Evaluating diagnostic horizons as indicators of ecosystem services through digital soil mapping in urban Gauteng, South Africa

Evaluating diagnostic horizons as indicators of ecosystem services through digital soil mapping in urban Gauteng, South Africa | 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 Evaluating diagnostic horizons as indicators of ecosystem services through digital soil mapping in urban Gauteng, South Africa Trevan Flynn This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8006960/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Urban areas are strongly influenced by the ecosystem services provided by soil, which help mitigate environmental pressures generated by metropolises themselves and protect surrounding natural and agricultural landscapes. However, it remains unclear whether existing soil classification systems adequately capture the complex multifunctionality of soil that produce these services. This study examined the spatial patterns and ecosystem services represented in a numerical and South Africa’s taxonomic (“Blue Book”) soil classification system in Gauteng, South Africa (~ 26.3° S, 28.1° E), a region characterised by extensive urban development. A Gradient Tree Boosting model implemented in Google Earth Engine was used to predict topsoil and subsoil horizons for both systems using identical soil observations. The model achieved high spatial accuracy for topsoil and moderate accuracy for subsoil horizons in both systems. Despite this, the topsoil predictions of both systems (accuracy > 85%) showed limited relevance to ecosystem services across the urban landscape. In contrast, the subsoil predictions (accuracy ~ 54–68%) of the taxonomic system exhibited clear spatial patterns, greater interpretability and stronger ecological relevance. The primary limitation identified was the over-classification of the Ochric topsoil together with the moderate accuracy of the subsoil, obscured functional differences and constrained the interpretation of soil-based ecosystem services across Gauteng. Nevertheless, emerging numerical frameworks that integrate categorical data and dynamic processes, supported by advances in Earth observation and machine learning, are expected to enhance the ability to map soil multifunctionality and its representation of ecosystem services. Agronomy Digital soil mapping Soil services Multifunctionality Urban soils Figures Figure 1 Figure 2 Figure 3 Figure 4 Highlights • Urban soils mediate critical ecosystem services across rapidly expanding African cities. • A numerical and taxonomic soil classification system revealed distinct functional strengths. • Taxonomic subsoil horizons better captured ecosystem service potential and soil multifunctionality. • Classification should reflect the multidimensionality of soil and not only the vertical distribution. 1. Introduction Metropolitan expansion, which is increasingly common across Africa (Awumbila, 2017 ), places substantial pressure on the environment and shapes how populations grow, directly influencing the quality of life within and around expanding urban areas (Bhatta, 2010 ). Urban soils play a critical role in providing ecosystem services: they support stormwater management (Hossain Anni et al., 2020 ), serve as a medium for urban horticulture that enhances soil health (Wu and Congreves, 2024 ), help to mitigate urban heat (Li and Wang, 2021 ), and supply essential materials for construction (Reddi et al., 2012 ), culture, and recreation (O’Riordan et al., 2021 ). Furthermore, industrial and mining wastes, along with other pollutants, often leach into surrounding environments, affecting not only urban soils but also adjacent natural and agricultural lands (Karthika et al., 2022 ). The ecosystem services that soil provides arise from the underlying soil functions expressed through their physical, chemical, and biological properties. Consequently, digital soil mapping often focuses on producing detailed and accurate maps of these soil properties. Such studies commonly employ approaches to assess soil sealing (Villa et al., 2018 ), carbon stocks (Fiorentino et al., 2025 ; Villa et al., 2018 ), bare-soil exposure (Liu et al., 2022 ), and pollution loads and transport (Magiera et al., 2007 ; Shi et al., 2021 ). However, the spatial patterns of soil properties and processes are often difficult to visualize and conceptualize, and mapping multiple soil properties fails to capture the true multifunctional behavior of soil in a form that is easily interpretable. Due to this inherent complexity, a disconnect often arises between the ecosystem services that soil provides and how they are perceived or utilised within a metropolis, emphasising the need for classification systems (Konovalov and Krajbich, 2018 ). Classification frameworks help translate the continuous variability of soil properties into discrete, interpretable categories that encapsulate multiple functions simultaneously. Unlike single-property maps, classified soil maps can support broader applications, including land-use planning, environmental assessment and ecosystem service evaluation as they encapsulate multiple soil attributes into one. However, the process of classification can also lead to the loss of detail and spatial heterogeneity (Zhu, 1997 ), potentially over simplifying the true nature of soil. To address these challenges, various numerical classification systems have emerged at both local and national scales (Bezdek et al., 1984 ; Carré and Jacobson, 2009 ; Flynn et al., 2021 ; Hartigan, 1975 ; Triantafilis et al., 2001 ; Webster and Burrough, 1972 ; Zhang and Hartemink, 2019 ), offering data-driven approaches that classify soil more or less, objectively. The main advantage of numerical systems lies in their quantitative and less subjective nature, while traditional taxonomic systems benefit from hierarchical structure and interpretability, facilitating communication among scientists, policymakers and practitioners (Finisdore et al., 2020 ). Despite their differences, both approaches provide valuable perspectives for understanding and mapping ecosystem services, forming a complementary foundation for soil-informed decision-making that supports environmental and societal wellbeing. The aim of this study was to evaluate how effectively two distinct soil classification systems represent the ecosystem services that soil provides within and around the metropolitan areas of Gauteng Province, South Africa (ZA). It is hypothesized that both the numeric and ZA’s taxonomic classification systems would capture meaningful spatial patterns of ecosystem services within the subsoil, where long-term processes occur, but not within the topsoil, which is more influenced by anthropogenic disturbance. The level of detail, accuracy and interpretability of the resulting maps is crucial for disseminating soil information, guiding sustainable urban planning, and identifying opportunities for improving classification systems. Although traditional classification systems are largely vertical in conception, soil functions as a four-dimensional living system—integrating space, depth, time, and function—and modern classification must evolve to reflect this inherent complexity. 2. Materials and Methods 2.1 Site Gauteng Province (Figure 1) is located in the high-altitude grasslands of the Highveld in central South Africa, centered approximately at 26.3° S and 28.1° E. The province is landlocked, bordered by the Free State to the south, North West to the west, Mpumalanga to the east, and Limpopo to the north. Despite being the smallest province by area, Gauteng is the most densely populated region in the country, with an average population density exceeding 715 people km⁻² (Statistics South Africa, 2016). It encompasses two major metropolitan areas of Johannesburg and Pretoria along with their suburbs, and rapidly expanding cities such as Soshanguve and Vereeniging. Gauteng was selected as the study area because it represents a diverse and dynamic landscape that includes extensive urban areas, productive arable lands and natural reserves making it ideal for examining the spatial relationships between soil functions, ecosystem services and metropolitan expansion. Gauteng has an average elevation of approximately 1,500 m, a mean annual temperature of about 20 °C, and a mean annual precipitation of around 650 mm. Sitting on the Kaapvaal Craton, the province is well known for the Witwatersrand region, which has a long history of gold, diamond and silver mining and remains an important center of industrial activity and economic growth (Abiye et al., 2011). Although Gauteng’s economy is largely driven by the mining and financial sectors, it also supports a productive agricultural industry that includes livestock, vegetables, maize, cotton and sorghum (Nesamvuni et al., 2016). 2.2 Soil data Soil profile data w as obtained from the South African Land Type Survey (LTS; Land Type Survey Staff, 2006), comprising 565 observations across Gauteng (Figure 1b). These same observations were used to evaluate both the numerical and taxonomic classification systems. The numerical system was previously developed by Flynn et al. (2021), who clustered soil properties into five topsoil and nine subsoil classes (Table 1) using a clustering algorithm (Kaufman and Rousseeuw, 1960). The algorithm was selected for its robustness to outliers and computational efficiency with large datasets, producing medoid-based (median) clusters representative of soil property distributions. As the only numerical classification system derived from national-scale LTS data, and one that clusters on master horizons, it provides a suitable basis for comparison with the South African taxonomic system (Soil Classification Working Group, 1991) at the province scale. Although not the most recent system, the South African “Blue Book” (Soil Classification Working Group, 1991-2006) soil classification was used as the ZA taxonomic system in this study (Table 2), as it had already been fully incorporated into the LTS at the time of the research. For interpretability, the ZA taxonomic names were translated into the corresponding United States Department of Agriculture (USDA) Soil Taxonomy names or equivalent Latin names. In the database, Mollic, Ochric and Vertic soil were the only topsoil types identified, and the subsoil was grouped into eight categories based on morphological similarities and to account for the limited number of observations in some subsoil classes. 2.3 Covariates In total, 60 covariates (Table 3) were obtained or loaded into Google Earth Engine (GEE; Gorelick et al., 2017), representing the scorpan factors (McBratney et al., 2003). To capture the soil ( s ), parent material ( p ) and organism (o) factors, bands and indices derived from cloud-masked surface reflectance Landsat 8 OLI/TIRS (USGS, 2021) imagery at a 30 m resolution were used. For each band and index, the median value of the time series from 2013 to 2021 was computed. The s factor was also represented by iSDAsoil (Miller et al., 2021), a 30 m soil database of Africa. Climatic factors were acquired from worldclim (Fick and Hijmans, 2017), at a 90 m resolution. All bioclimatic variables (19) were used. The r factor was represented by the 30 m NASADEM (NASA JPL, 2022) with local and global derivatives. Six oblique geographic coordinates (OGC; Møller et al., 2020) were derived to represent the n factor. Oblique geographic coordinates were used instead of latitude and longitude to prevent highly correlated variables and map artifacts. 2.4 Model development The digital soil mapping framework combined machine learning and pedological reasoning to evaluate how well numerical and taxonomic systems capture the spatial organization of diagnostic horizons. (Figure 2). By applying identical observations and environmental covariates to both classification frameworks, the model isolated differences in interpretability rather than data structure. Model evaluation emphasised not only statistical performance, through overall accuracy and Cohen’s , but also the ecological and pedological realism of the resulting horizon patterns. This enabled the direct comparison between data-driven and diagnostic approaches in representing soil variability and ecosystem service potential across the landscape. 2.4.1 Model A gradient tree boosting (GTB) model is a generalized additive model in which the base learners are small decision trees. With the addition of each tree, the model iteratively corrects or improves upon the residuals of the previous ensemble (Hastie et al., 2009). Although GTB is additive in structure, it is nonlinear and stochastic, making it highly suitable and accurate for predicting soil attributes (Flynn et al., 2019). The additive nature of the model promotes regularization and stability, helping to prevent overfitting (Friedman and Hastie, 1998), while the nonlinear and stochastic components enable the model to capture complex relationships (Kuhn and Johnson, 2013), such as those between soil classes and satellite derived variables. Mathematically, the predictions of each class can be represented as: A SMILE (“Statistical Machine Intelligence and Learning Machine”; Weininger et al., 1989) implementation of GTB was used, with model parameters optimised in GEE. The model was controlled through the learning rate, tree depth, number of trees grown, and subsampling rate. Models were trained using covariates at 10, 30, 90, and 180 m resolutions, with the number of boosting iterations ranging from 100 to 1000 in increments of 100. Additionally, learning rates of 0.1, 0.01, 0.001, and 0.0001 were tested. The default subsampling rate in SMILE models is 0.7; however, values of 0.1 and 0.9 were also evaluated. 2.4.2 Evaluation To evaluate model performance, the data ( n = 565) was split into 80% training and 20% testing sets. The performance was evaluated by how well it predicted on the 20% training data. Evaluation of performance was based on and overall accuracy (%). However, the model with the highest was taken as the top performing model. In other words, the model with the highest agreement for a given soil class (Cohen, 1960). Evaluation was conducted to make sure that the interpretation of the spatial patterns and ecosystem services was plausible; therefore, only and accuracy were reported. Covariates were also evaluated on how important they were for predictions. This rank represents the percentage increase of the loss function when the covariate is permutated (Weininger et al., 1989). It is a relative importance index and gives an indication of which covariates correlate well to that soil class. As a large pool of covariates was used in the model, the validation of the model with covariates was needed to help with the assessment of the soil horizon spatial patterns. 