310048,\"identity\":\"f563390e-df00-4c4f-bb50-3d96a95d4b5f\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:45\",\"extension\":\"png\",\"order_by\":31,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":68580,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"OnlineFigure2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/82ee5b8653972d227df8e71b.png\"},{\"id\":99309896,\"identity\":\"a9277454-cbb8-4d4c-bda7-e9dfd01a38ad\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:21\",\"extension\":\"png\",\"order_by\":32,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":109700,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"OnlineFigure3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/750a0e528592b17180ba6d0c.png\"},{\"id\":99309688,\"identity\":\"b25181c4-0f8f-4716-acdf-5453d4fc53e9\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:10:57\",\"extension\":\"png\",\"order_by\":33,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":26229,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"OnlineFigure4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/68684591fd861e25982f8f95.png\"},{\"id\":98899086,\"identity\":\"f6d965cb-8f31-4f55-8850-69cfb6fe9f57\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"png\",\"order_by\":34,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":244151,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"OnlineFigure5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/e94e575f1cfe08dc310a92d8.png\"},{\"id\":99309904,\"identity\":\"33e3b1a0-b49c-49a3-b1ee-e2913fac899c\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:21\",\"extension\":\"png\",\"order_by\":35,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":48790,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"OnlineFigure6.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/cf9c0f8b387b29447daa5eab.png\"},{\"id\":99309903,\"identity\":\"87e2b6cd-5db5-469f-9e2f-55402b890be4\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:21\",\"extension\":\"png\",\"order_by\":36,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":304643,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"OnlineFigure7.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/8e7c3f6f5e3eaa235cdc15a9.png\"},{\"id\":98899089,\"identity\":\"badaf2d2-de44-4f81-a100-b82942a5ff21\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"png\",\"order_by\":37,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":84592,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"OnlineFigure8.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/3eeaa45634baf411324435cd.png\"},{\"id\":99309733,\"identity\":\"97f84944-019b-4c28-a3d6-749a8ca8941d\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:02\",\"extension\":\"png\",\"order_by\":38,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":93017,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"OnlineFigure9.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/f6dd2d80ed8437801b9035ca.png\"},{\"id\":99309967,\"identity\":\"3a37c683-7fb2-40de-9cbb-eec94c201139\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:31\",\"extension\":\"xml\",\"order_by\":39,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":123501,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"e1814ff6c5a34f51897bbcf0ada1b3b31structuring.xml\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/f12ced7872f5100af37fdc61.xml\"},{\"id\":98899087,\"identity\":\"069ea577-04af-49b6-a816-b288bc6f75b2\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:09\",\"extension\":\"html\",\"order_by\":40,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"acdc-reference\",\"size\":133352,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"earlyproof.html\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/fc09250615f34651cc5fec19.html\"},{\"id\":98899054,\"identity\":\"f5d3e1ef-81e2-43e5-aa8a-8c2a8d63d26c\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:08\",\"extension\":\"jpg\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":5601108,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eMap of the study area showing Lagos State, Nigeria. The map highlights major urban centers, coastal zones, and surrounding lagoon systems that influence surface temperature distribution.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure1.jpg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/11aa5f947e2343446fc68c23.jpg\"},{\"id\":98899048,\"identity\":\"8623055b-5287-45e3-86bc-371968c429eb\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:08\",\"extension\":\"png\",\"order_by\":2,\"title\":\"Figure 2\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":651486,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eWorkflow diagram illustrating the data processing and modeling framework. The process includes data acquisition from Landsat 8, SRTM, and ESA WorldCover datasets, preprocessing, feature integration, Random Forest model training, and spatial prediction of land surface temperature (LST).\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/200fd6960679460496f2ae46.png\"},{\"id\":99309882,\"identity\":\"944e81d7-429b-404e-8900-169c23d5a9b7\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:20\",\"extension\":\"png\",\"order_by\":3,\"title\":\"Figure 3\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":275569,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eCorrelation heatmap of predictor variables used in the Random Forest models. Positive correlations are shown in blue and negative correlations in red, indicating inter-variable relationships among NDVI, canopy height, built-up density, elevation, NDBI, LULC, and albedo.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/abc019245e3d8c821bfd5dfc.png\"},{\"id\":99309811,\"identity\":\"7a20ff0d-a6bd-471c-a326-d84837cd3e3d\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:12\",\"extension\":\"png\",\"order_by\":4,\"title\":\"Figure 4\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":66071,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eVariable importance of predictors used in Random Forest models for estimating Land Surface Temperature (LST) in Lagos.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/98ed52faf3efd6ef40463dfd.png\"},{\"id\":99310043,\"identity\":\"d241471e-eef3-48ba-8fb1-72f6adbd8a67\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:44\",\"extension\":\"png\",\"order_by\":5,\"title\":\"Figure 5\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1083549,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003ePredicted Land Surface Temperature (LST) in Lagos using the best-performing Random Forest model (Model 3). Red tones indicate high temperatures, and blue tones indicate cooler areas.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/46efde767a71bf9c18a68d50.png\"},{\"id\":99309997,\"identity\":\"7f8cccd9-67d3-4556-a8c0-f1fb3deddd32\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:36\",\"extension\":\"png\",\"order_by\":6,\"title\":\"Figure 6\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":165097,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eBoxplot showing the distribution of predicted Land Surface Temperature (lst_l8) across five major LULC classes in Lagos. Built-up and bare areas exhibit the highest temperatures, while vegetation and wetlands remain substantially cooler.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure6.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/9c03d0e11c39104f738942dc.png\"},{\"id\":98899070,\"identity\":\"9930e264-675c-43d1-bf0d-312b8ad3cd32\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:08\",\"extension\":\"png\",\"order_by\":7,\"title\":\"Figure 7\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1300612,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eResidual map for the Random Forest LST model in Lagos. Blue indicates over-prediction, red indicates under-prediction. Model 3 shows the lowest and most spatially random residuals.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure7.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/419d6fcec6fe89e15acfdd01.png\"},{\"id\":99309753,\"identity\":\"0283ad14-3579-4e6f-9f25-c472bc751cd0\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:04\",\"extension\":\"png\",\"order_by\":8,\"title\":\"Figure 8\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":260795,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003ePartial Dependence Plots (PDPs) showing the marginal effects of NDVI, Built_pct, NDBI, and Albedo on Land Surface Temperature (LST) in Lagos.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure8.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/1cd9d72c1d9b0193e2073106.png\"},{\"id\":98899073,\"identity\":\"3ce58d29-e0da-44a5-b2e1-cfcc5175a703\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:08\",\"extension\":\"png\",\"order_by\":9,\"title\":\"Figure 9\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":331860,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eScatter plot showing the relationship between observed and predicted Land Surface Temperature (LST) for Model 3. The red dashed line represents the 1:1 reference. The tight clustering of points around this line indicates strong model agreement and predictive accuracy.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure9.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/ad4d3c807f7b901cf0812be6.png\"},{\"id\":99788363,\"identity\":\"45e70d21-7039-4780-bad0-ab7768a7db6f\",\"added_by\":\"auto\",\"created_at\":\"2026-01-08 12:46:30\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":9942745,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/b49d2c66-e521-4361-8578-72df07351c06.pdf\"},{\"id\":99309774,\"identity\":\"ba1a28ab-96fe-455b-ae21-4789a8f09bf8\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:07\",\"extension\":\"png\",\"order_by\":1,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":46712,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eFigure S1:\\u003c/strong\\u003e Correlation heatmap of predictor variables.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"FigureS1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/14df8b63b6266c6f42675fe1.png\"},{\"id\":99309760,\"identity\":\"8388362e-9645-48e9-a6e1-ad0c1aef6aa2\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:04\",\"extension\":\"png\",\"order_by\":2,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":53865,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eFigure S2:\\u003c/strong\\u003e Predicted LST map from Random Forest Model 1.