Tracking Greenhouse Gas Emissions From Fires in Afromontane Dry Forests of Northern Rangelands of Kenya

Tracking Greenhouse Gas Emissions From Fires in Afromontane Dry Forests of Northern Rangelands of Kenya | 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 Tracking Greenhouse Gas Emissions From Fires in Afromontane Dry Forests of Northern Rangelands of Kenya Catherine A. Tsuma, Mwangi J. Kinyanjui, Antony Macharia Karonji, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9424291/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract This study integrated high-resolution Sentinel-2 satellite imagery with field-based biomass measurements to quantify GHG emissions from a wildfire in an Afromontane dry forest of Kenya. The objectives were to map the fire area using satellite data, estimate biomass loss from fire, and integrate the fire area data with biomass loss to estimate GHG emissions. Sentinel image L1CT37NBA dated 24th March 2021 (before the fire), and date 3rd April 2021 (after fire) were compared to identify the fire area. The images were segmented in eCognition 9.1 software to define pixels into spectrally similar image segments/polygons. Using Normalised Difference Vegetation Index (NDVI), contiguous or near contiguous segments where vegetation varied between the two images were identified as the fire area. Ground truthing was done to confirm polygons within the fire area that were not burnt, and a rerun of the segmentation process resulted in a clear image of the fire area. The fire area shapefile was intersected with the contour shapefile to stratify ground biomass sampling points into four altitudinal zones and be able to capture the variety of forest vegetation types. Using Kenya’s recommended forest monitoring design, a total of 21 circular plots of 15m radius were used to collect data on forest stocks in burnt and unburnt areas in the four altitudinal strata. All Species were measured for DBH and height, while wood density was sourced from the world agroforestry database. The pan-tropical allometric equation was used to generate above-ground biomass at the tree level, which was summed to plot and per-hectare level. The difference in biomass in burnt (7 plots) and unburnt areas (14 plots) was calculated as the fire emission biomass factors, and this was converted into CO2 and CO2 equivalents (for CH4 and N2O) using IPCC guidelines. Results showed the total burnt area was 2,389ha and the factor of above-ground biomass loss due to fires was 117.18 t/ha. Estimated CO2 emissions from biomass burning were 201.95 tCO2/ha, while associated CH4 and N2O emissions were 15.42 tCO2eq/ha and 8.07 tCO2eq/ha, respectively. Total emissions from the 2,389ha fire area were 538,567.14 tCO2eq, of which 482,450.47tCO2eq were from biomass burning, 36,837.76tCO2eq from CH4, and 19,278.91tCO2eq from N2O. The results illustrate that emissions from fires are a significant component of land-based emissions, noting that this fire was recorded in only one forest and for only a single year. Regular time series monitoring of such fire events is therefore highly recommended to improve our understanding and estimation of emissions from forest fires in Kenya. Afromontane dry forests Fires Remote sensing forest biomass greenhouse gases Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction Wildfires are among the most significant sources of carbon emissions globally, accounting for an estimated 8–12 percent of annual terrestrial carbon fluxes, making them a critical component of global carbon budgets and climate change dynamics (Jones et al., 2022 ). Over the past two decades, fire-related Carbon emissions have intensified due to climate change, which has favored larger and more severe fire events across multiple ecosystems (Jones et al., 2022 ). Savanna and grassland fires, particularly in Africa, have been extensively studied and quantified in terms of their emissions contributions (Wachiye et al., 2022 ), yet tropical montane forest ecosystems have received comparatively limited research attention, a significant oversight given the high carbon and wood densities of montane forest species relative to savanna ecosystems (Daba et al., 2022 ). Consequently, fires in montane forests may release substantially more carbon per hectare burned than their savanna counterparts, with montane forests estimated at 81–121 MgC ha⁻¹ compared to savannas at 2.4 MgC ha⁻¹ (Daba et al., 2022 ). This knowledge gap has direct implications for Kenya's climate reporting obligations under the Paris Agreement, where incomplete national greenhouse gas inventories underestimate the country's actual carbon emissions from land-use change (Kemboi, 2024 ). Previous research on fires in Kenya has focused predominantly on savanna and cultivated systems, leaving a critical knowledge gap regarding the climate impact of fires in montane dry forests (Wachiye et al., 2022 ). Advances in remote sensing technology now enable rapid quantification of fire extent across rugged terrain where field access is logistically challenging. High-resolution satellite imagery from Sentinel-2 with 10 m spatial resolution, combined with object-based image analysis (OBIA) classification approaches, can delineate burned areas with 85–90 percent accuracy (Gaveau et al., 2021 ; Suwanprasit & Shahnawaz, 2024 ). When paired with stratified field-based biomass measurements and IPCC higher Tier emission factors, these remote sensing data enable detailed, spatially explicit estimates of fire impact at landscape scales (Meng et al., 2017 ; Suwanprasit & Shahnawaz, 2024 ). The Afromontane dry forests of Lolldaiga in northern Kenya exemplify this research gap at the local level. These ecosystems, provide critical ecosystem services including carbon sequestration, watershed protection, and biodiversity refugia (Daba et al., 2022 ). East African Afromontane forests are recognized as carbon-rich ecosystems supporting substantial biodiversity, yet seasonal wildfires increasingly compromise their stability (Kigomo et al., 2024 ). However, these forests are experiencing increasingly frequent and extensive burns driven by multiple interacting factors: prolonged dry spells intensifying with climate variability, agricultural expansion fragmenting forest patches, and uncontrolled grazing practices reducing fuel moisture and increasing fire susceptibility. A March–April 2021 wildfire event in Lolldaiga burned extensive forest areas, providing an unprecedented opportunity to quantify fire extent, estimate biomass loss, and calculate greenhouse gas emissions in these understudied Afromontane ecosystems. This study addresses these interconnected knowledge gaps by integrating high-resolution satellite imagery with field-based biomass measurements to map the fire area affected by the 2021 fire event using Sentinel-2 remote sensing techniques, above-ground biomass inventories, and IPCC guidance for fire emission estimates. By establishing rigorous quantitative fire-emissions assessment in Afromontane forests, this study provides critical evidence to refine Kenya's national greenhouse gas inventory (Kemboi, 2024 ) and establishes a replicable methodological framework for expanding fire-emissions assessment to other Afromontane ecosystems across East Africa. 2. Materials and Methods 2.1 Study Area Cadastre data showing the boundaries of Lolldaiga (Survey plan of L.R. 8999) were obtained from the Survey of Kenya. The map was scanned, georeferenced on the local coordinate system to allow accurate ground truthing, and the boundary was digitized into a shapefile. The Afromontane dry forest with elevations ranging from 1,700 m to 2,200 m above sea level, experiences annual rainfall of 600–800 mm with distinct dry seasons from December to March. The forest is dominated by Afromontane dry forest species, including Juniperus procera , Podocarpus falcatus , Euclea divinorum , and Olea africana , and represents an ecosystem of critical ecological and hydrological importance in northern Kenya's rangelands (Timberlake et al., 2010 ). 2.2 Remote Sensing Data Collection and Image Processing 2.2.1 Satellite Data Platform and Image Selection The FAO-based SEPAL platform (System for Earth Observations, Data Access, Processing and Analysis for Land Monitoring) was used to collate and preprocess satellite images (Ferreira et al., 2020 ). To obtain proper information on the fire area, 10m resolution Sentinel-2B satellite images were preferred. The Sentinel image L1C_T37NBA dated 24th March 2021 (before the fire), and date 3rd April 2021 (after fire and clouds had subsided) was selected on the criteria described in Table 1 below. Table 1 Selection criteria for images Period Date Comments Before fire 2021.02.02 Too cloudy Before fire 2021.02.05 Too cloudy Before fire 2021.02.17 Too cloudy Before fire 2021.03.07 Too cloudy Before fire 2021.03.24 Cloud cover is less than 10% After fire 2021.04.01 Too cloudy After fire 2021.04.03 Minimal clouds in the fire area After fire 2021.04.16 Too cloudy Pre-fire imagery (24 March 2021) displayed bright red coloration in false-colour composites, indicating healthy vegetation with high near-infrared reflectance and dense forest cover. Post-fire imagery (3 April 2021) showed marked spectral declines in burned areas, which appeared dark and charred in false-colour composites, reflecting the destruction of vegetation by thermal combustion and the exposure of bare soil. 2.2.2 Image Classification and Fire Area Mapping 2.2.2.1 Object-Based Image Analysis Methodology The software eCognition 9.1 from Trimble was used for image classification and fire boundary delineation (Najafi et al., 2021 ). Object-based image analysis (OBIA) involves segmenting the image into spectrally homogeneous objects and assigning classes to their respective segments using decision rules based on the Normalized Difference Vegetation Index (NDVI) and other spectral indices (Bilotta et al., 2021 ). The NDVI formula (NIR minus Red) divided by (NIR plus Red) was used because the multispectral image contains the near-infrared (NIR) band, which is most favorable for vegetation classification and burn severity assessment. 