3. Results and Discussion 3.1 Model performance Before interpreting the predictive performance and spatial patterns of the numerical and taxonomic soil classification systems, the GTB model was first evaluated to ensure reliability. This evaluation was essential to determine whether the observed spatial distributions were genuinely influenced by environmental factors. Establishing a robust level of model accuracy provided the foundation for confidently analysing soil horizon predictions and their implications for ecosystem service assessment. 3.1.1 Agreement and accuracy The numerical system achieved a greater for topsoil (0.63) and subsoil (0.54) relative to the subsoil (Table 4). Numerical topsoil achieved a substantial agreement, while the subsoil achieved moderate agreement according to Landis and Koch (1977). On the other hand, only topsoil achieved moderate agreement (0.56) for the taxonomic system and subsoil achieved fair agreement (0.40). However, the taxonomic topsoil achieved the highest accuracy (98%) followed by numerical topsoil (85%), numerical subsoil (68%) and taxonomic subsoil (54%). This was expected, as accuracy generally decreases with an increasing number of soil classes (Herold et al., 2008). Although the fair taxonomic subsoil agreement raises concerns regarding reliability, the results remain within the range commonly reported in digital soil mapping studies. They are therefore regarded as sufficient for further interpretation. These results are broadly consistent with previous findings from South Africa. For instance, Van Zijl and Le Roux (2014) achieved an accuracy of 73% when mapping seven hydrological response unit classes in Kruger National Park, while Van Zijl et al. (2013) reported 67% accuracy for six soil classes using fuzzy logic in KwaZulu-Natal Province. Similarly, Van Zijl et al. (2019) obtained an accuracy of 69% and a κ value of 0.59 when mapping five soil classes within the city of Johannesburg using a hillslope-based approach. Additionally, these findings represent an improvement over those reported by Flynn et al. (2025), where the numerical subsoil system achieved an agreement of 0.30 and an accuracy of 57% using the same covariates and model structure. In that earlier study, the algorithm had not been fully optimised, indicating that GTB performance is highly sensitive to model calibration and parameter tuning. Such optimisation can substantially enhance the predictive potential of both classification systems and has important implications for improving the taxonomic subsoil model in this study. All models performed well with a learning rate of 0.001; however, each class exhibited a different optimal resolution and number of trees. Interestingly, this trend was opposite between the two classification systems. For the numerical topsoil model, the best performance was achieved with the largest number of trees (500) and the finest spatial resolution (30 m), whereas the taxonomic topsoil model required the fewest trees (50) but performed best at the coarsest resolution (180 m). For subsoil predictions, both systems achieved optimal outcomes at a 90 m resolution, with the numerical system requiring 300 trees and the taxonomic system needing 80 trees. 3.1.2 Covariate importance With the exception of taxonomic topsoil, all soil classes showed strong correlations with multiple climatic factors (Table 5), indicating that climate is an important variable at this spatial support. However, each class responded to different climatic influences. For example, numerical topsoil was strongly correlated with isothermality (i.e., difference between nighttime and daytime temperature), mean diurnal range (i.e., difference between the maximum and minimum temperature) and temperature seasonality, suggesting that its distribution is influenced primarily by temperature variability rather than mean temperature. In contrast, numerical subsoil was most closely correlated with mean annual temperature and the mean temperature of the coldest quarter. This is a reasonable outcome, as the topsoil responds more readily to short-term environmental fluctuations, whereas the subsoil remains comparatively insulated from such variation (Horton and Wierenga, 1983; Rubio et al., 2012). The taxonomic topsoil was also the only class to rank iSDAsoil (clay and CEC) among the ten most important predictors. These soil properties were initially assumed to correlate strongly with all horizons. However, iSDAsoil data is already highly correlated with the environmental covariates used in the model. Due to GTB being a highly regularised algorithm, it may have reduced the coefficients of the iSDAsoil variables, resulting in their lower ranking for most horizons. An alternative would be to apply principal component analysis prior to model training and combining all iSDAsoil variables into its principle components. This approach could also reduce computational demand and multicollinearity among predictors. Neighborhood was represented by OGCs, which were ranked high in both classification systems. This is noteworthy, as OGCs are not commonly cited among the most influential covariates in digital soil mapping studies. Their importance likely arises from their ability to capture spatial trends in a manner similar to geostatistical methods (Møller et al., 2020). The horizons exhibit spatial autocorrelation; however, they cannot be interpolated using ordinary kriging because they represent categorical data, and such computation is impractical at this scale. Therefore, OGCs provide a powerful alternative for capturing spatial structure and enhancing model performance. 3.2 Spatial predictions A GTB model was used to predict soil horizons for both a numerical and ZA taxonomic classification systems in and around the metropolises of Gauteng. The model evaluation statistics indicated performance that was sufficiently reliable to justify a detailed examination of the spatial patterns observed and the ecosystem services inferred. Establishing this level of predictive accuracy was essential; without it, the assessment of soil derived ecosystem services and their spatial interpretation would have remained too uncertain to support meaningful conclusions. 3.2.1 Topsoil The predictions showed a similar spatial distribution for both the numerical (Figure 3a) and taxonomic systems (Figure 3b) predominantly predicting A3 (95% of area) and Ochric (98% of area) horizons, respectively. Although the accuracy was high for both systems, the maps conceivably do not give much information about ecosystem services. The lack of pedodiversity seen in the topsoil is of concern to both classification systems as they clearly are not distinguishing a key region for microbial activity. Nevertheless, there was an area between Pretoria and Soshanguve just outside the De Onderstepoort Nature Reserve as well as the far southeast near the town of Devon, where both systems predicted the same horizon other than the Ochric. In the numeric system, this was A5 and is characteristic of saline soils with a high pH, CEC and clay content. In the taxonomic system, these soils were classified as Vertic topsoil that have the same saline properties (Fey et al., 2001). Surprisingly, this Vertic topsoil zone was also predicted through categorical downscaling texture classes from SoilGrid’s (Flynn and Kostecki, 2024) The numerical system exhibited greater topsoil pedodiversity in the northeast, where classes A1, A2 and A4 were predicted. Their distribution appeared to be influenced by the region’s more diverse cumulative geology, vegetation and topography. Both A2 and A4 were also identified within the city of Johannesburg, although their distribution was limited to a few predictions. Overall, the topsoil predictions lacked substantial or meaningful spatial patterns, detail or pedological interpretability. 3.2.2 Subsoil Like the topsoil, the subsoil exhibited the greatest pedodiversity in the northeast, where all horizons of the numerical classification were predicted (Figure 4a). Horizon B1 (low clay, coarse sand) followed areas of high relief across Gauteng, occurring most abundantly in and around Johannesburg. Horizons B2 and B3 were spatially associated with B1 in the northeast but appeared more scattered in areas of lower relief. Horizon B2 (low clay, fine sand) extended southward and, to a lesser extent, occurred within Johannesburg. Horizon B3 was predicted primarily in the West Rand (western Johannesburg) and adjacent valleys. Given its occurrence in low-lying areas and its properties of high clay content, elevated CEC and mildly alkaline, this horizon likely has strong structural development characteristic of Luvic subsoil. Horizon B4 was predominantly confined to the De Onderstepoort Nature Reserve, occurring along the foothills of the quartzite mountains. These quartzite-derived soils are typically coarse-textured, acidic and nutrient-poor due to the parent material’s resistance to weathering. In contrast, horizon B5 (very high clay, high Fe) was the most frequently predicted subsoil, occupying planar surfaces across much of the Highveld. Horizon B6 (high clay, high Mg) occurred east of Pretoria and exhibited a structured distribution along terrain transects, while horizons B7 and B8 were limited to a few localized occurrences in the northeast. Horizon B9 was similarly restricted to the northeastern part of the province, mainly in and around the Ezemvelo Game Reserve. The taxonomic system showed a similar distribution (Figure 4b), which was more diverse in the northeastern areas of Gauteng. Predictions followed clear toposequences at the 90 m resolution, reflecting consistent hillslope transitions across much of the province. As with B5 in the numerical system, Rhodic horizons were predicted to dominate much of Gauteng, which were prevalent in the west, encircled Johannesburg to the south and extended northeastward across planar surfaces. This correspondence is noteworthy, as both B5 and Rhodic horizons are characterized by high Fe content, suggesting that the numerical and taxonomic systems classified analogous subsoil. Further east, Xanthic horizons appeared on comparable surfaces and extended into the Johannesburg area, although these regions were marked by more dissected topography. The distribution of the Cambic horizon closely resembled that of B1, occurring mainly in high-relief areas and within Johannesburg, but absent from low-lying valleys. As expected, these valley positions were instead occupied by Aquic horizons, which were not represented in the numerical system, highlighting an area for improvement in capturing redox morphology in the numeric system. Along the eastern foothills, typically above the elevation of Aquic zones, Luvic horizons were predicted and frequently associated with A5/Vertic topsoil horizons between Pretoria and Soshanguve. Plinthic horizons occurred intermittently near the transitional boundaries of Cambic and Rhodic/Xanthic horizons across the province. As anticipated, Hardpan and Podzolic horizons were only occasionally predicted and exhibited no consistent spatial pattern. 3.3 Ecosystem services and limitations Both the numerical and taxonomic classification systems identified a type of Ochric horizon as the predominant topsoil across Gauteng. However, this horizon encompasses a wide range of physical, chemical and biological conditions and insufficient differentiation hides important functional variation. The topsoil represents a zone of high biological activity and strong interactions with the atmosphere and biosphere, and inadequate detail limits interpretation of key soil functions such as nutrient cycling (Ding et al., 2024), carbon sequestration (Mason et al., 2023) and aggregate stability (Liu et al., 2024). This drawback is especially significant given that the topsoil forms a key connection between ecosystem regulation and human health. The A2 horizon, characterised by high fertility and SOC content, supports multiple soil functions related to crop productivity (Ma et al., 2023), erosion control (Rugendo et al., 2023) and biodiversity maintenance (Wang et al., 2023). In urban contexts, A2 may further contribute to temperature regulation and energy balance through enhanced water retention (Ramírez et al., 2023). Despite this multifunctionality, A2 topsoil was predicted only intermittently. Their under-representation likely reflects the generalisation inherent in both systems rather than their true scarcity. The A5 horizon shares similar structural and textural characteristics with A2 but is alkaline, reducing its agricultural suitability. Nonetheless, its high clay, CEC and SOC content may provide regulating services through adsorption of organic pollutants (Ewis et al., 2022), particularly along the rapidly closing urban corridor between Soshanguve and Pretoria. The failure of both classification systems to distinguish topsoil variation reflects a broader global challenge in soil classification. In other words, soil classification systems, whether South Africa’s or those used globally, tend to define topsoil by exclusion. For example, in the USDA Soil Taxonomy (Soil Survey Staff, 2014), strict diagnostic criteria govern Mollic, Melanic, Umbric, Anthropic, Plaggen and Histic epipedons, resulting in most topsoil being classified as Ochric, regardless of its distinct physical or biological characteristics. Consequently, the true multifunctionality of soil cannot be fully represented through categorical mapping as seen in Figure 3. In the present study, the numerical framework exhibited a similar limitation, mirroring the constraints of traditional taxonomic systems. This challenge extends beyond pedology, influencing how other scientific disciplines, policy frameworks and stakeholders perceive and apply soil information. Nonetheless, numerical systems offer a distinct advantage: they can be revised, optimised and evaluated far more rapidly than national taxonomic systems, providing a flexible option toward a system that better reflects ecosystem services in the topsoil. For the subsoil, the Rhodic and Xanthic horizons predominated, representing mildly acidic horizons enriched in Fe oxides. The Rhodic horizon, with a higher hematite:goethite ratio (Schwertmann, 1971), exhibits slightly stronger reactivity towards organic pollutants (Voelz et al., 2018), whereas goethite rich Xanthic horizons are more reactive to phosphorus (Colombo et al., 1991). However, texture appears to be the major differentiating factor between these two horizons in terms of ecosystem services, yet this is not explicitly recognised in the taxonomic system but is in the numerical (e.g., B1 vs. B2). Regardless, both