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"FigureS2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/44848d1a925b96072a90446e.png\"},{\"id\":98899065,\"identity\":\"a6c43d0e-372b-4ffe-81bc-8a2432fd21c1\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:08\",\"extension\":\"png\",\"order_by\":3,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":442954,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eFigure S3:\\u003c/strong\\u003e Residual map of Random Forest Model 1.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"FigureS3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/722165b325e52335039f2bab.png\"},{\"id\":99309816,\"identity\":\"7593ca74-4f77-41c7-91ec-070811928014\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:12\",\"extension\":\"png\",\"order_by\":4,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":437177,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eFigure S4:\\u003c/strong\\u003e Residual map of Random Forest Model 2.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"FigureS4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/c67e3b56d61281642deaf6a2.png\"},{\"id\":99309831,\"identity\":\"666dc316-a3ab-4559-805b-70a3420c3e35\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:15\",\"extension\":\"tif\",\"order_by\":5,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":2738794,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eFile S1:\\u003c/strong\\u003e Predicted Land Surface Temperature (LST) raster from Model 3\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"FileS1.tif\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/c440c94e90dd961c142c6182.tif\"},{\"id\":99309824,\"identity\":\"e97250f1-0f78-49e1-ae11-5d37a77c8a9e\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:11:15\",\"extension\":\"csv\",\"order_by\":6,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":1976,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eFile S2:\\u003c/strong\\u003e Correlation matrix of predictor variables\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"FileS2.csv\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/936407ffec925f453c181723.csv\"},{\"id\":98899067,\"identity\":\"e612f231-c617-450c-8c64-fbcfbcf3c4ba\",\"added_by\":\"auto\",\"created_at\":\"2025-12-23 18:39:08\",\"extension\":\"ipynb\",\"order_by\":7,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":399620,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eFile S3:\\u003c/strong\\u003e Jupyter Notebook containing all analytical workflows, including data integration, Random Forest modeling, and validation steps (\\u003cem\\u003ebiophysical_model_for_lagos_canopy-temperature_CLEANED.ipynb\\u003c/em\\u003e).\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"FileS3.ipynb\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/aa99fe96066883efa37a7a1b.ipynb\"},{\"id\":99309560,\"identity\":\"b1b8e9a4-e15e-410f-8c36-147f3444949d\",\"added_by\":\"auto\",\"created_at\":\"2025-12-31 16:10:46\",\"extension\":\"tif\",\"order_by\":8,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":27365268,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eFile S4:\\u003c/strong\\u003e Stacked predictor raster dataset used for modeling (\\u003cem\\u003eLagos_full_stack_100m.tif\\u003c/em\\u003e).\\u003c/p\\u003e\\n\\u003cp\\u003eThe Google Earth Engine (GEE) platform was used for initial image preprocessing and data export; these scripts are not included but are fully described in Sections 2.2–2.3.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"FileS4.tif\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8034609/v1/f44457429b23fae60ea43ca9.tif\"}],\"financialInterests\":\"No competing interests reported.\",\"formattedTitle\":\"Machine Learning-Based Modeling of Land Surface Temperature in Lagos, Nigeria: Integrating Canopy Structure, Built Environment, and Surface Reflectance Variables \",\"fulltext\":[{\"header\":\"1. Introduction\",\"content\":\"\\u003cp\\u003eUrbanization is intensifying global heat exposure, especially in tropical megacities such as Lagos, Nigeria, where dense built environments, impervious surfaces, and reduced vegetation exacerbate the urban heat island (UHI) effect (Obe et al., \\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). Elevated surface temperatures contribute to increased energy demand, health risks (e.g., heat stress, respiratory problems), and challenges for vulnerable populations, particularly in informal settlements(Abou Samra, \\u003cspan citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e; Tanoori et al., \\u003cspan citationid=\\\"CR34\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). Vegetation plays a critical moderating role through evapotranspiration and shade; recent advances in remote sensing have enabled more detailed mapping of canopy structure and vegetation indices such as NDVI at increasingly fine spatial and temporal scales (Lang et al., \\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e; Nie et al., \\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003eMachine learning (ML), particularly ensemble methods such as Random Forest and Gradient Boosting, has become a powerful tool to relate remotely sensed predictors (e.g., canopy height, spectral indices, built-up area) to land surface temperature (LST), allowing spatially explicit predictions and scenario analysis (Tanoori et al., \\u003cspan citationid=\\\"CR34\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). Global datasets such as the GlobeFCH (Global Forest Canopy Height 2020, 30 m) offer vegetation structural variables over large spatial extents, and high fidelity canopy products at 10 m resolution are now available to support such analyses (Lang et al., \\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e; Nie et al., \\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003eIn the context of Lagos, prior work has documented increasing LST trends and land cover changes (Emekwuru \\u0026amp; Ejohwomu, \\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e; Obe et al., \\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). However, few studies have performed spatially explicit ML-based LST modelling that integrates canopy height, vegetation indices, and other environmental predictors at high resolution to identify thermal hotspots and prioritize urban greening zones.\\u003c/p\\u003e \\u003cp\\u003eThis research bridges that gap by developing and rigorously comparing three distinct Random Forest models designed to estimate Land Surface Temperature (LST) at a fine spatial resolution (~\\u0026thinsp;100 m) throughout Lagos. The models are structured to progressively test the predictive power of various biophysical and surface parameters. Model 1 employs an initial set of vegetation-related predictors (NDVI and canopy height). Model 2 introduces key urban and elevation characteristics by adding built-up proportion and elevation. The full-featured Model 3 incorporates a complete set of surface descriptors, including NDBI, Land-Use/Land-Cover (LULC), and albedo.\\u003c/p\\u003e \\u003cp\\u003eThe objectives are to (1) evaluate the predictive performance of each model, (2) analyze spatial variations in predicted LST, and (3) assess the relative importance of vegetation, built-up density, and surface properties in shaping urban heat patterns. We hypothesize that the inclusion of built‐up and spectral surface variables will significantly enhance model accuracy, and that NDVI, NDBI, and built‐up percentage will emerge as dominant predictors of LST variability. The results provide spatially explicit evidence for urban heat mitigation and greening strategies in Lagos.\\u003c/p\\u003e\"},{\"header\":\"2. Methodology\",\"content\":\"\\u003cdiv id=\\\"Sec3\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.1 Study Area\\u003c/h2\\u003e \\u003cp\\u003eLagos State, Nigeria\\u0026rsquo;s largest metropolitan area, is a rapidly expanding coastal megacity located along the Gulf of Guinea between 6.35\\u0026deg;\\u0026ndash;6.75\\u0026deg; N and 3.0\\u0026deg;\\u0026ndash;3.6\\u0026deg; E. The city\\u0026rsquo;s low-lying coastal plain, complex lagoon network, and limited elevation (mostly\\u0026thinsp;\\u0026lt;\\u0026thinsp;50 m above sea level) make it particularly prone to heat accumulation (Omosanya et al., \\u003cspan citationid=\\\"CR25\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e). The climate is tropical monsoon (K\\u0026ouml;ppen Am), characterized by high humidity and two main seasons: a wet season (April\\u0026ndash;October) and a dry season (November\\u0026ndash;March). Annual rainfall exceeds 1,500 mm, while mean daily temperatures range from 26\\u0026deg;C to 34\\u0026deg;C (Akinwumi et al., \\u003cspan citationid=\\\"CR2\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e; Emekwuru \\u0026amp; Ejohwomu, \\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e; Njoku et al., \\u003cspan citationid=\\\"CR21\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e; Ogunjo et al., \\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). Figure\\u0026nbsp;\\u003cspan refid=\\\"Fig1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e presents the geographic setting of Lagos State, highlighting major roads, urban centers, waterbodies, and coastal features that characterize its low-lying terrain and urban sprawl.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003eLagos\\u0026rsquo;s rapid urbanization has led to extensive conversion of vegetated land to impervious surfaces, with built-up areas increasing from ~\\u0026thinsp;23.6% in 2000 to 47.2% in 2022 (Emekwuru \\u0026amp; Ejohwomu, \\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e). This has intensified the UHI effect, producing surface temperature differences of up to ~\\u0026thinsp;5\\u0026deg;C between urban cores and peri-urban zones(Obiefuna et al., \\u003cspan citationid=\\\"CR23\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). Vegetation loss and canopy degradation have further reduced urban cooling capacity (Ayanlade, \\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e2017\\u003c/span\\u003e; Sulaiman et al., \\u003cspan citationid=\\\"CR32\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e). Recent advances such as GEDI LiDAR and Sentinel-2 data now enable fine‐scale mapping of canopy height and structure (Tsao et al., \\u003cspan citationid=\\\"CR35\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e), providing new opportunities to quantify vegetation\\u0026ndash;temperature interactions. Given these conditions, Lagos serves as an ideal case study for investigating how vegetation, surface composition, and built-up intensity jointly influence LST in tropical urban environments.