horizons generally support high agricultural productivity and support carbon stabilisation (He et al., 2025); however, when overlain by an Ochric topsoil they may undergo bleaching (Clarke et al., 2020), which promotes surface crusting and erosion. In most circumstances, Rhodic and Xanthic horizons are suitable for metropolitan expansion and can contribute to buffering or attenuating contaminants within the urban landscape. The light texture and low Fe content of B1 subsoil suggests unstable clays with limited regulating capacity, although intensive urban horticulture, particularly in Sandton city, may locally enhance infiltration and reduce surface heat (Wu and Congreves, 2024). The balanced sand fraction of the B1 horizon also indicates potential suitability for sustainable local building materials (Aaquib et al., 2018), linking soil properties to ecosystem services beyond agriculture and the environment. From a land-use perspective, Aquic and Luvic horizons are unsuitable for urban development owing to their hydromorphic and shrink–swell behaviour. Aquic horizons indicate a shallow water table, while Luvic horizons favour lateral subsurface flow (Van Tol et al., 2011) and may cause structural damage through expansion and contraction processes, depending on their mineralogy and dominant cation species (Sabtan, 2005). These areas are therefore better reserved for maintained natural vegetation due to their inherent water cycling capacity (Basset et al., 2023), keeping their true inherent ability to perform ecosystem services. The numerical classification system misclassified many of these hydromorphic areas as B1 horizons, which are typically associated with high-relief terrain. This confirms that the numerical framework lacks diagnostic sensitivity to colour, structure and drainage attributes, a limitation previously noted by Flynn et al. (2021) and emphasized by Flynn et al. (2025). Inclusion of such morphological features would improve the system’s capability to represent hydrological functions, and perhaps more globally relevant, greenhouse emissions such as methane emissions (Le Mer and Roger, 2001). Although the taxonomic subsoil achieved only modest accuracy (54%) and agreement (κ = 0.40), its interpretive strength demonstrates that fundamental soil elements remain essential for linking pedological understanding to ecosystem services. However, modern pedometric techniques are essential to implement, interpret, visualise, and continuously refine such data-driven frameworks. For instance, emerging Earth-observation embedding models such AlphaEarth (Brown et al., 2025) and TESSERA (Feng et al., 2025) offer uncorrelated, multitemporal and high-resolution (10 m) representations that surpass the informational capacity of conventional covariates. Their integration could markedly improve predictive accuracy and ecological interpretability, particularly in heterogeneous urban landscapes where the interactions between soil processes and anthropogenic activity are subtle yet profoundly dynamic. Ultimately, this study demonstrated that while both systems effectively captured the spatial variability of soil, neither fully represented its multifunctionality. Moving beyond static or property-based numerical classifications toward function-oriented frameworks that integrate soil dynamics, hydromorphic behaviour and possibly features from taxonomic systems will enable a more explicit linkage between soil horizons and the expression of ecosystem services. Such integration bridges pedology and a multidisciplinary approach, reaffirming soil’s role not merely as a biophysical substrate but as a living, regulating system fundamental to climate resilience, agricultural productivity and human wellbeing in an increasingly urbanised world. 4. Conclusion This study evaluated a numerical and South Africa’s taxonomic soil classification for their ability to represent soil variability and ecosystem service potential within the urban expanses of Gauteng, South Africa. Both systems effectively captured spatial patterns but differed in interpretability and functional resolution. The numerical framework achieved strong quantitative accuracy (topsoil = 85%, subsoil = 68%), whereas the taxonomic system provided clearer ecological meaning grounded in pedological reasoning (clear toposequences). However, both frameworks generalised the topsoil predominantly as an Ochric topsoil, limiting the interpretation of biologically active surface processes. Subsoil differentiation proved far more informative: Rhodic and Xanthic horizons reflected productive, Fe-rich conditions, while Aquic and Luvic horizons indicated hydromorphic constraints for urban development but remain integral components of the regional water cycle. Although the taxonomic subsoil achieved a moderate accuracy (54%) and agreement (κ = 0.40), its interpretive strength demonstrated the value of diagnostic horizons for linking soil structure, composition and function. Future frameworks that integrate diagnostic and functional properties, supported by advances in high-resolution, multitemporal Earth observation embeddings, could help connect numerical classification with ecosystem understanding. Such developments will enhance the capacity of soil mapping to inform sustainable land management, climate adaptation and ecosystem-service planning across rapidly urbanising landscapes. Declarations Acknowledgements We are grateful to the Agriculture Research Council-Institute of South Africa for providing the data to conduct this study. References Aaquib A, Nandurkar B, Bhagat R, Raut J, Ganvir V, Agrawal V, Kedar A, Sahare P (2018) Influence of Sand Particles on Strength and Durability of Mortar (1:3). SJ Impact Factor: 6 887 Awumbila M (2017) Drivers of Migration and Urbanization in Africa: Key Trends and Issues Basset C, Abou Najm M, Ghezzehei T, Hao X, Daccache A (2023) How does soil structure affect water infiltration? A meta-data systematic review. Soil Tillage Res 226:105577. https://doi.org/10.1016/j.still.2022.105577 Bezdek JC, Ehrlich R, Full W (1984) FCM: The fuzzy c-means clustering algorithm. Comput Geosci 10:191–203. https://doi.org/10.1016/0098-3004(84)90020-7 Bhatta B (2010) Causes and Consequences of Urban Growth and Sprawl. 17–36. https://doi.org/10.1007/978-3-642-05299-6_2 Brown CF, Kazmierski MR, Pasquarella VJ, Rucklidge WJ, Samsikova M, Zhang C, Shelhamer E, Lahera E, Wiles O, Ilyushchenko S, Gorelick N, Zhang LL, Alj S, Schechter E, Askay S, Guinan O, Moore R, Boukouvalas A, Kohli P (2025) AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data. https://doi.org/10.48550/arXiv.2507.22291 Carré F, Jacobson M (2009) Numerical classification of soil profile data using distance metrics. Geoderma 148:336–345 Clarke CE, Roux JL, Ellis F, Clercq WPD, Waals JHVD (2020) Pedogenesis of bleached topsoils occurring on weakly structured, high chroma subsoils in South Africa. CATENA 193:104634–104634. https://doi.org/10.1016/j.catena.2020.104634 Cohen J (1960) A Coefficient of Agreement for Nominal Scales. Educ Psychol Meas 20:37–46. https://doi.org/10.1177/001316446002000104 Colombo C, Buondonno A, Violante A, Torrent J (1991) The Contrasting Effect of Goethite and Hematite on Phosphate Sorption and Desorption by Terre Rosse. Z für Pflanzenernährung und Bodenkunde 154:301–305. https://doi.org/10.1002/jpln.19911540410 Ding Y, Gao X, Shu D, Siddique KHM, Song X, Wu P, Li C, Zhao X (2024) Enhancing soil health and nutrient cycling through soil amendments: Improving the synergy of bacteria and fungi. Sci Total Environ 923:171332. https://doi.org/10.1016/j.scitotenv.2024.171332 Ewis D, Ba-Abbad MM, Benamor A, El-Naas MH (2022) Adsorption of organic water pollutants by clays and clay minerals composites: A comprehensive review. Appl Clay Sci 229:106686. https://doi.org/10.1016/j.clay.2022.106686 Feng Z, Atzberger C, Jaffer S, Knezevic J, Sormunen S, Young R, Lisaius MC, Immitzer M, Jackson T, Ball J, Coomes DA, Madhavapeddy A, Blake A, Keshav S (2025) TESSERA: Precomputed FAIR Global Pixel Embeddings for Earth Representation and Analysis. https://doi.org/10.48550/arXiv.2506.20380 Fey M, Hughes J, Lambrechts J, Dohse T (2001) Chapter 2: The soil groups: distribution, properties, classification, genesis and use, in: Soils of South Africa. pp. 9–10. https://doi.org/10.1017/CBO9780511491757.014 Fick SE, Hijmans RJ (2017) WorldClim 2: new 1km spatial resolution climate surface for global land areas. Int J Climatol 37:4302–4315 Finisdore J, Rhodes C, Haines-Young R, Maynard S, Wielgus J, Dvarskas A, Houdet J, Quétier F, Lamothe KA, Ding H, Soulard F, Van Houtven G, Rowcroft P (2020) The 18 benefits of using ecosystem services classification systems. Ecosyst Serv 45:101160–101160. https://doi.org/10.1016/j.ecoser.2020.101160 Fiorentino A, Rajput FZ, Di Serio A, Baldi V, Guarino F, Baldantoni D, Ronga D, Mazzei P, Motta O, Falanga M, Cicatelli A, Castiglione S (2025) Role of Plants and Urban Soils in Carbon Stock: Status, Modulators, and Sustainable Management Practices. Plants 14:546. https://doi.org/10.3390/plants14040546 Flynn T, Clarke C, Kostecki R, Rebi A (2025) A comparison of country-scale subsoil predictions between a numeric and a taxonomic soil classification system. Geoderma Reg 40:e00902. https://doi.org/10.1016/j.geodrs.2024.e00902 Flynn T, de Clercq W, Rozanov A, Clarke C (2019) High-resolution digital soil mapping of multiple soil properties: an alternative to the traditional field survey? South Afr J Plant Soil 1–11. https://doi.org/10.1080/02571862.2019.1570566 Flynn T, Kostecki R (2024) Spatial downscaling of global soil texture classes into 30 m images at the province scale. Geomatica 76:100028. https://doi.org/10.1016/j.geomat.2024.100028 Flynn T, Triantafilis J, Rozanov A, Ellis F, Lázaro-López A, Watson A, Clarke C (2021) Numerical soil horizon classification from South Africa’s legacy database. CATENA 206:105543. https://doi.org/10.1016/j.catena.2021.105543 Friedman J, Hastie T (1998) Additive Logistic Regression: a Statistical View of Boosting Hartigan (1975) Clustering Algorithms. Wiley, New York He Y, Wang Z, Zhu J, Lin X, Qi J (2025) Soil Carbon Sequestration: Role of Fe Oxides and Polyphenol Oxidase Across Temperature and Cultivation Systems. Plants 14:927. https://doi.org/10.3390/plants14060927 Herold M, Mayaux P, Woodcock CE, Baccini A, Schmullius C (2008) Some challenges in global land cover mapping: An assessment of agreement and accuracy in existing 1 km datasets. Remote Sens Environ 112:2538–2556. https://doi.org/10.1016/j.rse.2007.11.013 Horton R, Wierenga PJ (1983) Estimating the Soil Heat Flux from Observations of Soil Temperature Near the Surface. Soil Sci Soc Amer J 47:14–20. https://doi.org/10.2136/sssaj1983.03615995004700010003x Hossain Anni A, Cohen S, Praskievicz S (2020) Sensitivity of urban flood simulations to stormwater infrastructure and soil infiltration. J Hydrol 588:125028–125028. https://doi.org/10.1016/j.jhydrol.2020.125028 Karthika V, Sekaran U, Jainullabudeen GB, Nagarathinam A (2022) Advances in bioremediation of industrial wastewater containing metal pollutants. Biological Approaches to Controlling Pollutants. Elsevier, pp 163–177. https://doi.org/10.1016/B978-0-12-824316-9.00001-X Kaufman L, Rousseeuw PJ (1960) Clustering by means of medoids. In: Dodge Y (ed) Statisical Data Analysis Based on the L1 Norm. North-Holland, Amsterdam, pp 405–416 Konovalov A, Krajbich I (2018) Neurocomputational Dynamics of Sequence Learning. Neuron 98:1282–1293e4. https://doi.org/10.1016/j.neuron.2018.05.013 Kuhn M, Johnson K (2013) Applied Predictive Modeling. Springer New York. https://doi.org/10.1007/978-1-4614-6849-3 Land Type Survey Staff (2006) Land Types of South Africa on 1:250 000 scale. Pretoria, South Africa Landis JR, Koch GG (1977) The Measurement of Observer Agreement for Categorical Data. Biometrics 33:159–174 Le Mer J, Roger P (2001) Production, oxidation, emission and consumption of methane by soils: A review. Eur J Soil Biol 37:25–50. https://doi.org/10.1016/S1164-5563(01)01067-6 Li P, Wang Z-H (2021) Environmental co-benefits of urban greening for mitigating heat and carbon emissions. J Environ Manage 293:112963–112963. https://doi.org/10.1016/j.jenvman.2021.112963 Liu S, Lin Z, Duan X, Deng Y (2024) Effects of soil microorganisms on aggregate stability during vegetation recovery in degraded granitic red soil areas. Appl Soil Ecol 204:105734. https://doi.org/10.1016/j.apsoil.2024.105734 Liu Y, Meng Q, Zhang L, Wu C (2022) NDBSI: A normalized difference bare soil index for remote sensing to improve bare soil mapping accuracy in urban and rural areas. CATENA 214:106265–106265. https://doi.org/10.1016/j.catena.2022.106265 Ma Y, Woolf D, Fan M, Qiao L, Li R, Lehmann J (2023) Global crop production increase by soil organic carbon. Nat Geosci 16:1159–1165. https://doi.org/10.1038/s41561-023-01302-3 Magiera T, Strzyszcz Z, Rachwal M (2007) Mapping particulate pollution loads using soil magnetometry in urban forests in the Upper Silesia Industrial Region, Poland. For Ecol Manag 248:36–42. https://doi.org/10.1016/j.foreco.2007.02.034 Mason ARG, Salomon MJ, Lowe AJ, Cavagnaro TR (2023) Microbial solutions to soil carbon sequestration. J Clean Prod 417:137993. https://doi.org/10.1016/j.jclepro.2023.137993 Miller MAE, Shepherd KD, Kisitu B, Collinson J (2021) iSDAsoil: The first continent-scale soil property map at 30 m resolution provides a soil information revolution for Africa. PLoS Biol 19:e3001441–e3001441. https://doi.org/10.1371/journal.pbio.3001441 Møller AB, Beucher AM, Pouladi N, Humlekrog Greve M (2020) Oblique geographic coordinates as covariates for digital soil mapping. Soil 6:269–289. https://doi.org/10.5194/soil-6-269-2020 NASA JPL (2022) NASADEM Merged DEM Global 1 arc second V001. NASA EOSDIS Land Processes DAAC O’Riordan R, Davies J, Stevens C, Quinton JN, Boyko C (2021) The ecosystem services of urban soils: A review. Geoderma 395:115076–115076. https://doi.org/10.1016/j.geoderma.2021.115076 Ramírez PB, Machado S, Singh S, Plunkett R, Calderón FJ (2023) Addressing the effects of soil organic carbon on water retention in US Pacific Northwest wheat–soil systems. Front Soil Sci 3:1233886. https://doi.org/10.3389/fsoil.2023.1233886 Reddi LN, Jain AK, Yun H-B (2012) Soil materials for earth construction: properties, classification and suitability testing. Modern Earth Buildings. Elsevier, pp 155–171. https://doi.org/10.1533/9780857096166.2.155 Rubio CM, Ubeda X, Ferrer F (2012) Response of the Thermal Conductivity as a Function of Water Content of a Burnt Mediterranean Loam Soil. OJSS 02, 1–6. https://doi.org/10.4236/ojss.2012.21001 Rugendo MK, Gichimu BM, Mugwe JN, Mucheru-Muna M, Mugendi DN (2023) Surface runoff and soil erosion from Nitisols and Ferralsols as