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec4\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.2 Data Sources\\u003c/h2\\u003e \\u003cp\\u003eThis study integrated multi-source remote sensing datasets to model and map spatial patterns of land surface temperature (LST) across Lagos. The datasets included canopy height, vegetation indices, built-up area, elevation, surface reflectance, and derived spectral indices. The dependent and predictor variables were derived from multiple Earth observation sources, including Landsat 8, GlobeFCH 2020, ESA WorldCover, and SRTM DEM. LST was derived from Landsat 8 TIRS Band 10 (Collection 2, Tier 1) after atmospheric correction and conversion from Kelvin to Celsius following USGS guidelines(Barsi et al., \\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e2014\\u003c/span\\u003e). Vegetation structure and greenness were represented by canopy height GlobeFCH 2020 (Nie et al., \\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e) and the Normalized Difference Vegetation Index (NDVI = (NIR \\u0026ndash; RED)/(NIR\\u0026thinsp;+\\u0026thinsp;RED)), which captures the cooling influence of vegetated surfaces. Built-up density was quantified using the built-up percentage (Built_pct) and the Normalized Difference Built-up Index (NDBI = (SWIR \\u0026ndash; NIR)/(SWIR\\u0026thinsp;+\\u0026thinsp;NIR)), both indicators of impervious surface extent(Huang et al., \\u003cspan citationid=\\\"CR11\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). Elevation (from SRTM DEM) and surface albedo (computed from Landsat reflectance bands following (Liang, \\u003cspan citationid=\\\"CR15\\\" class=\\\"CitationRef\\\"\\u003e2001\\u003c/span\\u003e)) were included to account for topographic and radiative influences on surface temperature. Land-use/land-cover (LULC) data were obtained from ESA WorldCover 2020 (10 m) and resampled to 100 m for modelling consistency.(Zhang et al., \\u003cspan citationid=\\\"CR39\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e). A summary of all variables, their descriptions, and data sources is provided in Table\\u0026nbsp;\\u003cspan refid=\\\"Tab1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e.\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"Yes\\\" id=\\\"Tab1\\\" border=\\\"1\\\"\\u003e \\u003ccaption language=\\\"En\\\"\\u003e \\u003cdiv class=\\\"CaptionNumber\\\"\\u003eTable 1\\u003c/div\\u003e \\u003cdiv class=\\\"CaptionContent\\\"\\u003e \\u003cp\\u003eDescription and data sources of predictor variables used in the Random Forest models.\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/caption\\u003e \\u003ccolgroup cols=\\\"5\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c5\\\" colnum=\\\"5\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eVariable\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eDescription\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eUnits / Index\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eSource\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003eResolution\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLST (Target)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eLand Surface Temperature (dependent variable)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e\\u0026deg;C\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eLandsat 8 TIRS (USGS)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e100 m\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eNDVI\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eNormalized Difference Vegetation Index\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eUnitless\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eLandsat 8 OLI\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e100 m\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eCanopy Height\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eForest canopy height\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eMeters\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eGlobeFCH 2020\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e30 m resampled\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eBuilt_pct\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eBuilt-up percentage per pixel\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e%\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eESA WorldCover (2020)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e10 m resampled\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eElevation\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eDigital elevation model\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eMeters\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eSRTM DEM (NASA JPL, 2013)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e30 m resampled\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eNDBI\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eNormalized Difference Built-up Index\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eUnitless\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eLandsat 8 OLI\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e100 m\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLULC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eLand-use / land-cover class\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eClass (1\\u0026ndash;5)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eESA WorldCover (2020)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e10 m resampled\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eAlbedo\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eBroadband surface albedo\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eUnitless\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eComputed from Landsat 8 OLI\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e100 m\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec5\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.3 Data Pre-processing and Integration\\u003c/h2\\u003e \\u003cp\\u003ePre-processing was conducted using Google Earth Engine (GEE) and Python libraries (\\u003cem\\u003erasterio, numpy, geopandas, pandas\\u003c/em\\u003e). The Lagos administrative boundary was used to mask all datasets. Raster values for each predictor layer (NDVI, canopy height, Built_pct, elevation, NDBI, LULC, albedo) were extracted and stacked into a multiband GeoTIFF. Missing values were assessed: predictors with minor gaps; like NDBI, albedo were imputed using column‐wise mean substitution. A stratified random sample of approximately 10,000 points was extracted from valid pixels for modelling. The overall methodological workflow is summarized in Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003e, which outlines the data sources, preprocessing procedures, model training, and spatial prediction steps used in the Random Forest framework.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec6\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.4 Model Development and Refinement\\u003c/h2\\u003e \\u003cp\\u003eA three-tier Random Forest (RF) regression framework was implemented in Python using the \\u003cem\\u003escikit-learn\\u003c/em\\u003e library to model land surface temperature (LST) across Lagos. Each model incorporated a progressively richer set of predictors to evaluate the relative contribution of vegetation, built-up intensity, and surface properties. The three models were defined as follows: Model 1 included NDVI and canopy height; Model 2 included NDVI, canopy height, Built_pct, and elevation; while Model 3 incorporated NDVI, canopy height, Built_pct, elevation, NDBI, LULC, and albedo. The dataset was divided into an 80:20 train\\u0026ndash;test split to ensure robust evaluation and avoid overfitting.\\u003c/p\\u003e \\u003cp\\u003eHyperparameters, including the number of trees (\\u003cem\\u003en_estimators\\u003c/em\\u003e\\u0026thinsp;=\\u0026thinsp;500), maximum tree depth, and minimum samples per split, were optimized using grid search and 10-fold cross-validation. The parameter \\u003cem\\u003en_jobs = \\u0026minus;\\u0026thinsp;1\\u003c/em\\u003e enabled full parallel computation using all available CPU cores. Model performance was assessed using the coefficient of determination (R\\u0026sup2;), root mean square error (RMSE), and mean absolute error (MAE). Feature importance scores were derived from the RF\\u0026rsquo;s internal Gini importance metric to quantify the relative contribution of each predictor to LST prediction. The configuration of the three Random Forest models, including predictor composition and key hyperparameters, is summarized in Table\\u0026nbsp;\\u003cspan refid=\\\"Tab2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003e.\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"Yes\\\" id=\\\"Tab2\\\" border=\\\"1\\\"\\u003e \\u003ccaption language=\\\"En\\\"\\u003e \\u003cdiv class=\\\"CaptionNumber\\\"\\u003eTable 2\\u003c/div\\u003e \\u003cdiv class=\\\"CaptionContent\\\"\\u003e \\u003cp\\u003eRandom Forest model configurations and predictors used for land surface temperature (LST) estimation in Lagos.\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/caption\\u003e \\u003ccolgroup cols=\\\"6\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c5\\\" colnum=\\\"5\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c6\\\" colnum=\\\"6\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eModel ID\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003ePredictors Used\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eNumber of Trees (n_estimators)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eMax Depth\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003eCross-validation\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003en_jobs\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eModel 1\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eNDVI, Canopy Height\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e500\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eAuto\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e10-fold\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e-1\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eModel 2\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eNDVI, Canopy Height, Built_pct, Elevation\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e500\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eAuto\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e10-fold\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e-1\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eModel 3\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eNDVI, Canopy Height, Built_pct, Elevation, NDBI, LULC, Albedo\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e500\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eAuto\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e10-fold\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e-1\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec7\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.5 Mapping and Output Generation\\u003c/h2\\u003e \\u003cp\\u003ePredicted LST surfaces for each model were mapped in Python using matplotlib and rasterio, and further refined in QGIS 3.34. The outputs included predicted LST maps, residual maps (Observed \\u0026ndash; Predicted), and feature importance charts. All raster datasets were projected to WGS 84 (EPSG:4326) to ensure spatial compatibility and were exported in both GeoTIFF and PNG formats. Results from Models 1 and 2 were archived as supplementary figures, while Model 3 outputs were used for the main analysis.