influenced by different soil organic carbon levels under simulated rainfall conditions. Heliyon 9:e17684. https://doi.org/10.1016/j.heliyon.2023.e17684 Sabtan AA (2005) Geotechnical properties of expansive clay shale in Tabuk, Saudi Arabia. J Asian Earth Sci 25:747–757. https://doi.org/10.1016/j.jseaes.2004.07.003 Schwertmann U (1971) Transformation of Hematite to Goethite in Soils. Nature 232:624–625. https://doi.org/10.1038/232624a0 Shi T, Hu X, Guo L, Su F, Tu W, Hu Z, Liu H, Yang C, Wang J, Zhang J, Wu G (2021) Digital mapping of zinc in urban topsoil using multisource geospatial data and random forest. Sci Total Environ 792:148455–148455. https://doi.org/10.1016/j.scitotenv.2021.148455 Soil Classification Working Group (1991) Soil Classificatioin: a Taxonomic System for South Africa, 2nd edn. Department of Agricultural Development, Pretoria, South Africa Soil Survey Staff (2014) Keys to soil taxonomy, 12th edn. USDA-Natural Resources Conservation Service, Washington, DC Triantafilis J, Ward WT, Odeh IOA, McBratney AB (2001) Creation and Interpolation of Continuous Soil Layer Classes in the Lower Namoi Valley. Soil Sci Soc Am J 65:403–413. https://doi.org/10.2136/sssaj2001.652403x USGS (2021) Landsat collection 2 Van Tol JJ, Le Roux PA, Hensley M (2011) Soil Indicators of Hillslope Hydrology. Principles-Application Assessments Soil Sci 209–240. https://doi.org/10.5772/1860 Van Zijl G, Le Roux P (2014) Creating a conceptual hydrological soil response map for the Stevenson Hamilton Research Supersite, Kruger National Park, South Africa. Water SA 40:331–336. https://doi.org/10.4314/wsa.v40i2.15 Van Zijl G, van Tol J, Tinnefeld M, Le Roux P (2019) A hillslope based digital soil mapping approach, for hydropedological assessments. Geoderma 354:113888–113888. https://doi.org/10.1016/j.geoderma.2019.113888 Van Zijl GM, Le Roux PA, Turner DP (2013) Disaggregation of land types using terrain analysis, expert knowledge and GIS methods. South Afr J Plant Soil 30:123–129. https://doi.org/10.1080/02571862.2013.806679 Villa P, Malucelli F, Scalenghe R (2018) Sci Total Environ 612:590–604. https://doi.org/10.1016/j.scitotenv.2017.08.250 . Multitemporal mapping of peri-urban carbon stocks and soil sealing from satellite data Voelz JL, Arnold WA, Penn RL (2018) Redox-induced nucleation and growth of goethite on synthetic hematite nanoparticles. Am Mineral 103:1021–1029. https://doi.org/10.2138/am-2018-6342 Wang Y, Wang S, Liu C, Zhu E, Jia J, Feng X (2023) Shifting relationships between SOC and molecular diversity in soils of varied carbon concentrations: Evidence from drained wetlands. Geoderma 433:116459. https://doi.org/10.1016/j.geoderma.2023.116459 Webster R, Burrough PA (1972) Computer-based soil mapping of small areas from sample data II classification smoothing. J Soil Sci 23:222–234. https://doi.org/10.1111/j.1365-2389.1972.tb01655.x Weininger D, Weininger A, Weininger JL (1989) SMILES. 2. Algorithm for generation of unique SMILES notation. J Chem Inf Comput Sci 29:97–101. https://doi.org/10.1021/ci00062a008 Wu Q, Congreves KA (2024) Soil health benefits associated with urban horticulture. Sci Total Environ 912:168852. https://doi.org/10.1016/j.scitotenv.2023.168852 Zhang Y, Hartemink AE (2019) A method for automated soil horizon delineation using digital images. Geoderma 343:97–115. https://doi.org/10.1016/j.geoderma.2019.02.002 Zhu AX (1997) A similarity model for representing soil spatial information. Geoderma 77:217–242 Tables Tables 1 to 5 are available in the Supplementary Files section Additional Declarations The authors declare no competing interests. Supplementary Files Tables.docx Cite Share Download PDF Status: Posted Version 1 posted 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-8006960","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":538505610,"identity":"47fa93e1-50e6-4ad1-8d0b-a942511782dc","order_by":0,"name":"Trevan Flynn","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4klEQVRIiWNgGAWjYPACORkgwfgASPDwEVLLA6GMQTSzAUiAjRQtbBIgJkEt9tKHj0n83GHAY3C791jl1xw7GTYG5oePbuCzhS8tTbL3DFDLnXNpt2W3JQMdxmZsnINPCw+PsQFv2x8egxs5ZrcltzEDtfCwSePXwv/Z8G+bAVhLseS2emK08DA+5oVqYfy47TARWs6wGT6WBWqRvJFjLM247TgPGzMBv7D3MD84+LbNQI7vRo7hx5/bqu352ZsfPsanBQ4UDgDjEhxLzMQoBwH5BmCK+UGs6lEwCkbBKBhRAAARLj4MnmTtpgAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0002-5187-9946","institution":"University of Fort Hare","correspondingAuthor":true,"prefix":"","firstName":"Trevan","middleName":"","lastName":"Flynn","suffix":""}],"badges":[],"createdAt":"2025-11-01 16:34:42","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8006960/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8006960/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":95223523,"identity":"4ae02916-1632-4903-9e15-d1df746856b4","added_by":"auto","created_at":"2025-11-05 16:22:24","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":18071640,"visible":true,"origin":"","legend":"","description":"","filename":"manuscriptTrevanFlynn.docx","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/2fd53ca64f25851eb520b9f7.docx"},{"id":95077907,"identity":"cb99e67b-1cfa-4bb9-a4e8-c9ef1658506c","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":342,"visible":true,"origin":"","legend":"","description":"","filename":"rs8006960.json","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/91ca45c1ae860802b407902d.json"},{"id":95077911,"identity":"b687d96f-2f1d-452b-a5a0-75fd73b28cba","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":169681,"visible":true,"origin":"","legend":"","description":"","filename":"rs80069600enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/9cd796bc5f572ed6220fb8d4.xml"},{"id":95077909,"identity":"da8a6ed8-74b8-4e73-a505-0fb5c1b8dc85","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4383166,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/1196608995d4fd64c7e622ca.png"},{"id":95077910,"identity":"53c8073a-dc10-49a5-8455-0ec2a69af01a","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":88104,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/3e2fc5a0db74471e6d942501.jpeg"},{"id":95077913,"identity":"0a30befe-2e8c-43c6-a8c3-786a78554837","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3802662,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/4be6f1f75a1fd5d7c2399fe0.png"},{"id":95077921,"identity":"7f75c479-a502-4a4f-9c4b-ac25c584fbae","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4940825,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/7135874eca2d9035dca8644f.png"},{"id":95077915,"identity":"9e4e6be1-6ec9-436a-b07a-ef016813b7fe","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":571320,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/ed0f88a5b6046d45fd783c1a.png"},{"id":95224034,"identity":"1cac78d7-417e-430e-a7cb-bf97087680ce","added_by":"auto","created_at":"2025-11-05 16:23:13","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15926,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/d4fef2fd7270a5eb14902e14.png"},{"id":95077917,"identity":"9a9342fb-f78f-4b3f-bd24-631fe53b3198","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":506981,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/a881a737e90bf2c563dbf78f.png"},{"id":95077918,"identity":"e24c43cc-d0fb-472c-b142-43e8593693c9","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":702820,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/d855962368bbe9e1c1c1e57c.png"},{"id":95077920,"identity":"75632d29-16da-4108-810e-1a52577f999c","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"xml","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":168911,"visible":true,"origin":"","legend":"","description":"","filename":"rs80069600structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/4e0fbac59b60927768f02b93.xml"},{"id":95077916,"identity":"9a0c9c29-812d-43b7-9d09-79c9f4d66f80","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"html","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":176607,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/f73de6bf9a21c5b894eb6fbf.html"},{"id":95077904,"identity":"cd3c1a53-beea-4ade-bcea-e3e02c307cee","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":571047,"visible":true,"origin":"","legend":"\u003cp\u003eGauteng’s location in South Africa (a) and Gauteng with major urban areas and soil observations (b).\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/729e4a16fc59956e6c1c0ba9.png"},{"id":95077905,"identity":"ffec43c0-9ca0-43af-a523-aa57e4183bd5","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":39205,"visible":true,"origin":"","legend":"\u003cp\u003eFlow chart of the method taken to determine ecosystem services provided by soil in Gauteng.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/e82b4e2e528d69e059aa543c.png"},{"id":95077906,"identity":"87e9d3ea-4190-498a-b394-40fe8bb19b2e","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":530153,"visible":true,"origin":"","legend":"\u003cp\u003eTopsoil class predictions for the numerical system (a) and the South African taxonomic system (b).\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/0167319f248c03f30c3a8d36.png"},{"id":95224515,"identity":"bbd65cdd-ae4c-4e62-8732-1650ba092254","added_by":"auto","created_at":"2025-11-05 16:23:50","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":676218,"visible":true,"origin":"","legend":"\u003cp\u003eSubsoil predictions for (a) the numeric and (b) the South African classification systems.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/1c173029197f30fde3647253.png"},{"id":95312125,"identity":"62fee460-067f-4a20-94ac-e31e760aedab","added_by":"auto","created_at":"2025-11-06 15:47:17","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2304448,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/0fc11c08-f2f8-47e2-8d46-52046f9915b8.pdf"},{"id":95077912,"identity":"764c0f58-b302-4382-be45-eed1192f471c","added_by":"auto","created_at":"2025-11-04 05:25:07","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":4675191,"visible":true,"origin":"","legend":"","description":"","filename":"Tables.docx","url":"https://assets-eu.researchsquare.com/files/rs-8006960/v1/1c948d2aefccf97a1be8c302.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eEvaluating diagnostic horizons as indicators of ecosystem services through digital soil mapping in urban Gauteng, South Africa\u003c/p\u003e","fulltext":[{"header":"Highlights","content":"\u003cp\u003e\u0026bull; Urban soils mediate critical ecosystem services across rapidly expanding African cities.\u003c/p\u003e\u003cp\u003e\u0026bull; A numerical and taxonomic soil classification system revealed distinct functional strengths.\u003c/p\u003e\u003cp\u003e\u0026bull; Taxonomic subsoil horizons better captured ecosystem service potential and soil multifunctionality.\u003c/p\u003e\u003cp\u003e\u0026bull; Classification should reflect the multidimensionality of soil and not only the vertical distribution.\u003c/p\u003e"},{"header":"1. Introduction","content":"\u003cp\u003eMetropolitan expansion, which is increasingly common across Africa (Awumbila, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), places substantial pressure on the environment and shapes how populations grow, directly influencing the quality of life within and around expanding urban areas (Bhatta, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Urban soils play a critical role in providing ecosystem services: they support stormwater management (Hossain Anni et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), serve as a medium for urban horticulture that enhances soil health (Wu and Congreves, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), help to mitigate urban heat (Li and Wang, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), and supply essential materials for construction (Reddi et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), culture, and recreation (O\u0026rsquo;Riordan et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Furthermore, industrial and mining wastes, along with other pollutants, often leach into surrounding environments, affecting not only urban soils but also adjacent natural and agricultural lands (Karthika et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe ecosystem services that soil provides arise from the underlying soil functions expressed through their physical, chemical, and biological properties. Consequently, digital soil mapping often focuses on producing detailed and accurate maps of these soil properties. Such studies commonly employ approaches to assess soil sealing (Villa et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), carbon stocks (Fiorentino et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Villa et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), bare-soil exposure (Liu et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and pollution loads and transport (Magiera et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Shi et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). However, the spatial patterns of soil properties and processes are often difficult to visualize and conceptualize, and mapping multiple soil properties fails to capture the true multifunctional behavior of soil in a form that is easily interpretable.\u003c/p\u003e\u003cp\u003eDue to this inherent complexity, a disconnect often arises between the ecosystem services that soil provides and how they are perceived or utilised within a metropolis, emphasising the need for classification systems (Konovalov and Krajbich, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Classification frameworks help translate the continuous variability of soil properties into discrete, interpretable categories that encapsulate multiple functions simultaneously. Unlike single-property maps, classified soil maps can support broader applications, including land-use planning, environmental assessment and ecosystem service evaluation as they encapsulate multiple soil attributes into one. However, the process of classification can also lead to the loss of detail and spatial heterogeneity (Zhu, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e1997\u003c/span\u003e), potentially over simplifying the true nature of soil.