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec8\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.6 Validation and Reproducibility\\u003c/h2\\u003e \\u003cp\\u003eModel validation combined statistical metrics (R\\u0026sup2;, RMSE, and MAE), residual mapping to assess spatial bias, and scatterplots comparing observed and predicted LST values. The complete Python-based Random Forest workflow is provided in the supplementary notebook (File S3), while preliminary GEE preprocessing steps are described in the Methods section\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec9\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.7 Model Interpretability\\u003c/h2\\u003e \\u003cp\\u003eTo gain deeper insights into how individual predictors influence modelled Land Surface Temperature (LST), Partial Dependence Plots (PDPs) were generated for the best-performing model (Model 3). PDPs illustrate the marginal effect of one or more predictors on the model\\u0026rsquo;s predicted outcome while averaging over the influence of other variables (He et al., \\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e). This approach provides an interpretable visualization of nonlinear or threshold relationships between predictors and LST that may not be evident from feature importance rankings alone.\\u003c/p\\u003e \\u003cp\\u003eFour primary predictors were selected for interpretability analysis: NDVI, Built-up Percentage (Built_pct), Normalized Difference Built-up Index (NDBI), and Albedo because they represent distinct biophysical and surface-energy dimensions of the urban environment. NDVI was expected to show a negative relationship with LST, reflecting vegetation-induced cooling, while Built_pct and NDBI were hypothesized to increase LST due to higher impervious surface density. Albedo was examined for its moderating influence on surface reflectance and heat absorption. Partial Dependence Plots (PDPs) were computed using the \\u003cem\\u003ePartialDependenceDisplay\\u003c/em\\u003e function from \\u003cem\\u003escikit-learn\\u003c/em\\u003e and visualized in \\u003cem\\u003eMatplotlib\\u003c/em\\u003e with a grid resolution of 100 to ensure smooth response curves. For each selected predictor, the average predicted LST across all observations was plotted against the predictor\\u0026rsquo;s value range. The resulting plots provided a mechanistic understanding of the model\\u0026rsquo;s behavior and were later referenced in the Discussion to interpret urban heat dynamics in Lagos.\\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"3. Results\",\"content\":\"\\u003cdiv id=\\\"Sec11\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.1 Model Performance Evaluation\\u003c/h2\\u003e \\u003cp\\u003eThe three Random Forest (RF) models exhibited progressive improvements in predicting land surface temperature (LST) as additional predictors were incorporated (Table\\u0026nbsp;\\u003cspan refid=\\\"Tab3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e). Model 1, which used only vegetation-based predictors (NDVI and canopy height), achieved an R\\u0026sup2; of 0.51 and RMSE of 2.44\\u0026deg;C, indicating moderate explanatory power primarily driven by vegetation greenness and canopy structure. Introducing built-up and topographic variables in Model 2 increased predictive accuracy (R\\u0026sup2; = 0.66, RMSE\\u0026thinsp;=\\u0026thinsp;2.04\\u0026deg;C, MAE\\u0026thinsp;=\\u0026thinsp;1.43\\u0026deg;C), suggesting that structural and elevation factors considerably influence LST variability across Lagos. The full-feature Model 3, which integrated additional predictors including NDBI, LULC, and surface albedo, provided the highest performance (R\\u0026sup2; = 0.74, RMSE\\u0026thinsp;=\\u0026thinsp;1.77\\u0026deg;C, MAE\\u0026thinsp;=\\u0026thinsp;1.21\\u0026deg;C).\\u003c/p\\u003e \\u003cp\\u003eThis model was therefore selected for spatial prediction and further analysis. The sequential improvement across models highlights the value of integrating spectral and structural variables for explaining the thermal heterogeneity of Lagos\\u0026rsquo; urban landscape.\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"Yes\\\" id=\\\"Tab3\\\" border=\\\"1\\\"\\u003e \\u003ccaption language=\\\"En\\\"\\u003e \\u003cdiv class=\\\"CaptionNumber\\\"\\u003eTable 3\\u003c/div\\u003e \\u003cdiv class=\\\"CaptionContent\\\"\\u003e \\u003cp\\u003eSummary of Random Forest model performance for LST prediction in Lagos.\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/caption\\u003e \\u003ccolgroup cols=\\\"6\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c5\\\" colnum=\\\"5\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c6\\\" colnum=\\\"6\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eModel\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003ePredictor Variables\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eR\\u0026sup2;\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eRMSE (\\u0026deg;C)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003eMAE (\\u0026deg;C)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003eInterpretation\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eModel 1\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eNDVI, Canopy Height\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.5113\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e2.44\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e1.75\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003eVegetation-only predictors provide moderate explanatory power \\u0026mdash; NDVI and canopy height account for about 51% of LST variance.\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eModel 2\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eNDVI, Canopy Height, Built_pct, Elevation\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.6582\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e2.04\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e1.43\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003eAddition of built-up and topographic variables improves prediction by ~\\u0026thinsp;15%, reflecting structural and terrain influence.\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eModel 3\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eNDVI, Canopy Height, Built_pct, Elevation, NDBI, LULC, Albedo\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.7438\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e1.77\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e1.21\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003eComprehensive model incorporating spectral and surface properties yields the best performance; ~74% of LST variance explained.\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003cp\\u003eThe performance metrics obtained for all three models indicate a robust and credible predictive framework for LST estimation across Lagos. The final Random Forest model (Model 3) achieved an R\\u0026sup2; of 0.74, RMSE of 1.77\\u0026deg;C, and MAE of 1.21\\u0026deg;C, demonstrating strong predictive accuracy in a complex tropical urban environment. These results are consistent with similar Random Forest-based LST models reported in recent studies by (Z. Li et al., \\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e; Tanoori et al., \\u003cspan citationid=\\\"CR34\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e; \\u0026Uuml;nsal et al., \\u003cspan citationid=\\\"CR36\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e), where R\\u0026sup2; values typically range between 0.65 and 0.78 for heterogeneous urban landscapes. Given the inherent variability of land surface conditions and sensor-related uncertainties in remote sensing data, this level of performance reflects a well-generalized and reliable model capable of accurately reproducing spatial temperature patterns across the Lagos metropolis.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec12\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.2 Correlation Analysis of Predictor Variables\\u003c/h2\\u003e \\u003cp\\u003eA correlation heatmap was generated to examine relationships among all predictor variables and their association with land surface temperature (lst_l8) (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e). NDVI showed a strong negative correlation with lst_l8 (r \\u0026asymp; \\u0026minus;\\u0026thinsp;0.72), confirming its significant cooling influence through vegetation cover and evapotranspiration. In contrast, Built_pct and NDBI exhibited strong positive correlations (r\\u0026thinsp;\\u0026asymp;\\u0026thinsp;+\\u0026thinsp;0.64 and +\\u0026thinsp;0.68, respectively), highlighting the warming effect of impervious surfaces. Albedo and Elevation demonstrated weak-to-moderate negative correlations with lst_l8, suggesting minor cooling effects linked to reflective and elevated surfaces. These relationships emphasize the opposing roles of vegetation and built-up surfaces in shaping the city\\u0026rsquo;s thermal environment and provide a foundation for understanding variable interactions before modeling. Importantly, the moderate intercorrelations among predictors like between Built_pct and NDBI, underscore the need for algorithms like Random Forest, which can handle multicollinearity and complex nonlinear effects.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec13\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.3 Variable Importance\\u003c/h2\\u003e \\u003cp\\u003eFeature importance analysis revealed distinct shifts in predictor influence as model complexity increased (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003e). In Model 1, NDVI dominated, accounting for more than 80% of the total predictive power, indicating that vegetation greenness alone strongly influences surface cooling in the absence of built-environment variables. In Model 2, Built_pct became the most influential predictor (~\\u0026thinsp;40%), followed by NDVI (~\\u0026thinsp;30%) and Elevation (~\\u0026thinsp;20%). The growing importance of Built_pct underscores the role of impervious surfaces in controlling LST variability. Model 3 exhibited a more balanced influence distribution. Built_pct, NDVI, NDBI, and LULC contributed comparably to LST prediction, while Albedo and Elevation had moderate effects. Canopy height showed the lowest relative importance across all models. These results highlight that both vegetation cover and built-up density jointly determine spatial variations in surface temperature across Lagos.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec14\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.4 Spatial Pattern of Predicted Land Surface Temperature (LST)\\u003c/h2\\u003e \\u003cp\\u003eThe spatial distribution of predicted LST from the best-performing Random Forest model (Model 3) is presented in Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig5\\\" class=\\\"InternalRef\\\"\\u003e5\\u003c/span\\u003e, revealing clear spatial heterogeneity in surface temperature across Lagos. Higher LST values (approximately 40\\u0026ndash;45\\u0026deg;C) were concentrated in densely built-up and industrial zones such as Ikeja, Apapa, Lagos Island, Surulere, and Agege, while lower temperatures (around 20\\u0026ndash;30\\u0026deg;C) were observed in vegetated and coastal areas of Lekki, Epe, and Badagry. Intermediate temperature zones corresponded to mixed-use landscapes. Compared with Models 1 and 2, Model 3 produced more coherent and realistic LST patterns. The inclusion of NDBI, LULC, and Albedo enhanced the model\\u0026rsquo;s ability to distinguish between impervious and vegetated surfaces, resulting in a more accurate depiction of urban heat island (UHI) patterns.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec15\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.5 Land Surface Temperature Variation Across Land-Use Classes\\u003c/h2\\u003e \\u003cp\\u003eTo further interpret LST variability, predicted lst_l8 values were analyzed across five major land-use/land-cover (LULC) classes (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig6\\\" class=\\\"InternalRef\\\"\\u003e6\\u003c/span\\u003e). Distinct temperature gradients were evident: built-up and bare/transitional areas recorded the highest median LST (~\\u0026thinsp;33\\u0026ndash;35\\u0026deg;C), reflecting the dominance of impervious materials and low vegetation cover. Vegetation and wetland classes exhibited markedly lower median temperatures (~\\u0026thinsp;28\\u0026ndash;30\\u0026deg;C), consistent with the cooling effects of canopy cover and evapotranspiration. Water bodies were the coolest overall but displayed moderate variability, likely due to shallow lagoon heating and mixed-pixel effects near shorelines. These contrasts provide clear empirical evidence of the urban heat island effect in Lagos and validate the model\\u0026rsquo;s capacity to capture thermal differences among land-cover types.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003eThe findings emphasize that expanding vegetative cover can significantly mitigate surface heating in densely urbanized districts.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec16\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.6 Residual Analysis\\u003c/h2\\u003e \\u003cp\\u003eResidual maps (Observed \\u0026ndash; Predicted) were used to assess spatial bias and uncertainty in model performance (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig11\\\" class=\\\"InternalRef\\\"\\u003e7\\u003c/span\\u003e). In Model 1, residuals exhibited strong clustering, with underestimation (positive residuals) in dense urban areas and overestimation (negative residuals) in vegetated zones. Model 2 reduced this bias but still showed localized under-prediction in built-up zones. Model 3 demonstrated the lowest and most spatially uniform residuals, with errors typically within \\u0026plusmn;\\u0026thinsp;5\\u0026deg;C, indicating enhanced predictive stability and minimal bias. Residual maps for Models 1 and 2 are provided in the Supplementary Materials (Figures \\u003cspan refid=\\\"MOESM3\\\" class=\\\"InternalRef\\\"\\u003eS3\\u003c/span\\u003e\\u0026ndash;S4).\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec17\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.7 Partial Dependence Analysis\\u003c/h2\\u003e \\u003cp\\u003ePartial Dependence Plots (PDPs) were used to examine the marginal effects of key predictors on LST (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig12\\\" class=\\\"InternalRef\\\"\\u003e8\\u003c/span\\u003e). NDVI displayed a strong negative association with LST, confirming that increasing vegetation greenness reduces surface temperature through evapotranspiration and shading. Built_pct and NDBI exhibited positive nonlinear relationships, with rapid temperature increases at moderate urban densities that stabilized at higher built-up fractions.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003eAlbedo had a mild, nearly linear positive relationship with LST, suggesting that reflective materials in Lagos do not significantly mitigate heating where vegetation is absent. Collectively, the PDPs affirm that vegetation cover is the most effective biophysical control on surface temperature, while built-up intensity and surface composition drive localized heat accumulation.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec18\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.8 Predicted vs. Observed Validation\\u003c/h2\\u003e \\u003cp\\u003eThe scatter plot comparing observed and predicted LST values for Model 3 (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig13\\\" class=\\\"InternalRef\\\"\\u003e9\\u003c/span\\u003e) shows a strong linear relationship, with most points clustering along the 1:1 reference line.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003eThe Random Forest model achieved R\\u0026sup2; = 0.74, RMSE\\u0026thinsp;=\\u0026thinsp;1.77\\u0026deg;C, and MAE\\u0026thinsp;=\\u0026thinsp;1.21\\u0026deg;C, confirming robust predictive performance consistent with the official evaluation metrics reported in Table\\u0026nbsp;\\u003cspan refid=\\\"Tab3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e.\\u003c/p\\u003e \\u003cp\\u003eThe model slightly underestimates extreme high-temperature values in dense urban cores and marginally overestimates lower values in vegetated and coastal areas, reflecting the spatial heterogeneity typical of tropical megacities. However, these deviations are minor and do not affect the overall model accuracy or spatial coherence. The consistency between independent validation and model evaluation underscores the reliability and generalization capacity of the Random Forest approach. Overall, the model provides a stable and accurate spatial representation of land surface temperature in Lagos suitable for urban heat analysis and policy planning.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec19\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.7 Summary of Findings\\u003c/h2\\u003e \\u003cp\\u003eThe Random Forest models effectively captured the spatial variability of land surface temperature (LST) across Lagos, with Model 3, integrating NDVI, canopy height, built-up percentage, elevation, NDBI, LULC, and albedo, achieving the highest predictive accuracy (R\\u0026sup2; = 0.74; RMSE\\u0026thinsp;=\\u0026thinsp;1.77\\u0026deg;C; MAE\\u0026thinsp;=\\u0026thinsp;1.21\\u0026deg;C). Feature importance results indicated that Built_pct, NDVI, and NDBI were the most influential predictors, while Albedo, Elevation, and Canopy Height showed moderate contributions. Spatial predictions revealed clear temperature contrasts between densely built-up districts such as Ikeja, Apapa, Lagos Island, Surulere, and Agege, and the cooler, vegetated or coastal areas of Lekki, Epe, and Badagry. Residual and validation analyses confirmed that Model 3 produced the most spatially coherent and unbiased predictions, demonstrating that the LST distribution in Lagos is shaped by the combined influence of vegetation cover, urban density, and surface reflectance characteristics.\\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"4. Discussion\",\"content\":\"\\u003cdiv id=\\\"Sec21\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.1 Urban heat in tropical megacities\\u003c/h2\\u003e \\u003cp\\u003eThe results of our study reinforce the understanding that tropical megacities such as Lagos experience pronounced surface heating due to rapid urbanization, dense built environments, and loss of vegetation. Prior research demonstrated that urban growth in tropical climates intensifies land-surface and air temperature differentials more rapidly than in many temperate cities (Marcotullio et al., \\u003cspan citationid=\\\"CR18\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). In Lagos in particular, long‐term analysis of the UHI (urban heat island) indicates that land‐surface and near‐surface warming has been accelerating alongside built‐up expansion and vegetation decline (Bassett et al., \\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e2020\\u003c/span\\u003e). The relatively high coefficient of determination (R\\u0026sup2; = 0.74) achieved by our best-performing model demonstrates that a combined set of canopy, spectral, and built-cover variables can explain a substantial portion of the spatial variability in LST across Lagos, reflecting the dominant influence of urban form and land cover in tropical heat dynamics. Additionally, recent global studies of tropical city heat exposure highlight that even a 1\\u0026deg;C increase in surface temperature can significantly raise human heat‐stress risks and energy demand in low-income settlements (Ramsay et al., \\u003cspan citationid=\\\"CR28\\\" class=\\\"CitationRef\\\"\\u003e2021a\\u003c/span\\u003e; Zhou et al., \\u003cspan citationid=\\\"CR40\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e).