\u003c/p\u003e\u003cp\u003eTo address these challenges, various numerical classification systems have emerged at both local and national scales (Bezdek et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1984\u003c/span\u003e; Carr\u0026eacute; and Jacobson, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Flynn et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Hartigan, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e1975\u003c/span\u003e; Triantafilis et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Webster and Burrough, \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e1972\u003c/span\u003e; Zhang and Hartemink, \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), offering data-driven approaches that classify soil more or less, objectively. The main advantage of numerical systems lies in their quantitative and less subjective nature, while traditional taxonomic systems benefit from hierarchical structure and interpretability, facilitating communication among scientists, policymakers and practitioners (Finisdore et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Despite their differences, both approaches provide valuable perspectives for understanding and mapping ecosystem services, forming a complementary foundation for soil-informed decision-making that supports environmental and societal wellbeing.\u003c/p\u003e\u003cp\u003eThe aim of this study was to evaluate how effectively two distinct soil classification systems represent the ecosystem services that soil provides within and around the metropolitan areas of Gauteng Province, South Africa (ZA). It is hypothesized that both the numeric and ZA\u0026rsquo;s taxonomic classification systems would capture meaningful spatial patterns of ecosystem services within the subsoil, where long-term processes occur, but not within the topsoil, which is more influenced by anthropogenic disturbance. The level of detail, accuracy and interpretability of the resulting maps is crucial for disseminating soil information, guiding sustainable urban planning, and identifying opportunities for improving classification systems. Although traditional classification systems are largely vertical in conception, soil functions as a four-dimensional living system\u0026mdash;integrating space, depth, time, and function\u0026mdash;and modern classification must evolve to reflect this inherent complexity.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003ch2\u003e2.1\u0026nbsp; \u0026nbsp; \u0026nbsp;Site\u003c/h2\u003e\n\u003cp\u003eGauteng Province (Figure 1) is located in the high-altitude grasslands of the Highveld in central South Africa, centered approximately at 26.3\u0026deg; S and 28.1\u0026deg; E. The province is landlocked, bordered by the Free State to the south, North West to the west, Mpumalanga to the east, and Limpopo to the north. Despite being the smallest province by area, Gauteng is the most densely populated region in the country, with an average population density exceeding 715 people km⁻\u0026sup2; (Statistics South Africa, 2016). It encompasses two major metropolitan areas of Johannesburg and Pretoria along with their suburbs, and rapidly expanding cities such as Soshanguve and Vereeniging. Gauteng was selected as the study area because it represents a diverse and dynamic landscape that includes extensive urban areas, productive arable lands and natural reserves making it ideal for examining the spatial relationships between soil functions, ecosystem services and metropolitan expansion.\u003c/p\u003e\n\u003cp\u003eGauteng has an average elevation of approximately 1,500 m, a mean annual temperature of about 20 \u0026deg;C, and a mean annual precipitation of around 650 mm. Sitting on the Kaapvaal Craton, the province is well known for the Witwatersrand region, which has a long history of gold, diamond and silver mining and remains an important center of industrial activity and economic growth (Abiye et al., 2011). Although Gauteng\u0026rsquo;s economy is largely driven by the mining and financial sectors, it also supports a productive agricultural industry that includes livestock, vegetables, maize, cotton and sorghum (Nesamvuni et al., 2016).\u003c/p\u003e\n\u003ch2\u003e2.2\u0026nbsp; \u0026nbsp; \u0026nbsp;Soil data\u003c/h2\u003e\n\u003cp\u003eSoil profile data w\u003cem\u003eas\u003c/em\u003e obtained from the South African Land Type Survey (LTS; Land Type Survey Staff, 2006), comprising 565 observations across Gauteng (Figure 1b). These same observations were used to evaluate both the numerical and taxonomic classification systems. The numerical system was previously developed by Flynn et al. (2021), who clustered soil properties into five topsoil and nine subsoil classes (Table 1) using a clustering algorithm (Kaufman and Rousseeuw, 1960). The algorithm was selected for its robustness to outliers and computational efficiency with large datasets, producing medoid-based (median) clusters representative of soil property distributions. As the only numerical classification system derived from national-scale LTS data, and one that clusters on master horizons, it provides a suitable basis for comparison with the South African taxonomic system (Soil Classification Working Group, 1991) at the province scale.\u003c/p\u003e\n\u003cp\u003eAlthough not the most recent system, the South African \u0026ldquo;Blue Book\u0026rdquo; (Soil Classification Working Group, 1991-2006) soil classification was used as the ZA taxonomic system in this study (Table 2), as it had already been fully incorporated into the LTS at the time of the research. For interpretability, the ZA taxonomic names were translated into the corresponding United States Department of Agriculture (USDA) Soil Taxonomy names or equivalent Latin names. In the database, Mollic, Ochric and Vertic soil were the only topsoil types identified, and the subsoil was grouped into eight categories based on morphological similarities and to account for the limited number of observations in some subsoil classes.\u003c/p\u003e\n\u003ch2\u003e2.3\u0026nbsp; \u0026nbsp; \u0026nbsp;Covariates\u003c/h2\u003e\n\u003cp\u003eIn total, 60 covariates (Table 3) were obtained or loaded into Google Earth Engine (GEE; Gorelick et al., 2017), representing the \u003cem\u003escorpan\u003c/em\u003e factors (McBratney et al., 2003). To capture the soil (\u003cem\u003es\u003c/em\u003e), parent material (\u003cem\u003ep\u003c/em\u003e) and organism (o) factors, bands and indices derived from cloud-masked surface reflectance Landsat 8 OLI/TIRS (USGS, 2021) imagery at a 30 m resolution were used. For each band and index, the median value of the time series from 2013 to 2021 was computed. The \u003cem\u003es\u003c/em\u003e factor was also represented by iSDAsoil (Miller et al., 2021), a 30 m soil database of Africa. Climatic factors were acquired from worldclim (Fick and Hijmans, 2017), at a 90 m resolution. All bioclimatic variables (19) were used. The r factor was represented by the 30 m NASADEM (NASA JPL, 2022) with local and global derivatives. Six oblique geographic coordinates (OGC; M\u0026oslash;ller et al., 2020) were derived to represent the n factor. Oblique geographic coordinates were used instead of latitude and longitude to prevent highly correlated variables and map artifacts.\u003c/p\u003e\n\u003ch2\u003e2.4\u0026nbsp; \u0026nbsp; \u0026nbsp;Model development\u003c/h2\u003e\n\u003cp\u003eThe digital soil mapping framework combined machine learning and pedological reasoning to evaluate how well numerical and taxonomic systems capture the spatial organization of diagnostic horizons. (Figure 2). By applying identical observations and environmental covariates to both classification frameworks, the model isolated differences in interpretability rather than data structure. Model evaluation emphasised not only statistical performance, through overall accuracy and Cohen\u0026rsquo;s \u003cem\u003e\u003cbr\u003e\u003c/em\u003e\u003cimg width=\"9\" height=\"46\" src=\"data:image/png;base64,R0lGODlhCQAuAHcAMSH+GlNvZnR3YXJlOiBNaWNyb3NvZnQgT2ZmaWNlACH5BAEAAAAALAAAEAAJAAcAhAAAAAAAAAAAOgAAZgA6ZgBmtjoAADoAOjoAZjo6OjqQ22YAAGY6AGa2/5A6AJA6OpC2kJC225Db/7ZmALZmOraQkLb/29uQOtv///+2Zv/bkP//tv//2wECAwECAwECAwUqIDAFCsBRggRsS8M9ESBrCHSoMpAFidHkogLHUcgNS5eBpYLR3FYMgioEADs=\" alt=\"image\"\u003e, but also the ecological and pedological realism of the resulting horizon patterns. This enabled the direct comparison between data-driven and diagnostic approaches in representing soil variability and ecosystem service potential across the landscape.\u003c/p\u003e\n\u003ch3\u003e2.4.1\u0026nbsp; \u0026nbsp;\u0026nbsp;Model\u003c/h3\u003e\n\u003cp\u003eA gradient tree boosting (GTB) model is a generalized additive model in which the base learners are small decision trees. With the addition of each tree, the model iteratively corrects or improves upon the residuals of the previous ensemble (Hastie et al., 2009). Although GTB is additive in structure, it is nonlinear and stochastic, making it highly suitable and accurate for predicting soil attributes (Flynn et al., 2019). The additive nature of the model promotes regularization and stability, helping to prevent overfitting (Friedman and Hastie, 1998), while the nonlinear and stochastic components enable the model to capture complex relationships (Kuhn and Johnson, 2013), such as those between soil classes and satellite derived variables. Mathematically, the predictions of each class \u003cimg width=\"9\" height=\"46\" src=\"data:image/png;base64,R0lGODlhCQAuAHcAMSH+GlNvZnR3YXJlOiBNaWNyb3NvZnQgT2ZmaWNlACH5BAEAAAAALAAADAAJAAsAhAAAAAAAAAAAOgAAZgA6ZgA6kABmtjoAADoAOjoAZjo6ZjpmtjqQ22YAAGa2/5A6AJBmkJDb25Db/7ZmALaQZrbb/7b/27b//9uQOtv/29v////bkP/btv//tgECAwECAwU2INA1AQOc6IZIaJpc7YkVWgxMJtcI7OgAG0XlpIpQFjVZYPlDTQqZhwE1MmEGFohGxdocCMMQADs=\" alt=\"image\"\u003e\u0026nbsp;can be represented as:\u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/p\u003e\n\u003cp\u003eA SMILE (\u0026ldquo;Statistical Machine Intelligence and Learning Machine\u0026rdquo;; Weininger et al., 1989) implementation of GTB was used, with model parameters optimised in GEE. The model was controlled through the learning rate, tree depth, number of trees grown, and subsampling rate. Models were trained using covariates at 10, 30, 90, and 180 m resolutions, with the number of boosting iterations ranging from 100 to 1000 in increments of 100. Additionally, learning rates of 0.1, 0.01, 0.001, and 0.0001 were tested. The default subsampling rate in SMILE models is 0.7; however, values of 0.1 and 0.9 were also evaluated.\u003c/p\u003e\n\u003ch3\u003e2.4.2\u0026nbsp; \u0026nbsp;\u0026nbsp;Evaluation\u003c/h3\u003e\n\u003cp\u003eTo evaluate model performance, the data (\u003cem\u003en\u003c/em\u003e = 565) was split into 80% training and 20% testing sets. The performance was evaluated by how well it predicted on the 20% training data. Evaluation of performance was based on \u003cimg width=\"9\" height=\"46\" src=\"data:image/png;base64,R0lGODlhCQAuAHcAMSH+GlNvZnR3YXJlOiBNaWNyb3NvZnQgT2ZmaWNlACH5BAEAAAAALAAAEAAJAAcAhAAAAAAAAAAAOgAAZgA6ZgBmtjoAADoAOjoAZjo6OjqQ22YAAGY6AGa2/5A6AJA6OpC2kJC225Db/7ZmALZmOraQkLb/29uQOtv///+2Zv/bkP//tv//2wECAwECAwECAwUqIDAFCsBRggRsS8M9ESBrCHSoMpAFidHkogLHUcgNS5eBpYLR3FYMgioEADs=\" alt=\"image\"\u003e\u0026nbsp;and overall accuracy (%). However, the model with the highest \u003cimg width=\"9\" height=\"46\" src=\"data:image/png;base64,R0lGODlhCQAuAHcAMSH+GlNvZnR3YXJlOiBNaWNyb3NvZnQgT2ZmaWNlACH5BAEAAAAALAAAEAAJAAcAhAAAAAAAAAAAOgAAZgA6ZgBmtjoAADoAOjoAZjo6OjqQ22YAAGY6AGa2/5A6AJA6OpC2kJC225Db/7ZmALZmOraQkLb/29uQOtv///+2Zv/bkP//tv//2wECAwECAwECAwUqIDAFCsBRggRsS8M9ESBrCHSoMpAFidHkogLHUcgNS5eBpYLR3FYMgioEADs=\" alt=\"image\"\u003e\u0026nbsp;was taken as the top performing model. In other words, the model with the highest agreement for a given soil class (Cohen, 1960). Evaluation was conducted to make sure that the interpretation of the spatial patterns and ecosystem services was plausible; therefore, only \u003cimg width=\"9\" height=\"46\" src=\"data:image/png;base64,R0lGODlhCQAuAHcAMSH+GlNvZnR3YXJlOiBNaWNyb3NvZnQgT2ZmaWNlACH5BAEAAAAALAAAEAAJAAcAhAAAAAAAAAAAOgAAZgA6ZgBmtjoAADoAOjoAZjo6OjqQ22YAAGY6AGa2/5A6AJA6OpC2kJC225Db/7ZmALZmOraQkLb/29uQOtv///+2Zv/bkP//tv//2wECAwECAwECAwUqIDAFCsBRggRsS8M9ESBrCHSoMpAFidHkogLHUcgNS5eBpYLR3FYMgioEADs=\" alt=\"image\"\u003e\u0026nbsp;and accuracy were reported.\u003c/p\u003e\n\u003cp\u003eCovariates were also evaluated on how important they were for predictions. This rank represents the percentage increase of the loss function when the covariate is permutated (Weininger et al., 1989). It is a relative importance index and gives an indication of which covariates correlate well to that soil class. As a large pool of covariates was used in the model, the validation of the model with covariates was needed to help with the assessment of the soil horizon spatial patterns.\u003c/p\u003e"},{"header":"3. Results and Discussion","content":"\u003ch2\u003e3.1\u0026nbsp; \u0026nbsp; \u0026nbsp;Model performance\u003c/h2\u003e\n\u003cp\u003eBefore interpreting the predictive performance and spatial patterns of the numerical and taxonomic soil classification systems, the GTB model was first evaluated to ensure reliability. This evaluation was essential to determine whether the observed spatial distributions were genuinely influenced by environmental factors. Establishing a robust level of model accuracy provided the foundation for confidently analysing soil horizon predictions and their implications for ecosystem service assessment.\u003c/p\u003e\n\u003ch3\u003e3.1.1\u0026nbsp; \u0026nbsp;\u0026nbsp;Agreement and accuracy\u003c/h3\u003e\n\u003cp\u003eThe numerical system achieved a greater\u003cbr\u003e\u003cimg width=\"9\" height=\"46\" src=\"data:image/png;base64,R0lGODlhCQAuAHcAMSH+GlNvZnR3YXJlOiBNaWNyb3NvZnQgT2ZmaWNlACH5BAEAAAAALAAAEAAJAAcAhAAAAAAAAAAAOgAAZgA6ZgBmtjoAADoAOjoAZjo6OjqQ22YAAGY6AGa2/5A6AJA6OpC2kJC225Db/7ZmALZmOraQkLb/29uQOtv///+2Zv/bkP//tv//2wECAwECAwECAwUqIDAFCsBRggRsS8M9ESBrCHSoMpAFidHkogLHUcgNS5eBpYLR3FYMgioEADs=\" alt=\"image\"\u003e\u0026nbsp;for topsoil (0.63) and subsoil (0.54) relative to the subsoil (Table 4). Numerical topsoil achieved a substantial agreement, while the subsoil achieved moderate agreement according to Landis and Koch (1977). On the other hand, only topsoil achieved moderate agreement (0.56) for the taxonomic system and subsoil achieved fair agreement (0.40). However, the taxonomic topsoil achieved the highest accuracy (98%) followed by numerical topsoil (85%), numerical subsoil (68%) and taxonomic subsoil (54%). This was expected, as accuracy generally decreases with an increasing number of soil classes (Herold et al., 2008). Although the fair taxonomic subsoil agreement raises concerns regarding reliability, the results remain within the range commonly reported in digital soil mapping studies. They are therefore regarded as sufficient for further interpretation.