\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec22\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.2 Role of vegetation and canopy structure\\u003c/h2\\u003e \\u003cp\\u003eOne of our key findings is the strong negative relationship between NDVI and canopy height and predicted LST, confirming that vegetation and canopy structure are critical for urban cooling. This aligns with studies showing that higher vegetation greenness and forest-canopy height reduce surface temperatures through shading and evapotranspiration (Alonzo et al., \\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e; Bedra \\u0026amp; Li, \\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e; Lima Alves \\u0026amp; Lopes, \\u003cspan citationid=\\\"CR16\\\" class=\\\"CitationRef\\\"\\u003e2017\\u003c/span\\u003e; Pfeifer et al., \\u003cspan citationid=\\\"CR27\\\" class=\\\"CitationRef\\\"\\u003e2019\\u003c/span\\u003e). Our partial dependence plot for NDVI showed a continuous cooling effect across the range of values, with no clear saturation observed. The boxplot results, which show lower median predicted LST values for vegetation and wetland land-cover classes (~\\u0026thinsp;28\\u0026ndash;30\\u0026deg;C) compared to built-up zones (~\\u0026thinsp;33\\u0026ndash;35\\u0026deg;C), further demonstrate how vegetation loss in Lagos increases vulnerability to elevated surface temperature. These findings support the argument that urban greening and canopy preservation should be top priority in tropical cities. Emerging research also suggests that mixed-layer vegetation (trees\\u0026thinsp;+\\u0026thinsp;shrubs\\u0026thinsp;+\\u0026thinsp;grasses) may deliver enhanced cooling when integrated into urban form (Gallay et al., \\u003cspan citationid=\\\"CR9\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e; Tan et al., \\u003cspan citationid=\\\"CR33\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e)\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec23\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.3 Built environment, imperviousness and surface reflectance\\u003c/h2\\u003e \\u003cp\\u003eIn contrast, our results show that built-up percentage (Built_pct) and the Normalized Difference Built‐up Index (NDBI) are strong positive predictors of LST. These findings mirror the notion that impervious surfaces absorb and re‐emit more radiation, reduce evaporative cooling, and trap heat in urban fabric (Lima Alves \\u0026amp; Lopes, \\u003cspan citationid=\\\"CR16\\\" class=\\\"CitationRef\\\"\\u003e2017\\u003c/span\\u003e). The marginal effect curves for Built_pct and NDBI show steep increases in LST for moderate to high values of urban cover, indicating non‐linear sensitivities: once impervious cover passes a threshold, surface heating may escalate rapidly. These results have practical importance for urban design: limiting impervious cover, ensuring pervious/vegetated surfaces, and using higher‐reflectance materials may reduce surface heating. Interestingly, the positive but moderate effect of Albedo suggests that while surface reflectance has some mitigation potential, in the Lagos context it appears secondary to vegetation and imperviousness. Similar studies suggest that reflectance and albedo changes must be coupled with vegetation interventions to yield meaningful cooling (Bedra \\u0026amp; Li, \\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e; Luo et al., \\u003cspan citationid=\\\"CR17\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e).\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec24\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.4 Spatial-temporal patterns and implications for urban planning\\u003c/h2\\u003e \\u003cp\\u003eOur spatial maps derived from Model 3 clearly delineated high-temperature \\u0026ldquo;hotspots\\u0026rdquo; in dense built-up districts (e.g., Ikeja, Surulere, Lagos Island) and cooler zones in vegetated and coastal districts like Lekki, Epe and Badagry. This pattern aligns with previously observed UHI gradients in Lagos and other tropical cities (Bassett et al., \\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e2020\\u003c/span\\u003e). The residual analysis; showing that our model achieved enhanced spatial coherence and lower bias compared to simpler models, provides confidence that the predictions are reliable for planning purposes. For urban planners and policy-makers in Lagos, these results mean that targeted interventions should focus on built-up hotspots: greening, canopy enhancement, and surface material redesign. Equally, preserving existing vegetated and coastal zones is critical to maintaining cooling capacity. Because surface temperature correlates strongly with energy demand, outdoor thermal comfort, and heat stress; especially under high humidity in tropical climates where every degree of surface temperature reduction may translate into meaningful public health and energy savings (Ramsay et al., \\u003cspan citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e2021b\\u003c/span\\u003e; Santamouris, \\u003cspan citationid=\\\"CR30\\\" class=\\\"CitationRef\\\"\\u003e2020\\u003c/span\\u003e; Ziter et al., \\u003cspan citationid=\\\"CR41\\\" class=\\\"CitationRef\\\"\\u003e2019\\u003c/span\\u003e)\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec25\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.5 Limitations and future research\\u003c/h2\\u003e \\u003cp\\u003eDespite strong model performance, there are important caveats. First, our dependent variable is land surface temperature (lst_l8) rather than air temperature; while LST is a widely accepted proxy for surface heating, human-experienced thermal impact may differ. Second, the model uses a single raster stack at 100 m resolution and dry-season imagery, meaning that seasonal or diurnal variations may not be fully captured. Future work should incorporate multi-temporal data (wet and dry seasons), diurnal cycles, and time-series of heating patterns (Naserikia et al., \\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e; Winckler et al., \\u003cspan citationid=\\\"CR37\\\" class=\\\"CitationRef\\\"\\u003e2019\\u003c/span\\u003e). Third, although vegetation indices and canopy height were included, micro-scale urban morphological variables like building height, street canyon geometry and ventilation were not explicitly integrated; the inclusion of such features could further refine predictions and urban niche-scale gradients (Y. Li et al., \\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e2020\\u003c/span\\u003e; Peng \\u0026amp; Huang, \\u003cspan citationid=\\\"CR26\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e; Yin et al., \\u003cspan citationid=\\\"CR38\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e). Finally, the observed moderate effect of albedo suggests that additional field-based data on material reflectance and thermal emissivity may enhance the understanding of surface energy dynamics (Solanki et al., \\u003cspan citationid=\\\"CR31\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e)\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec26\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.6 Policy relevance and actionable recommendations\\u003c/h2\\u003e \\u003cp\\u003eFrom a policy perspective, our evidence supports several actionable planning strategies for Lagos and comparable tropical urban centers:\\u003c/p\\u003e \\u003cp\\u003e \\u003cul\\u003e \\u003cli\\u003e \\u003cp\\u003ePrioritize tree-canopy planting and preservation in dense built-up zones, especially informal settlements and industrial estates, to maximize cooling benefits.\\u003c/p\\u003e \\u003c/li\\u003e \\u003cli\\u003e \\u003cp\\u003eLimit the expansion of impervious surfaces and promote permeable pavements and green-infrastructure corridors.\\u003c/p\\u003e \\u003c/li\\u003e \\u003cli\\u003e \\u003cp\\u003eIntegrate thermal-model outputs (as demonstrated here) into urban-greening policy to prioritize intervention zones based on high-predicted LST hotspots.\\u003c/p\\u003e \\u003c/li\\u003e \\u003cli\\u003e \\u003cp\\u003eUse LST maps and predictor-importance rankings to inform municipal bylaws on land-cover change, building-material standards, and urban-planning codes.\\u003c/p\\u003e \\u003c/li\\u003e \\u003c/ul\\u003e \\u003c/p\\u003e \\u003cp\\u003eIn summary, surface-temperature mitigation in tropical cities must prioritize structural redesign of the built environment and vegetation interventions, rather than relying solely on reflective-surface or albedo enhancements. By linking remotely-sensed datasets with machine-learning models, this study provides a transferrable framework for urban-climate adaptation in data-limited tropical environments.\\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"5. Conclusion\",\"content\":\"\\u003cp\\u003eThis study applied a Random Forest regression framework to model the spatial distribution of land surface temperature (LST) across Lagos, Nigeria, using high-resolution remote sensing data and biophysical predictors. Among the three models developed, the most comprehensive configuration (Model 3); integrating NDVI, canopy height, built-up percentage, elevation, NDBI, LULC, and albedo, achieved the highest predictive accuracy (R\\u0026sup2; = 0.74, RMSE\\u0026thinsp;=\\u0026thinsp;1.77\\u0026deg;C, MAE\\u0026thinsp;=\\u0026thinsp;1.21\\u0026deg;C). These results demonstrate that machine-learning approaches can effectively capture the spatial heterogeneity of surface heating in complex tropical urban systems.\\u003c/p\\u003e \\u003cp\\u003eVegetation greenness (NDVI) and canopy structure emerged as dominant cooling variables, underscoring the crucial role of urban greenery in mitigating surface heat. In contrast, built-up density and impervious surfaces significantly amplified LST, while albedo exerted only a secondary moderating effect. The resulting LST maps revealed pronounced urban heat island (UHI) patterns, with the highest temperatures concentrated in industrial and densely populated zones (Ikeja, Apapa, Lagos Island, Surulere) and cooler areas aligned with vegetated and coastal regions (Lekki, Epe, Badagry).\\u003c/p\\u003e \\u003cp\\u003eThese findings have strong implications for sustainable urban planning and climate adaptation in tropical megacities. Strategic canopy restoration, integration of green infrastructure, and limitation of impervious surface expansion can yield substantial cooling benefits and enhance urban resilience. The Random Forest modeling framework presented here provides a transferable, data-driven approach for mapping urban thermal environments in other data-limited tropical cities.\\u003c/p\\u003e \\u003cp\\u003eFuture studies should extend this framework to multi-temporal and multi-seasonal analyses, incorporating diurnal variation, urban morphological parameters, and field-measured air temperature data to improve model generalization and deepen understanding of surface-energy interactions. Ultimately, integrating remote sensing, machine learning, and urban design insights can guide evidence-based policies aimed at reducing heat vulnerability and improving livability in rapidly urbanizing regions such as Lagos.\\u003c/p\\u003e\"},{\"header\":\"Declarations\",\"content\":\"\\u003cp\\u003e \\u003ch2\\u003eSupplementary Materials\\u003c/h2\\u003e \\u003cp\\u003eThe supplementary materials accompanying this article include additional figures, datasets, and code resources supporting the main analysis.\\u003c/p\\u003e \\u003c/p\\u003e\\u003ch2\\u003eFunding\\u003c/h2\\u003e \\u003cp\\u003eThe authors received no financial support for the research, authorship, and/or publication of this article.\\u003c/p\\u003e\\u003ch2\\u003eAuthor Contribution\\u003c/h2\\u003e\\u003cp\\u003eConcept and design was done by A.B.C.Acquisition, analysis, or interpretation of data was by A.B.C.D.Model Development was by A.A.C. Drafted the manuscriptCritical review of the manuscript for important intellectual content was by A.B.C.D.\\u003c/p\\u003e\\u003ch2\\u003eData Availability\\u003c/h2\\u003e\\u003cp\\u003eAll data supporting the findings of this study are available within the paper and its Supplementary Information\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\n\\u003cli\\u003eAbou Samra, R. M. (2023). \\u0026ldquo;Investigating and mapping day-night urban heat island and its driving factors using Sentinel/MODIS data and Google Earth Engine. Case study: Greater Cairo, Egypt.\\u0026rdquo; \\u003cem\\u003eUrban Climate\\u003c/em\\u003e, \\u003cem\\u003e52\\u003c/em\\u003e, 101729. https://doi.org/10.1016/j.uclim.2023.101729\\u003c/li\\u003e\\n\\u003cli\\u003eAkinwumi, S. A., Ayo-Akanbi, O. A., Omotosho, T. V., \\u0026amp; Mastorakis, Nikos. E. (2023). Monthly and Seasonal Variation of Cloud Cover, Humidity and Rainfall in Lagos, Nigeria. \\u003cem\\u003eWSEAS TRANSACTIONS ON ENVIRONMENT AND DEVELOPMENT\\u003c/em\\u003e, \\u003cem\\u003e19\\u003c/em\\u003e, 1371\\u0026ndash;1379. https://doi.org/10.37394/232015.2023.19.123\\u003c/li\\u003e\\n\\u003cli\\u003eAlonzo, M., Ibsen, P. C., \\u0026amp; Locke, D. H. (2025). Urban Trees and Cooling: A Review of the Recent Literature (2018 to 2024). \\u003cem\\u003eArboriculture \\u0026amp; Urban Forestry\\u003c/em\\u003e, \\u003cem\\u003e51\\u003c/em\\u003e(5), 420\\u0026ndash;444. https://doi.org/10.48044/jauf.2025.023\\u003c/li\\u003e\\n\\u003cli\\u003eAyanlade, A. (2017). Variations in urban surface temperature: An assessment of land use change impacts over Lagos metropolis. \\u003cem\\u003eWeather\\u003c/em\\u003e, \\u003cem\\u003e72\\u003c/em\\u003e(10), 315\\u0026ndash;319. https://doi.org/10.1002/wea.2925\\u003c/li\\u003e\\n\\u003cli\\u003eBarsi, J., Schott, J., Hook, S., Raqueno, N., Markham, B., \\u0026amp; Radocinski, R. (2014). Landsat-8 Thermal Infrared Sensor (TIRS) Vicarious Radiometric Calibration. \\u003cem\\u003eRemote Sensing\\u003c/em\\u003e, \\u003cem\\u003e6\\u003c/em\\u003e(11), 11607\\u0026ndash;11626. https://doi.org/10.3390/rs61111607\\u003c/li\\u003e\\n\\u003cli\\u003eBassett, R., Young, P. J., Blair, G. S., Samreen, F., \\u0026amp; Simm, W. (2020). The Megacity Lagos and Three Decades of Urban Heat Island Growth. \\u003cem\\u003eJournal of Applied Meteorology and Climatology\\u003c/em\\u003e, \\u003cem\\u003e59\\u003c/em\\u003e(12), 2041\\u0026ndash;2055. https://doi.org/10.1175/JAMC-D-20-0059.1\\u003c/li\\u003e\\n\\u003cli\\u003eBedra, K. B., \\u0026amp; Li, J. (2025). Relevance of Ground and Wall Albedo for Outdoor Thermal Comfort in Tropical Savanna Climates: Evidence from Parametric Simulations. \\u003cem\\u003eSustainability\\u003c/em\\u003e, \\u003cem\\u003e17\\u003c/em\\u003e(14), 6303. https://doi.org/10.3390/su17146303\\u003c/li\\u003e\\n\\u003cli\\u003eEmekwuru, N., \\u0026amp; Ejohwomu, O. (2023). Temperature, Humidity and Air Pollution Relationships during a Period of Rainy and Dry Seasons in Lagos, West Africa. \\u003cem\\u003eClimate\\u003c/em\\u003e, \\u003cem\\u003e11\\u003c/em\\u003e(5), 113. https://doi.org/10.3390/cli11050113\\u003c/li\\u003e\\n\\u003cli\\u003eGallay, I., Olah, B., Murtinov\\u0026aacute;, V., \\u0026amp; Gallayov\\u0026aacute;, Z. (2023). Quantification of the Cooling Effect and Cooling Distance of Urban Green Spaces Based on Their Vegetation Structure and Size as a Basis for Management Tools for Mitigating Urban Climate. \\u003cem\\u003eSustainability\\u003c/em\\u003e, \\u003cem\\u003e15\\u003c/em\\u003e(4), 3705. https://doi.org/10.3390/su15043705\\u003c/li\\u003e\\n\\u003cli\\u003eHe, Z., Yang, Y., Fang, R., Zhou, S., Zhao, W., Bai, Y., Li, J., \\u0026amp; Wang, B. (2023). Integration of shapley additive explanations with random forest model for quantitative precipitation estimation of mesoscale convective systems. \\u003cem\\u003eFrontiers in Environmental Science\\u003c/em\\u003e, \\u003cem\\u003e10\\u003c/em\\u003e, 1057081. https://doi.org/10.3389/fenvs.2022.1057081\\u003c/li\\u003e\\n\\u003cli\\u003eHuang, S., Tang, L., Hupy, J. P., Wang, Y., \\u0026amp; Shao, G. (2021). A commentary review on the use of normalized difference vegetation index (NDVI) in the era of popular remote sensing. \\u003cem\\u003eJournal of Forestry Research\\u003c/em\\u003e, \\u003cem\\u003e32\\u003c/em\\u003e(1), 1\\u0026ndash;6. https://doi.org/10.1007/s11676-020-01155-1\\u003c/li\\u003e\\n\\u003cli\\u003eLang, N., Jetz, W., Schindler, K., \\u0026amp; Wegner, J. D. (2023). A high-resolution canopy height model of the Earth. \\u003cem\\u003eNature Ecology \\u0026amp; Evolution\\u003c/em\\u003e, \\u003cem\\u003e7\\u003c/em\\u003e(11), 1778\\u0026ndash;1789. https://doi.org/10.1038/s41559-023-02206-6\\u003c/li\\u003e\\n\\u003cli\\u003eLi, Y., Schubert, S., Kropp, J. P., \\u0026amp; Rybski, D. (2020). On the influence of density and morphology on the Urban Heat Island intensity. \\u003cem\\u003eNature Communications\\u003c/em\\u003e, \\u003cem\\u003e11\\u003c/em\\u003e(1), 2647. https://doi.org/10.1038/s41467-020-16461-9\\u003c/li\\u003e\\n\\u003cli\\u003eLi, Z., Wu, H., Duan, S., Zhao, W., Ren, H., Liu, X., Leng, P., Tang, R., Ye, X., Zhu, J., Sun, Y., Si, M., Liu, M., Li, J., Zhang, X., Shang, G., Tang, B., Yan, G., \\u0026amp; Zhou, C. (2023). Satellite Remote Sensing of Global Land Surface Temperature: Definition, Methods, Products, and Applications. \\u003cem\\u003eReviews of Geophysics\\u003c/em\\u003e, \\u003cem\\u003e61\\u003c/em\\u003e(1), e2022RG000777. https://doi.org/10.1029/2022RG000777\\u003c/li\\u003e\\n\\u003cli\\u003eLiang, S. (2001). Narrowband to broadband conversions of land surface albedo I. \\u003cem\\u003eRemote Sensing of Environment\\u003c/em\\u003e, \\u003cem\\u003e76\\u003c/em\\u003e(2), 213\\u0026ndash;238. https://doi.org/10.1016/S0034-4257(00)00205-4\\u003c/li\\u003e\\n\\u003cli\\u003eLima Alves, E., \\u0026amp; Lopes, A. (2017). The Urban Heat Island Effect and the Role of Vegetation to Address the Negative Impacts of Local Climate Changes in a Small Brazilian City. \\u003cem\\u003eAtmosphere\\u003c/em\\u003e, \\u003cem\\u003e8\\u003c/em\\u003e(2), 18. https://doi.org/10.3390/atmos8020018\\u003c/li\\u003e\\n\\u003cli\\u003eLuo, Y., Yang, J., Shi, Q., Xu, Y., Menenti, M., \\u0026amp; Wong, M. S. (2023). Seasonal Cooling Effect of Vegetation and Albedo Applied to the LCZ Classification of Three Chinese Megacities. \\u003cem\\u003eRemote Sensing\\u003c/em\\u003e, \\u003cem\\u003e15\\u003c/em\\u003e(23), 5478. https://doi.org/10.3390/rs15235478\\u003c/li\\u003e\\n\\u003cli\\u003eMarcotullio, P. J., Ke\\u0026szlig;ler, C., Quintero Gonzalez, R., \\u0026amp; Schmeltz, M. (2021). Urban Growth and Heat in Tropical Climates. \\u003cem\\u003eFrontiers in Ecology and Evolution\\u003c/em\\u003e, \\u003cem\\u003e9\\u003c/em\\u003e, 616626. https://doi.org/10.3389/fevo.2021.616626\\u003c/li\\u003e\\n\\u003cli\\u003eNaserikia, M., Hart, M. A., Nazarian, N., Bechtel, B., Lipson, M., \\u0026amp; Nice, K. A. (2023). Land surface and air temperature dynamics: The role of urban form and seasonality. \\u003cem\\u003eScience of The Total Environment\\u003c/em\\u003e, \\u003cem\\u003e905\\u003c/em\\u003e, 167306. https://doi.org/10.1016/j.scitotenv.2023.167306\\u003c/li\\u003e\\n\\u003cli\\u003eNie, S., Zhu, X., Wang, C., \\u0026amp; Xi, X. (2023). \\u003cem\\u003eThe global 30-m forest canopy height map for 2020\\u003c/em\\u003e (Version v1) [Dataset]. Zenodo. https://doi.org/10.5281/ZENODO.7643402\\u003c/li\\u003e\\n\\u003cli\\u003eNjoku, C. A., Ojeh, V. N., Naabil, E., Ekwe, M. C., Joshua, M. M., Elijah, E., \\u0026amp; Malgwi, M. W. (2023). Analysis of Rainfall Trend and Variability in Lagos, Southwestern Nigeria. In J. C. Egbueri, J. O. Ighalo, \\u0026amp; C. B. Pande (Eds.), \\u003cem\\u003eClimate Change Impacts on Nigeria\\u003c/em\\u003e (pp. 115\\u0026ndash;134). Springer International Publishing. https://doi.org/10.1007/978-3-031-21007-5_7\\u003c/li\\u003e\\n\\u003cli\\u003eObe, O. B., Morakinyo, T. E., \\u0026amp; Mills, G. (2024). An assessment of WRF-urban schemes in simulating local meteorology for heat stress analysis in a tropical sub-Saharan African city, Lagos, Nigeria. \\u003cem\\u003eInternational Journal of Biometeorology\\u003c/em\\u003e, \\u003cem\\u003e68\\u003c/em\\u003e(5), 811\\u0026ndash;828. https://doi.org/10.1007/s00484-024-02627-3\\u003c/li\\u003e\\n\\u003cli\\u003eObiefuna, J. N., Okolie, C. J., Nwilo, P. C., Daramola, O. E., \\u0026amp; Isiofia, L. C. (2021). Potential Influence of Urban Sprawl and Changing Land Surface Temperature on Outdoor Thermal Comfort in Lagos State, Nigeria. \\u003cem\\u003eQuaestiones Geographicae\\u003c/em\\u003e, \\u003cem\\u003e40\\u003c/em\\u003e(1), 5\\u0026ndash;23. https://doi.org/10.2478/quageo-2021-0001\\u003c/li\\u003e\\n\\u003cli\\u003eOgunjo, S. T., Akinsusi, J. O., \\u0026amp; Fuwape, I. A. (2021). Trends in extreme temperature indices over Lagos, Nigeria. \\u003cem\\u003eIOP Conference Series: Earth and Environmental Science\\u003c/em\\u003e, \\u003cem\\u003e655\\u003c/em\\u003e(1), 012003. https://doi.org/10.1088/1755-1315/655/1/012003\\u003c/li\\u003e\\n\\u003cli\\u003eOmosanya, K. O. L., Oseni, R., Lawal, M. A., Bankole, A., \\u0026amp; Mosuro, G. O. (2025). Evolving dynamics of lagoons in a megacity: Insights on climate and urbanization over four decades. \\u003cem\\u003eUrban Climate\\u003c/em\\u003e, \\u003cem\\u003e61\\u003c/em\\u003e, 102439. https://doi.org/10.1016/j.uclim.2025.102439\\u003c/li\\u003e\\n\\u003cli\\u003ePeng, M., \\u0026amp; Huang, H. (2022). The Synergistic Effect of Urban Canyon Geometries and Greenery on Outdoor Thermal Comfort in Humid Subtropical Climates. \\u003cem\\u003eFrontiers in Environmental Science\\u003c/em\\u003e, \\u003cem\\u003e10\\u003c/em\\u003e, 851810. https://doi.org/10.3389/fenvs.2022.851810\\u003c/li\\u003e\\n\\u003cli\\u003ePfeifer, M., Boyle, M. J. W., Dunning, S., \\u0026amp; Olivier, P. I. (2019). Forest floor temperature and greenness link significantly to canopy attributes in South Africa\\u0026rsquo;s fragmented coastal