\u003c/p\u003e\n\u003cp\u003eThese results are broadly consistent with previous findings from South Africa. For instance, Van Zijl and Le Roux (2014) achieved an accuracy of 73% when mapping seven hydrological response unit classes in Kruger National Park, while Van Zijl et al. (2013) reported 67% accuracy for six soil classes using fuzzy logic in KwaZulu-Natal Province. Similarly, Van Zijl et al. (2019) obtained an accuracy of 69% and a \u0026kappa; value of 0.59 when mapping five soil classes within the city of Johannesburg using a hillslope-based approach. Additionally, these findings represent an improvement over those reported by Flynn et al. (2025), where the numerical subsoil system achieved an agreement of 0.30 and an accuracy of 57% using the same covariates and model structure. In that earlier study, the algorithm had not been fully optimised, indicating that GTB performance is highly sensitive to model calibration and parameter tuning. Such optimisation can substantially enhance the predictive potential of both classification systems and has important implications for improving the taxonomic subsoil model in this study.\u003c/p\u003e\n\u003cp\u003eAll models performed well with a learning rate of 0.001; however, each class exhibited a different optimal resolution and number of trees. Interestingly, this trend was opposite between the two classification systems. For the numerical topsoil model, the best performance was achieved with the largest number of trees (500) and the finest spatial resolution (30 m), whereas the taxonomic topsoil model required the fewest trees (50) but performed best at the coarsest resolution (180 m). For subsoil predictions, both systems achieved optimal outcomes at a 90 m resolution, with the numerical system requiring 300 trees and the taxonomic system needing 80 trees.\u003c/p\u003e\n\u003ch3\u003e3.1.2\u0026nbsp; \u0026nbsp;\u0026nbsp;Covariate importance\u003c/h3\u003e\n\u003cp\u003eWith the exception of taxonomic topsoil, all soil classes showed strong correlations with multiple climatic factors (Table 5), indicating that climate is an important variable at this spatial support. However, each class responded to different climatic influences. For example, numerical topsoil was strongly correlated with isothermality (i.e., difference between nighttime and daytime temperature), mean diurnal range (i.e., difference between the maximum and minimum temperature) and temperature seasonality, suggesting that its distribution is influenced primarily by temperature variability rather than mean temperature. In contrast, numerical subsoil was most closely correlated with mean annual temperature and the mean temperature of the coldest quarter. This is a reasonable outcome, as the topsoil responds more readily to short-term environmental fluctuations, whereas the subsoil remains comparatively insulated from such variation (Horton and Wierenga, 1983; Rubio et al., 2012).\u003c/p\u003e\n\u003cp\u003eThe taxonomic topsoil was also the only class to rank iSDAsoil (clay and CEC) among the ten most important predictors. These soil properties were initially assumed to correlate strongly with all horizons. However, iSDAsoil data is already highly correlated with the environmental covariates used in the model. Due to GTB being a highly regularised algorithm, it may have reduced the coefficients of the iSDAsoil variables, resulting in their lower ranking for most horizons. An alternative would be to apply principal component analysis prior to model training and combining all iSDAsoil variables into its principle components. This approach could also reduce computational demand and multicollinearity among predictors.\u003c/p\u003e\n\u003cp\u003eNeighborhood was represented by OGCs, which were ranked high in both classification systems. This is noteworthy, as OGCs are not commonly cited among the most influential covariates in digital soil mapping studies. Their importance likely arises from their ability to capture spatial trends in a manner similar to geostatistical methods (M\u0026oslash;ller et al., 2020). The horizons exhibit spatial autocorrelation; however, they cannot be interpolated using ordinary kriging because they represent categorical data, and such computation is impractical at this scale. Therefore, OGCs provide a powerful alternative for capturing spatial structure and enhancing model performance.\u003c/p\u003e\n\u003ch2\u003e3.2\u0026nbsp; \u0026nbsp; \u0026nbsp;Spatial predictions\u003c/h2\u003e\n\u003cp\u003eA GTB model was used to predict soil horizons for both a numerical and ZA taxonomic classification systems in and around the metropolises of Gauteng. The model evaluation statistics indicated performance that was sufficiently reliable to justify a detailed examination of the spatial patterns observed and the ecosystem services inferred. Establishing this level of predictive accuracy was essential; without it, the assessment of soil derived ecosystem services and their spatial interpretation would have remained too uncertain to support meaningful conclusions.\u003c/p\u003e\n\u003ch3\u003e3.2.1\u0026nbsp; \u0026nbsp;\u0026nbsp;Topsoil\u003c/h3\u003e\n\u003cp\u003eThe predictions showed a similar spatial distribution for both the numerical (Figure 3a) and taxonomic systems (Figure 3b) predominantly predicting A3 (95% of area) and Ochric (98% of area) horizons, respectively. Although the accuracy was high for both systems, the maps conceivably do not give much information about ecosystem services. The lack of pedodiversity seen in the topsoil is of concern to both classification systems as they clearly are not distinguishing a key region for microbial activity. Nevertheless, there was an area between Pretoria and Soshanguve just outside the De Onderstepoort Nature Reserve as well as the far southeast near the town of Devon, where both systems predicted the same horizon other than the Ochric. In the numeric system, this was A5 and is characteristic of saline soils with a high pH, CEC and clay content. In the taxonomic system, these soils were classified as Vertic topsoil that have the same saline properties (Fey et al., 2001). Surprisingly, this Vertic topsoil zone was also predicted through categorical downscaling texture classes from SoilGrid\u0026rsquo;s (Flynn and Kostecki, 2024)\u003c/p\u003e\n\u003cp\u003eThe numerical system exhibited greater topsoil pedodiversity in the northeast, where classes A1, A2 and A4 were predicted. Their distribution appeared to be influenced by the region\u0026rsquo;s more diverse cumulative geology, vegetation and topography. Both A2 and A4 were also identified within the city of Johannesburg, although their distribution was limited to a few predictions. Overall, the topsoil predictions lacked substantial or meaningful spatial patterns, detail or pedological interpretability.\u003c/p\u003e\n\u003ch3\u003e3.2.2 \u0026nbsp; \u0026nbsp;Subsoil\u003c/h3\u003e\n\u003cp\u003eLike the topsoil, the subsoil exhibited the greatest pedodiversity in the northeast, where all horizons of the numerical classification were predicted (Figure 4a). Horizon B1 (low clay, coarse sand) followed areas of high relief across Gauteng, occurring most abundantly in and around Johannesburg. Horizons B2 and B3 were spatially associated with B1 in the northeast but appeared more scattered in areas of lower relief. Horizon B2 (low clay, fine sand) extended southward and, to a lesser extent, occurred within Johannesburg. Horizon B3 was predicted primarily in the West Rand (western Johannesburg) and adjacent valleys. Given its occurrence in low-lying areas and its properties of high clay content, elevated CEC and mildly alkaline, this horizon likely has strong structural development characteristic of Luvic subsoil.\u003c/p\u003e\n\u003cp\u003eHorizon B4 was predominantly confined to the De Onderstepoort Nature Reserve, occurring along the foothills of the quartzite mountains. These quartzite-derived soils are typically coarse-textured, acidic and nutrient-poor due to the parent material\u0026rsquo;s resistance to weathering. In contrast, horizon B5 (very high clay, high Fe) was the most frequently predicted subsoil, occupying planar surfaces across much of the Highveld. Horizon B6 (high clay, high Mg) occurred east of Pretoria and exhibited a structured distribution along terrain transects, while horizons B7 and B8 were limited to a few localized occurrences in the northeast. Horizon B9 was similarly restricted to the northeastern part of the province, mainly in and around the Ezemvelo Game Reserve.\u003c/p\u003e\n\u003cp\u003eThe taxonomic system showed a similar distribution (Figure 4b), which was more diverse in the northeastern areas of Gauteng. Predictions followed clear toposequences at the 90 m resolution, reflecting consistent hillslope transitions across much of the province. As with B5 in the numerical system, Rhodic horizons were predicted to dominate much of Gauteng, which were prevalent in the west, encircled Johannesburg to the south and extended northeastward across planar surfaces. This correspondence is noteworthy, as both B5 and Rhodic horizons are characterized by high Fe content, suggesting that the numerical and taxonomic systems classified analogous subsoil. Further east, Xanthic horizons appeared on comparable surfaces and extended into the Johannesburg area, although these regions were marked by more dissected topography.\u003c/p\u003e\n\u003cp\u003eThe distribution of the Cambic horizon closely resembled that of B1, occurring mainly in high-relief areas and within Johannesburg, but absent from low-lying valleys. As expected, these valley positions were instead occupied by Aquic horizons, which were not represented in the numerical system, highlighting an area for improvement in capturing redox morphology in the numeric system. Along the eastern foothills, typically above the elevation of Aquic zones, Luvic horizons were predicted and frequently associated with A5/Vertic topsoil horizons between Pretoria and Soshanguve. Plinthic horizons occurred intermittently near the transitional boundaries of Cambic and Rhodic/Xanthic horizons across the province. As anticipated, Hardpan and Podzolic horizons were only occasionally predicted and exhibited no consistent spatial pattern.\u003c/p\u003e\n\u003ch2\u003e3.3\u0026nbsp; \u0026nbsp; \u0026nbsp;Ecosystem services and limitations\u003c/h2\u003e\n\u003cp\u003eBoth the numerical and taxonomic classification systems identified a type of Ochric horizon as the predominant topsoil across Gauteng. However, this horizon encompasses a wide range of physical, chemical and biological conditions and insufficient differentiation hides important functional variation. The topsoil represents a zone of high biological activity and strong interactions with the atmosphere and biosphere, and inadequate detail limits interpretation of key soil functions such as nutrient cycling (Ding et al., 2024), carbon sequestration (Mason et al., 2023) and aggregate stability (Liu et al., 2024). This drawback is especially significant given that the topsoil forms a key connection between ecosystem regulation and human health.\u003c/p\u003e\n\u003cp\u003eThe A2 horizon, characterised by high fertility and SOC content, supports multiple soil functions related to crop productivity (Ma et al., 2023), erosion control (Rugendo et al., 2023) and biodiversity maintenance (Wang et al., 2023). In urban contexts, A2 may further contribute to temperature regulation and energy balance through enhanced water retention (Ram\u0026iacute;rez et al., 2023). Despite this multifunctionality, A2 topsoil was predicted only intermittently. Their under-representation likely reflects the generalisation inherent in both systems rather than their true scarcity. The A5 horizon shares similar structural and textural characteristics with A2 but is alkaline, reducing its agricultural suitability. Nonetheless, its high clay, CEC and SOC content may provide regulating services through adsorption of organic pollutants (Ewis et al., 2022), particularly along the rapidly closing urban corridor between Soshanguve and Pretoria. The failure of both classification systems to distinguish topsoil variation reflects a broader global challenge in soil classification.\u003c/p\u003e\n\u003cp\u003eIn other words, soil classification systems, whether South Africa\u0026rsquo;s or those used globally, tend to define topsoil by exclusion. For example, in the USDA Soil Taxonomy (Soil Survey Staff, 2014), strict diagnostic criteria govern Mollic, Melanic, Umbric, Anthropic, Plaggen and Histic epipedons, resulting in most topsoil being classified as Ochric, regardless of its distinct physical or biological characteristics. Consequently, the true multifunctionality of soil cannot be fully represented through categorical mapping as seen in Figure 3. In the present study, the numerical framework exhibited a similar limitation, mirroring the constraints of traditional taxonomic systems. This challenge extends beyond pedology, influencing how other scientific disciplines, policy frameworks and stakeholders perceive and apply soil information. Nonetheless, numerical systems offer a distinct advantage: they can be revised, optimised and evaluated far more rapidly than national taxonomic systems, providing a flexible option toward a system that better reflects ecosystem services in the topsoil.\u003c/p\u003e\n\u003cp\u003eFor the subsoil, the Rhodic and Xanthic horizons predominated, representing mildly acidic horizons enriched in Fe oxides. The Rhodic horizon, with a higher hematite:goethite ratio (Schwertmann, 1971), exhibits slightly stronger reactivity towards organic pollutants (Voelz et al., 2018), whereas goethite rich Xanthic horizons are more reactive to phosphorus (Colombo et al., 1991). However, texture appears to be the major differentiating factor between these two horizons in terms of ecosystem services, yet this is not explicitly recognised in the taxonomic system but is in the numerical (e.g., B1 vs. B2). Regardless, both horizons generally support high agricultural productivity and support carbon stabilisation (He et al., 2025); however, when overlain by an Ochric topsoil they may undergo bleaching (Clarke et al., 2020), which promotes surface crusting and erosion. In most circumstances, Rhodic and Xanthic horizons are suitable for metropolitan expansion and can contribute to buffering or attenuating contaminants within the urban landscape.\u003c/p\u003e\n\u003cp\u003eThe light texture and low Fe content of B1 subsoil suggests unstable clays with limited regulating capacity, although intensive urban horticulture, particularly in Sandton city, may locally enhance infiltration and reduce surface heat (Wu and Congreves, 2024). The balanced sand fraction of the B1 horizon also indicates potential suitability for sustainable local building materials (Aaquib et al., 2018), linking soil properties to ecosystem services beyond agriculture and the environment.\u003c/p\u003e\n\u003cp\u003eFrom a land-use perspective, Aquic and Luvic horizons are unsuitable for urban development owing to their hydromorphic and shrink\u0026ndash;swell behaviour. Aquic horizons indicate a shallow water table, while Luvic horizons favour lateral subsurface flow (Van Tol et al., 2011) and may cause structural damage through expansion and contraction processes, depending on their mineralogy and dominant cation species (Sabtan, 2005). These areas are therefore better reserved for maintained natural vegetation due to their inherent water cycling capacity (Basset et al., 2023), keeping their true inherent ability to perform ecosystem services. The numerical classification system misclassified many of these hydromorphic areas as B1 horizons, which are typically associated with high-relief terrain. This confirms that the numerical framework lacks diagnostic sensitivity to colour, structure and drainage attributes, a limitation previously noted by Flynn et al. (2021) and emphasized by Flynn et al. (2025). Inclusion of such morphological features would improve the system\u0026rsquo;s capability to represent hydrological functions, and perhaps more globally relevant, greenhouse emissions such as methane emissions (Le Mer and Roger, 2001).\u003c/p\u003e\n\u003cp\u003eAlthough the taxonomic subsoil achieved only modest accuracy (54%) and agreement (\u0026kappa; = 0.40), its interpretive strength demonstrates that fundamental soil elements remain essential for linking pedological understanding to ecosystem services. However, modern pedometric techniques are essential to implement, interpret, visualise, and continuously refine such data-driven frameworks. For instance, emerging Earth-observation embedding models such AlphaEarth (Brown et al., 2025) and TESSERA (Feng et al., 2025) offer uncorrelated, multitemporal and high-resolution (10 m) representations that surpass the informational capacity of conventional covariates. Their integration could markedly improve predictive accuracy and ecological interpretability, particularly in heterogeneous urban landscapes where the interactions between soil processes and anthropogenic activity are subtle yet profoundly dynamic.\u003c/p\u003e\n\u003cp\u003eUltimately, this study demonstrated that while both systems effectively captured the spatial variability of soil, neither fully represented its multifunctionality. Moving beyond static or property-based numerical classifications toward function-oriented frameworks that integrate soil dynamics, hydromorphic behaviour and possibly features from taxonomic systems will enable a more explicit linkage between soil horizons and the expression of ecosystem services. Such integration bridges pedology and a multidisciplinary approach, reaffirming soil\u0026rsquo;s role not merely as a biophysical substrate but as a living, regulating system fundamental to climate resilience, agricultural productivity and human wellbeing in an increasingly urbanised world.\u003c/p\u003e"},{"header":"4. Conclusion","content":"\u003cp\u003eThis study evaluated a numerical and South Africa\u0026rsquo;s taxonomic soil classification for their ability to represent soil variability and ecosystem service potential within the urban expanses of Gauteng, South Africa. Both systems effectively captured spatial patterns but differed in interpretability and functional resolution. The numerical framework achieved strong quantitative accuracy (topsoil\u0026thinsp;=\u0026thinsp;85%, subsoil\u0026thinsp;=\u0026thinsp;68%), whereas the taxonomic system provided clearer ecological meaning grounded in pedological reasoning (clear toposequences). However, both frameworks generalised the topsoil predominantly as an Ochric topsoil, limiting the interpretation of biologically active surface processes. Subsoil differentiation proved far more informative: Rhodic and Xanthic horizons reflected productive, Fe-rich conditions, while Aquic and Luvic horizons indicated hydromorphic constraints for urban development but remain integral components of the regional water cycle. Although the taxonomic subsoil achieved a moderate accuracy (54%) and agreement (κ\u0026thinsp;=\u0026thinsp;0.40), its interpretive strength demonstrated the value of diagnostic horizons for linking soil structure, composition and function. Future frameworks that integrate diagnostic and functional properties, supported by advances in high-resolution, multitemporal Earth observation embeddings, could help connect numerical classification with ecosystem understanding. Such developments will enhance the capacity of soil mapping to inform sustainable land management, climate adaptation and ecosystem-service planning across rapidly urbanising landscapes.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAcknowledgements\u003c/h2\u003e\u003cp\u003eWe are grateful to the Agriculture Research Council-Institute of South Africa for providing the data to conduct this study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAaquib A, Nandurkar B, Bhagat R, Raut J, Ganvir V, Agrawal V, Kedar A, Sahare P (2018) Influence of Sand Particles on Strength and Durability of Mortar (1:3). SJ Impact Factor: 6 887\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAwumbila M (2017) Drivers of Migration and Urbanization in Africa: Key Trends and Issues\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBasset C, Abou Najm M, Ghezzehei T, Hao X, Daccache A (2023) How does soil structure affect water infiltration? A meta-data systematic review. Soil Tillage Res 226:105577. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.still.2022.105577\u003c/span\u003e\u003cspan address=\"10.1016/j.still.2022.105577\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBezdek JC, Ehrlich R, Full W (1984) FCM: The fuzzy c-means clustering algorithm. Comput Geosci 10:191\u0026ndash;203. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/0098-3004(84)90020-7\u003c/span\u003e\u003cspan address=\"10.1016/0098-3004(84)90020-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBhatta B (2010) Causes and Consequences of Urban Growth and Sprawl. 17\u0026ndash;36. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-642-05299-6_2\u003c/span\u003e\u003cspan address=\"10.1007/978-3-642-05299-6_2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBrown CF, Kazmierski MR, Pasquarella VJ, Rucklidge WJ, Samsikova M, Zhang C, Shelhamer E, Lahera E, Wiles O, Ilyushchenko S, Gorelick N, Zhang LL, Alj S, Schechter E, Askay S, Guinan O, Moore R, Boukouvalas A, Kohli P (2025) AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2507.22291\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2507.22291\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCarr\u0026eacute; F, Jacobson M (2009) Numerical classification of soil profile data using distance metrics. Geoderma 148:336\u0026ndash;345\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eClarke CE, Roux JL, Ellis F, Clercq WPD, Waals JHVD (2020) Pedogenesis of bleached topsoils occurring on weakly structured, high chroma subsoils in South Africa. CATENA 193:104634\u0026ndash;104634. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.catena.2020.104634\u003c/span\u003e\u003cspan address=\"10.1016/j.catena.2020.104634\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCohen J (1960) A Coefficient of Agreement for Nominal Scales. Educ Psychol Meas 20:37\u0026ndash;46. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/001316446002000104\u003c/span\u003e\u003cspan address=\"10.1177/001316446002000104\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eColombo C, Buondonno A, Violante A, Torrent J (1991) The Contrasting Effect of Goethite and Hematite on Phosphate Sorption and Desorption by Terre Rosse. Z f\u0026uuml;r Pflanzenern\u0026auml;hrung und Bodenkunde 154:301\u0026ndash;305. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/jpln.19911540410\u003c/span\u003e\u003cspan address=\"10.1002/jpln.19911540410\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDing Y, Gao X, Shu D, Siddique KHM, Song X, Wu P, Li C, Zhao X (2024) Enhancing soil health and nutrient cycling through soil amendments: Improving the synergy of bacteria and fungi. Sci Total Environ 923:171332. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.scitotenv.2024.171332\u003c/span\u003e\u003cspan address=\"10.1016/j.scitotenv.2024.171332\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEwis D, Ba-Abbad MM, Benamor A, El-Naas MH (2022) Adsorption of organic water pollutants by clays and clay minerals composites: A comprehensive review. Appl Clay Sci 229:106686. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.clay.2022.106686\u003c/span\u003e\u003cspan address=\"10.1016/j.clay.2022.106686\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFeng Z, Atzberger C, Jaffer S, Knezevic J, Sormunen S, Young R, Lisaius MC, Immitzer M, Jackson T, Ball J, Coomes DA, Madhavapeddy A, Blake A, Keshav S (2025) TESSERA: Precomputed FAIR Global Pixel Embeddings for Earth Representation and Analysis. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2506.20380\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2506.20380\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFey M, Hughes J, Lambrechts J, Dohse T (2001) Chapter 2: The soil groups: distribution, properties, classification, genesis and use, in: Soils of South Africa. pp. 9\u0026ndash;10. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/CBO9780511491757.014\u003c/span\u003e\u003cspan address=\"10.1017/CBO9780511491757.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFick SE, Hijmans RJ (2017) WorldClim 2: new 1km spatial resolution climate surface for global land areas. Int J Climatol 37:4302\u0026ndash;4315\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFinisdore J, Rhodes C, Haines-Young R, Maynard S, Wielgus J, Dvarskas A, Houdet J, Qu\u0026eacute;tier F, Lamothe KA, Ding H, Soulard F, Van Houtven G, Rowcroft P (2020) The 18 benefits of using ecosystem services classification systems. Ecosyst Serv 45:101160\u0026ndash;101160. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ecoser.2020.101160\u003c/span\u003e\u003cspan address=\"10.1016/j.ecoser.2020.101160\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFiorentino A, Rajput FZ, Di Serio A, Baldi V, Guarino F, Baldantoni D, Ronga D, Mazzei P, Motta O, Falanga M, Cicatelli A, Castiglione S (2025) Role of Plants and Urban Soils in Carbon Stock: Status, Modulators, and Sustainable Management Practices. Plants 14:546. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/plants14040546\u003c/span\u003e\u003cspan address=\"10.3390/plants14040546\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFlynn T, Clarke C, Kostecki R, Rebi A (2025) A comparison of country-scale subsoil predictions between a numeric and a taxonomic soil classification system. Geoderma Reg 40:e00902. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.geodrs.2024.e00902\u003c/span\u003e\u003cspan address=\"10.1016/j.geodrs.2024.e00902\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFlynn T, de Clercq W, Rozanov A, Clarke C (2019) High-resolution digital soil mapping of multiple soil properties: an alternative to the traditional field survey? South Afr J Plant Soil 1\u0026ndash;11. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/02571862.2019.1570566\u003c/span\u003e\u003cspan address=\"10.1080/02571862.2019.1570566\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFlynn T, Kostecki R (2024) Spatial downscaling of global soil texture classes into 30 m images at the province scale. Geomatica 76:100028. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.geomat.2024.100028\u003c/span\u003e\u003cspan address=\"10.1016/j.geomat.2024.100028\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFlynn T, Triantafilis J, Rozanov A, Ellis F, L\u0026aacute;zaro-L\u0026oacute;pez A, Watson A, Clarke C (2021) Numerical soil horizon classification from South Africa\u0026rsquo;s legacy database. CATENA 206:105543. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.catena.2021.105543\u003c/span\u003e\u003cspan address=\"10.1016/j.catena.2021.105543\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFriedman J, Hastie T (1998) Additive Logistic Regression: a Statistical View of Boosting\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHartigan (1975) Clustering Algorithms. Wiley, New York\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHe Y, Wang Z, Zhu J, Lin X, Qi J (2025) Soil Carbon Sequestration: Role of Fe Oxides and Polyphenol Oxidase Across Temperature and Cultivation Systems. Plants 14:927. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/plants14060927\u003c/span\u003e\u003cspan address=\"10.3390/plants14060927\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHerold M, Mayaux P, Woodcock CE, Baccini A, Schmullius C (2008) Some challenges in global land cover mapping: An assessment of agreement and accuracy in existing 1 km datasets. Remote Sens Environ 112:2538\u0026ndash;2556. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rse.2007.11.013\u003c/span\u003e\u003cspan address=\"10.1016/j.rse.2007.11.013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHorton R, Wierenga PJ (1983) Estimating the Soil Heat Flux from Observations of Soil Temperature Near the Surface. Soil Sci Soc Amer J 47:14\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2136/sssaj1983.03615995004700010003x\u003c/span\u003e\u003cspan address=\"10.2136/sssaj1983.03615995004700010003x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHossain Anni A, Cohen S, Praskievicz S (2020) Sensitivity of urban flood simulations to stormwater infrastructure and soil infiltration. J Hydrol 588:125028\u0026ndash;125028. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jhydrol.2020.125028\u003c/span\u003e\u003cspan address=\"10.1016/j.jhydrol.2020.125028\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKarthika V, Sekaran U, Jainullabudeen GB, Nagarathinam A (2022) Advances in bioremediation of industrial wastewater containing metal pollutants. Biological Approaches to Controlling Pollutants. Elsevier, pp 163\u0026ndash;177. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/B978-0-12-824316-9.00001-X\u003c/span\u003e\u003cspan address=\"10.1016/B978-0-12-824316-9.00001-X\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKaufman L, Rousseeuw PJ (1960) Clustering by means of medoids. In: Dodge Y (ed) Statisical Data Analysis Based on the L1 Norm. North-Holland, Amsterdam, pp 405\u0026ndash;416\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKonovalov A, Krajbich I (2018) Neurocomputational Dynamics of Sequence Learning. Neuron 98:1282\u0026ndash;1293e4. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.neuron.2018.05.013\u003c/span\u003e\u003cspan address=\"10.1016/j.neuron.2018.05.013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKuhn M, Johnson K (2013) Applied Predictive Modeling. Springer New York. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-1-4614-6849-3\u003c/span\u003e\u003cspan address=\"10.1007/978-1-4614-6849-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLand Type Survey Staff (2006) Land Types of South Africa on 1:250 000 scale. Pretoria, South Africa\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLandis JR, Koch GG (1977) The Measurement of Observer Agreement for Categorical Data. Biometrics 33:159\u0026ndash;174\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLe Mer J, Roger P (2001) Production, oxidation, emission and consumption of methane by soils: A review. Eur J Soil Biol 37:25\u0026ndash;50. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S1164-5563(01)01067-6\u003c/span\u003e\u003cspan address=\"10.1016/S1164-5563(01)01067-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi P, Wang Z-H (2021) Environmental co-benefits of urban greening for mitigating heat and carbon emissions. J Environ Manage 293:112963\u0026ndash;112963. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jenvman.2021.112963\u003c/span\u003e\u003cspan address=\"10.1016/j.jenvman.2021.112963\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu S, Lin Z, Duan X, Deng Y (2024) Effects of soil microorganisms on aggregate stability during vegetation recovery in degraded granitic red soil areas. Appl Soil Ecol 204:105734. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.apsoil.2024.105734\u003c/span\u003e\u003cspan address=\"10.1016/j.apsoil.2024.105734\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu Y, Meng Q, Zhang L, Wu C (2022) NDBSI: A normalized difference bare soil index for remote sensing to improve bare soil mapping accuracy in urban and rural areas. CATENA 214:106265\u0026ndash;106265. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.catena.2022.106265\u003c/span\u003e\u003cspan address=\"10.1016/j.catena.2022.106265\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMa Y, Woolf D, Fan M, Qiao L, Li R, Lehmann J (2023) Global crop production increase by soil organic carbon. Nat Geosci 16:1159\u0026ndash;1165. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41561-023-01302-3\u003c/span\u003e\u003cspan address=\"10.1038/s41561-023-01302-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMagiera T, Strzyszcz Z, Rachwal M (2007) Mapping particulate pollution loads using soil magnetometry in urban forests in the Upper Silesia Industrial Region, Poland. For Ecol Manag 248:36\u0026ndash;42. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.foreco.2007.02.034\u003c/span\u003e\u003cspan address=\"10.1016/j.foreco.2007.02.034\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMason ARG, Salomon MJ, Lowe AJ, Cavagnaro TR (2023) Microbial solutions to soil carbon sequestration. J Clean Prod 417:137993. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jclepro.2023.137993\u003c/span\u003e\u003cspan address=\"10.1016/j.jclepro.2023.137993\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMiller MAE, Shepherd KD, Kisitu B, Collinson J (2021) iSDAsoil: The first continent-scale soil property map at 30 m resolution provides a soil information revolution for Africa. PLoS Biol 19:e3001441\u0026ndash;e3001441. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pbio.3001441\u003c/span\u003e\u003cspan address=\"10.1371/journal.pbio.3001441\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eM\u0026oslash;ller AB, Beucher AM, Pouladi N, Humlekrog Greve M (2020) Oblique geographic coordinates as covariates for digital soil mapping. Soil 6:269\u0026ndash;289. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5194/soil-6-269-2020\u003c/span\u003e\u003cspan address=\"10.5194/soil-6-269-2020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNASA JPL (2022) NASADEM Merged DEM Global 1 arc second V001. NASA EOSDIS Land Processes DAAC\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eO\u0026rsquo;Riordan R, Davies J, Stevens C, Quinton JN, Boyko C (2021) The ecosystem services of urban soils: A review. Geoderma 395:115076\u0026ndash;115076. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.geoderma.2021.115076\u003c/span\u003e\u003cspan address=\"10.1016/j.geoderma.2021.115076\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRam\u0026iacute;rez PB, Machado S, Singh S, Plunkett R, Calder\u0026oacute;n FJ (2023) Addressing the effects of soil organic carbon on water retention in US Pacific Northwest wheat\u0026ndash;soil systems. Front Soil Sci 3:1233886. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fsoil.2023.1233886\u003c/span\u003e\u003cspan address=\"10.3389/fsoil.2023.1233886\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eReddi LN, Jain AK, Yun H-B (2012) Soil materials for earth construction: properties, classification and suitability testing. Modern Earth Buildings. Elsevier, pp 155\u0026ndash;171. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1533/9780857096166.2.155\u003c/span\u003e\u003cspan address=\"10.1533/9780857096166.2.155\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRubio CM, Ubeda X, Ferrer F (2012) Response of the Thermal Conductivity as a Function of Water Content of a Burnt Mediterranean Loam Soil. OJSS 02, 1\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4236/ojss.2012.21001\u003c/span\u003e\u003cspan address=\"10.4236/ojss.2012.21001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRugendo MK, Gichimu BM, Mugwe JN, Mucheru-Muna M, Mugendi DN (2023) Surface runoff and soil erosion from Nitisols and Ferralsols as influenced by different soil organic carbon levels under simulated rainfall conditions. Heliyon 9:e17684. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.heliyon.2023.e17684\u003c/span\u003e\u003cspan address=\"10.1016/j.heliyon.2023.e17684\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSabtan AA (2005) Geotechnical properties of expansive clay shale in Tabuk, Saudi Arabia. J Asian Earth Sci 25:747\u0026ndash;757. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jseaes.2004.07.003\u003c/span\u003e\u003cspan address=\"10.1016/j.jseaes.2004.07.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchwertmann U (1971) Transformation of Hematite to Goethite in Soils. Nature 232:624\u0026ndash;625. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/232624a0\u003c/span\u003e\u003cspan address=\"10.1038/232624a0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShi T, Hu X, Guo L, Su F, Tu W, Hu Z, Liu H, Yang C, Wang J, Zhang J, Wu G (2021) Digital mapping of zinc in urban topsoil using multisource geospatial data and random forest. Sci Total Environ 792:148455\u0026ndash;148455. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.scitotenv.2021.148455\u003c/span\u003e\u003cspan address=\"10.1016/j.scitotenv.2021.148455\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSoil Classification Working Group (1991) Soil Classificatioin: a Taxonomic System for South Africa, 2nd edn. Department of Agricultural Development, Pretoria, South Africa\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSoil Survey Staff (2014) Keys to soil taxonomy, 12th edn. USDA-Natural Resources Conservation Service, Washington, DC\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTriantafilis J, Ward WT, Odeh IOA, McBratney AB (2001) Creation and Interpolation of Continuous Soil Layer Classes in the Lower Namoi Valley. Soil Sci Soc Am J 65:403\u0026ndash;413. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2136/sssaj2001.652403x\u003c/span\u003e\u003cspan address=\"10.2136/sssaj2001.652403x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eUSGS (2021) Landsat collection 2\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVan Tol JJ, Le Roux PA, Hensley M (2011) Soil Indicators of Hillslope Hydrology. Principles-Application Assessments Soil Sci 209\u0026ndash;240. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5772/1860\u003c/span\u003e\u003cspan address=\"10.5772/1860\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVan Zijl G, Le Roux P (2014) Creating a conceptual hydrological soil response map for the Stevenson Hamilton Research Supersite, Kruger National Park, South Africa. Water SA 40:331\u0026ndash;336. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4314/wsa.v40i2.15\u003c/span\u003e\u003cspan address=\"10.4314/wsa.v40i2.15\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVan Zijl G, van Tol J, Tinnefeld M, Le Roux P (2019) A hillslope based digital soil mapping approach, for hydropedological assessments. Geoderma 354:113888\u0026ndash;113888. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.geoderma.2019.113888\u003c/span\u003e\u003cspan address=\"10.1016/j.geoderma.2019.113888\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVan Zijl GM, Le Roux PA, Turner DP (2013) Disaggregation of land types using terrain analysis, expert knowledge and GIS methods. South Afr J Plant Soil 30:123\u0026ndash;129. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/02571862.2013.806679\u003c/span\u003e\u003cspan address=\"10.1080/02571862.2013.806679\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVilla P, Malucelli F, Scalenghe R (2018) Sci Total Environ 612:590\u0026ndash;604. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.scitotenv.2017.08.250\u003c/span\u003e\u003cspan address=\"10.1016/j.scitotenv.2017.08.250\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Multitemporal mapping of peri-urban carbon stocks and soil sealing from satellite data\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVoelz JL, Arnold WA, Penn RL (2018) Redox-induced nucleation and growth of goethite on synthetic hematite nanoparticles. Am Mineral 103:1021\u0026ndash;1029. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2138/am-2018-6342\u003c/span\u003e\u003cspan address=\"10.2138/am-2018-6342\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang Y, Wang S, Liu C, Zhu E, Jia J, Feng X (2023) Shifting relationships between SOC and molecular diversity in soils of varied carbon concentrations: Evidence from drained wetlands. Geoderma 433:116459. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.geoderma.2023.116459\u003c/span\u003e\u003cspan address=\"10.1016/j.geoderma.2023.116459\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWebster R, Burrough PA (1972) Computer-based soil mapping of small areas from sample data II classification smoothing. J Soil Sci 23:222\u0026ndash;234. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.1365-2389.1972.tb01655.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1365-2389.1972.tb01655.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWeininger D, Weininger A, Weininger JL (1989) SMILES. 2. Algorithm for generation of unique SMILES notation. J Chem Inf Comput Sci 29:97\u0026ndash;101. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/ci00062a008\u003c/span\u003e\u003cspan address=\"10.1021/ci00062a008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWu Q, Congreves KA (2024) Soil health benefits associated with urban horticulture. Sci Total Environ 912:168852. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.scitotenv.2023.168852\u003c/span\u003e\u003cspan address=\"10.1016/j.scitotenv.2023.168852\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Y, Hartemink AE (2019) A method for automated soil horizon delineation using digital images. Geoderma 343:97\u0026ndash;115. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.geoderma.2019.02.002\u003c/span\u003e\u003cspan address=\"10.1016/j.geoderma.2019.02.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhu AX (1997) A similarity model for representing soil spatial information. Geoderma 77:217\u0026ndash;242\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 to 5 are available in the Supplementary Files section\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"University of Fort Hare","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Digital soil mapping, Soil services, Multifunctionality, Urban soils","lastPublishedDoi":"10.21203/rs.3.rs-8006960/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8006960/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eUrban areas are strongly influenced by the ecosystem services provided by soil, which help mitigate environmental pressures generated by metropolises themselves and protect surrounding natural and agricultural landscapes. However, it remains unclear whether existing soil classification systems adequately capture the complex multifunctionality of soil that produce these services. This study examined the spatial patterns and ecosystem services represented in a numerical and South Africa\u0026rsquo;s taxonomic (\u0026ldquo;Blue Book\u0026rdquo;) soil classification system in Gauteng, South Africa (~\u0026thinsp;26.3\u0026deg; S, 28.1\u0026deg; E), a region characterised by extensive urban development. A Gradient Tree Boosting model implemented in Google Earth Engine was used to predict topsoil and subsoil horizons for both systems using identical soil observations. The model achieved high spatial accuracy for topsoil and moderate accuracy for subsoil horizons in both systems. Despite this, the topsoil predictions of both systems (accuracy\u0026thinsp;\u0026gt;\u0026thinsp;85%) showed limited relevance to ecosystem services across the urban landscape. In contrast, the subsoil predictions (accuracy\u0026thinsp;~\u0026thinsp;54\u0026ndash;68%) of the taxonomic system exhibited clear spatial patterns, greater interpretability and stronger ecological relevance. The primary limitation identified was the over-classification of the Ochric topsoil together with the moderate accuracy of the subsoil, obscured functional differences and constrained the interpretation of soil-based ecosystem services across Gauteng. Nevertheless, emerging numerical frameworks that integrate categorical data and dynamic processes, supported by advances in Earth observation and machine learning, are expected to enhance the ability to map soil multifunctionality and its representation of ecosystem services.\u003c/p\u003e","manuscriptTitle":"Evaluating diagnostic horizons as indicators of ecosystem services through digital soil mapping in urban Gauteng, South Africa","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-04 05:25:02","doi":"10.21203/rs.3.rs-8006960/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"4ad1bf6a-8fbc-48f1-9983-e194f7c96cfd","owner":[],"postedDate":"November 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":57276942,"name":"Agronomy"}],"tags":[],"updatedAt":"2025-11-04T05:25:02+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-04 05:25:02","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8006960","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8006960","identity":"rs-8006960","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}