forests. \\u003cem\\u003ePeerJ\\u003c/em\\u003e, \\u003cem\\u003e7\\u003c/em\\u003e, e6190. https://doi.org/10.7717/peerj.6190\\u003c/li\\u003e\\n\\u003cli\\u003eRamsay, E. E., Fleming, G. M., Faber, P. A., Barker, S. F., Sweeney, R., Taruc, R. R., Chown, S. L., \\u0026amp; Duffy, G. A. (2021a). Chronic heat stress in tropical urban informal settlements. \\u003cem\\u003eiScience\\u003c/em\\u003e, \\u003cem\\u003e24\\u003c/em\\u003e(11), 103248. https://doi.org/10.1016/j.isci.2021.103248\\u003c/li\\u003e\\n\\u003cli\\u003eRamsay, E. E., Fleming, G. M., Faber, P. A., Barker, S. F., Sweeney, R., Taruc, R. R., Chown, S. L., \\u0026amp; Duffy, G. A. (2021b). Chronic heat stress in tropical urban informal settlements. \\u003cem\\u003eiScience\\u003c/em\\u003e, \\u003cem\\u003e24\\u003c/em\\u003e(11), 103248. https://doi.org/10.1016/j.isci.2021.103248\\u003c/li\\u003e\\n\\u003cli\\u003eSantamouris, M. (2020). Recent progress on urban overheating and heat island research. Integrated assessment of the energy, environmental, vulnerability and health impact. Synergies with the global climate change. \\u003cem\\u003eEnergy and Buildings\\u003c/em\\u003e, \\u003cem\\u003e207\\u003c/em\\u003e, 109482. https://doi.org/10.1016/j.enbuild.2019.109482\\u003c/li\\u003e\\n\\u003cli\\u003eSolanki, P., Amirtham, L. R., \\u0026amp; Deb, C. (2025). Optimizing sky view factor and vegetation to mitigate urban heat in hot-humid climates. \\u003cem\\u003eDiscover Cities\\u003c/em\\u003e, \\u003cem\\u003e2\\u003c/em\\u003e(1), 79. https://doi.org/10.1007/s44327-025-00108-9\\u003c/li\\u003e\\n\\u003cli\\u003eSulaiman, K., Mohd Haniffah, M. R., \\u0026amp; Shahid, S. (2025). Spatiotemporal Analysis of Urban Heat Island Effects in Nigerian Cities Using MODIS Data. \\u003cem\\u003eJournal of Advanced Geospatial Science \\u0026amp; Technology\\u003c/em\\u003e, \\u003cem\\u003e5\\u003c/em\\u003e(2), 168\\u0026ndash;229. https://doi.org/10.11113/jagst.v5n2.112\\u003c/li\\u003e\\n\\u003cli\\u003eTan, X., Liao, J., Bedra, K. B., \\u0026amp; Li, J. (2022). Evaluating the 3D cooling performances of different vegetation combinations in the urban area. \\u003cem\\u003eJournal of Asian Architecture and Building Engineering\\u003c/em\\u003e, \\u003cem\\u003e21\\u003c/em\\u003e(3), 1124\\u0026ndash;1136. https://doi.org/10.1080/13467581.2021.1903905\\u003c/li\\u003e\\n\\u003cli\\u003eTanoori, G., Soltani, A., \\u0026amp; Modiri, A. (2024). Machine Learning for Urban Heat Island (UHI) Analysis: Predicting Land Surface Temperature (LST) in Urban Environments. \\u003cem\\u003eUrban Climate\\u003c/em\\u003e, \\u003cem\\u003e55\\u003c/em\\u003e, 101962. https://doi.org/10.1016/j.uclim.2024.101962\\u003c/li\\u003e\\n\\u003cli\\u003eTsao, A., Nzewi, I., Jayeoba, A., Ayogu, U., \\u0026amp; Lobell, D. B. (2023). Canopy Height Mapping for Plantations in Nigeria Using GEDI, Landsat, and Sentinel-2. \\u003cem\\u003eRemote Sensing\\u003c/em\\u003e, \\u003cem\\u003e15\\u003c/em\\u003e(21), 5162. https://doi.org/10.3390/rs15215162\\u003c/li\\u003e\\n\\u003cli\\u003e\\u0026Uuml;nsal, \\u0026Ouml;., Lotfata, A., \\u0026amp; Avcı, S. (2023). Exploring the Relationships between Land Surface Temperature and Its Influencing Determinants Using Local Spatial Modeling. \\u003cem\\u003eSustainability\\u003c/em\\u003e, \\u003cem\\u003e15\\u003c/em\\u003e(15), 11594. https://doi.org/10.3390/su151511594\\u003c/li\\u003e\\n\\u003cli\\u003eWinckler, J., Reick, C. H., Luyssaert, S., Cescatti, A., Stoy, P. C., Lejeune, Q., Raddatz, T., Chlond, A., Heidkamp, M., \\u0026amp; Pongratz, J. (2019). Different response of surface temperature and air temperature to deforestation in climate models. \\u003cem\\u003eEarth System Dynamics\\u003c/em\\u003e, \\u003cem\\u003e10\\u003c/em\\u003e(3), 473\\u0026ndash;484. https://doi.org/10.5194/esd-10-473-2019\\u003c/li\\u003e\\n\\u003cli\\u003eYin, S., Liu, J., \\u0026amp; Han, Z. (2022). Relationship between urban morphology and land surface temperature\\u0026mdash;A case study of Nanjing City. \\u003cem\\u003ePLOS ONE\\u003c/em\\u003e, \\u003cem\\u003e17\\u003c/em\\u003e(2), e0260205. https://doi.org/10.1371/journal.pone.0260205\\u003c/li\\u003e\\n\\u003cli\\u003eZhang, N., Nex, F., Kerle, N., \\u0026amp; Vosselman, G. (2022). LISU: Low-light indoor scene understanding with joint learning of reflectance restoration. \\u003cem\\u003eISPRS Journal of Photogrammetry and Remote Sensing\\u003c/em\\u003e, \\u003cem\\u003e183\\u003c/em\\u003e, 470\\u0026ndash;481. https://doi.org/10.1016/j.isprsjprs.2021.11.010\\u003c/li\\u003e\\n\\u003cli\\u003eZhou, S., Liu, D., Zhu, M., Tang, W., Chi, Q., Ye, S., Xu, S., \\u0026amp; Cui, Y. (2022). Temporal and Spatial Variation of Land Surface Temperature and Its Driving Factors in Zhengzhou City in China from 2005 to 2020. \\u003cem\\u003eRemote Sensing\\u003c/em\\u003e, \\u003cem\\u003e14\\u003c/em\\u003e(17), 4281. https://doi.org/10.3390/rs14174281\\u003c/li\\u003e\\n\\u003cli\\u003eZiter, C. D., Pedersen, E. J., Kucharik, C. J., \\u0026amp; Turner, M. G. (2019). Scale-dependent interactions between tree canopy cover and impervious surfaces reduce daytime urban heat during summer. \\u003cem\\u003eProceedings of the National Academy of Sciences\\u003c/em\\u003e, \\u003cem\\u003e116\\u003c/em\\u003e(15), 7575\\u0026ndash;7580. https://doi.org/10.1073/pnas.1817561116\\u003c/li\\u003e\\n\\u003c/ol\\u003e\"}],\"fulltextSource\":\"\",\"fullText\":\"\",\"funders\":[],\"hasAdminPriorityOnWorkflow\":false,\"hasManuscriptDocX\":true,\"hasOptedInToPreprint\":true,\"hasPassedJournalQc\":\"\",\"hasAnyPriority\":false,\"hideJournal\":false,\"highlight\":\"\",\"institution\":\"\",\"isAcceptedByJournal\":false,\"isAuthorSuppliedPdf\":false,\"isDeskRejected\":\"\",\"isHiddenFromSearch\":false,\"isInQc\":false,\"isInWorkflow\":false,\"isPdf\":false,\"isPdfUpToDate\":true,\"isWithdrawnOrRetracted\":false,\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"modeling-earth-systems-and-environment\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":false,\"externalIdentity\":\"mese\",\"sideBox\":\"Learn more about [Modeling Earth Systems and Environment](http://link.springer.com/journal/40808)\",\"snPcode\":\"40808\",\"submissionUrl\":\"https://submission.springernature.com/new-submission/40808/3\",\"title\":\"Modeling Earth Systems and Environment\",\"twitterHandle\":\"\",\"acdcEnabled\":true,\"dfaEnabled\":true,\"editorialSystem\":\"stoa\",\"reportingPortfolio\":\"Springer Hybrid\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":false},\"keywords\":\"Land Surface Temperature, Random Forest, Urban Heat Island, NDVI, Canopy Height, Lagos, Remote Sensing\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-8034609/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-8034609/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003cp\\u003eUrban heat is an escalating environmental challenge in tropical megacities, where rapid urbanization and declining vegetation cover intensify surface warming. This study applies a machine learning approach to predict and map Land Surface Temperature (LST) across Lagos, Nigeria, by integrating multisource remote sensing variables within a Random Forest (RF) framework. Three models of increasing complexity were developed using combinations of vegetation, structural, and spectral predictors derived from Landsat 8, GlobeFCH canopy height, ESA WorldCover, and SRTM data. Model 1, using vegetation variables (NDVI and canopy height), achieved an R\\u0026sup2; of 0.51, while Model 2, incorporating built-up and elevation variables, improved performance to R\\u0026sup2; = 0.66. The final model (Model 3), combining NDVI, canopy height, built-up percentage, elevation, NDBI, LULC, and albedo, achieved the best accuracy (R\\u0026sup2; = 0.74; RMSE\\u0026thinsp;=\\u0026thinsp;1.77\\u0026deg;C; MAE\\u0026thinsp;=\\u0026thinsp;1.21\\u0026deg;C). Partial dependence analysis revealed that NDVI and canopy height exert strong cooling effects, whereas NDBI and albedo were positively associated with surface warming. Spatial predictions highlighted pronounced thermal gradients, with high LST values concentrated in industrial and densely built-up areas like Ikeja, Apapa, Lagos Island; and cooler conditions in vegetated and coastal zones. These findings underscore the role of vegetation structure in mitigating urban heat and provide actionable spatial insights for urban greening and climate adaptation planning. The reproducible workflow demonstrates the potential of machine learning and Earth observation data for urban climate monitoring in data-limited tropical regions.\\u003c/p\\u003e\",\"manuscriptTitle\":\"Machine Learning-Based Modeling of Land Surface Temperature in Lagos, Nigeria: Integrating Canopy Structure, Built Environment, and Surface Reflectance Variables\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2025-12-23 18:39:03\",\"doi\":\"10.21203/rs.3.rs-8034609/v1\",\"editorialEvents\":[{\"type\":\"communityComments\",\"content\":0},{\"type\":\"reviewersInvited\",\"content\":\"\",\"date\":\"2025-12-19T19:54:21+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"editorAssigned\",\"content\":\"\",\"date\":\"2025-11-08T14:23:21+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"checksComplete\",\"content\":\"\",\"date\":\"2025-11-08T14:22:51+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"submitted\",\"content\":\"Modeling Earth Systems and Environment\",\"date\":\"2025-11-05T05:52:32+00:00\",\"index\":\"\",\"fulltext\":\"\"}],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"modeling-earth-systems-and-environment\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":false,\"externalIdentity\":\"mese\",\"sideBox\":\"Learn more about [Modeling Earth Systems and Environment](http://link.springer.com/journal/40808)\",\"snPcode\":\"40808\",\"submissionUrl\":\"https://submission.springernature.com/new-submission/40808/3\",\"title\":\"Modeling Earth Systems and Environment\",\"twitterHandle\":\"\",\"acdcEnabled\":true,\"dfaEnabled\":true,\"editorialSystem\":\"stoa\",\"reportingPortfolio\":\"Springer Hybrid\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":false}}],\"origin\":\"\",\"ownerIdentity\":\"e75215af-08e4-46ad-82d9-92e2a3f8cdb7\",\"owner\":[],\"postedDate\":\"December 23rd, 2025\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"under-review\",\"subjectAreas\":[],\"tags\":[],\"updatedAt\":\"2025-12-23T18:39:03+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2025-12-23 18:39:03\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-8034609\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-8034609\",\"identity\":\"rs-8034609\",\"version\":[\"v1\"]},\"buildId\":\"8U1c8b4HqxoKbykW_rLl7\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}