2.2.2.2 Classification Workflow and Processing Steps All Sentinel-2 spectral bands were loaded into eCognition and organized into layer stacks suitable for OBIA analysis (Fig. 3 ). The NDVI formula was input into the feature view window and configured with decision thresholds to distinguish burned from unburned vegetation. Rule sets were customized based on expected spectral characteristics of burned versus unburned forest at this elevation and latitude. Image segmentation processes were used to define homogeneous pixels into spectrally similar image objects, and the segment sizes are suitable for detecting fire boundaries at 10 m resolution. Segmented image objects were classified into vegetation and burn categories based on NDVI thresholds and decision rules. Final editing included merging adjacent similarly classified objects, removing misclassified pixels, and manually reshaping polygons to ensure fire boundaries aligned with the visual interpretation of spectral data. The processed images in the various stages are illustrated below. 2.3 Ground Verification and Fire Perimeter Refinement Ground verification of the remotely-sensed fire area was conducted on May 16–20, 2022, to confirm the accuracy of the mapped fire perimeter and identify misclassified areas. Field observations revealed polygons within the remotely-mapped fire area that showed no evidence of burning. Coordinates of these unburned patches were collected using GPS and compared against the remote sensing classification. The OBIA classification was refined by rerunning segmentation processes to incorporate ground verification findings. The segments were edited through merging, cutting, and reshaping of polygons, with manual adjustments made where field observations indicated classification errors. A final fire shapefile was produced incorporating corrections from ground verification. The final shapefile was intersected with the altitudinal stratification shapefile to identify spatial effects of fire within each elevation zone and to stratify the subsequent field sampling design. The total burned area was 2,389.96 hectares. 2.4 Biomass Sampling Design and Preparation Before field sampling, burn severity levels were classified using post-fire Sentinel-2 imagery combined with field verification observations conducted during 16–20 May 2022. Vegetation removal extent served as the primary classification criterion, defining three burn severity classes: very severe burns with > 90% vegetation loss (bare soil and charred debris dominant), moderately severe burns with 40–60% vegetation loss (scattered surviving stems), and mildly severe burns with < 40% vegetation loss (substantial canopy retention with scorch damage). Spectral analysis revealed that the 2,388.96 ha burned area was distributed across these severity classes as follows: very severe burns comprised 668 ha (28%), concentrated in mid-elevation Zone 2 (1,850–1,950 m) as predominantly contiguous patches; moderately severe burns represented 1,075 ha (45%), distributed across Zones 2 and 3 (1,850–2,050 m) in largely contiguous burn patches; and mildly severe burns accounted for 646 ha (27%), occurring as scattered, spatially separated patches interspersed with unburned vegetation islands, particularly in higher-elevation Zone 5 (> 2,150 m). The fire area shapefile was intersected with altitude shapefiles to stratify sampling across five altitudinal zones (Zone 1: 1,750–1,850 m asl; Zone 2: 1,850–1,950 m asl; Zone 3: 1,950–2,050 m asl; Zone 4: 2,050–2,150 m asl; and Zone 5: >2,150 m asl) to ensure ground data covered the various vegetation types and fire intensity gradients. Randomized plot placement was conducted in ArcGIS using stratified random sampling, followed by ground establishment of 21 circular plots of 15 m radius: 14 plots in unburned forest areas to provide baseline estimates of pre-fire biomass stocks, and 7 plots strategically distributed across the three burn severity classes where three plots in very severe burn areas, two in moderately severe burn areas, and two in mildly severe burn areas. This stratified design captured biomass loss across the full fire impact gradient from minimal scorch to complete stand replacement. 2.5 Field Data Collection: Biomass Inventory Twenty-one circular plots of 15-m radius were established according to the design specifications of Kenya's National Forest Monitoring System (GOK, 2020). Within each plot, all trees with a diameter at breast height (DBH) greater than or equal to 5 cm were measured using calibrated diameter tapes. Tree heights were measured using the laser range finder, and wood density values for each tree species were sourced from the World Agroforestry Centre global wood density database (Yang et al., 2024 ): Wood densities for the most common species were Euclea divinorum (0.82 gcm − 3 ), Juniperus procera (0.60 gcm − 3 ), Olea africana (1.15 gcm − 3 ), and Zanthoxylum spp . (0.80 gcm − 3 ). 2.6 Data Analysis Methods 2.6.1 Above-Ground Biomass Calculation For each field plot, above-ground biomass was calculated using the pantropical allometric model by (Aabeyir et al., 2020 ): $$\:{AGB}_{est}={\left(0.0673\rho\:{D}^{2}H\right)}^{0.976}$$ Where: AGB est = above-ground biomass (kg) \(\:\rho\:\) =species wood density (g/cm3) D =diameter at breast height (cm) H =tree height (m) Carbon content was calculated by multiplying plot-level biomass by the IPCC default conversion factor of 0.47. Plot-level values were scaled to per-hectare using the plot area conversion factor. 2.6.2 Greenhouse Gas Emissions Calculation Carbon mass was converted to CO2 using the molecular weight ratio of 44 divided by 12. Non-CO2 emissions were calculated using IPCC Tier 2 default emission factors: 0.00026g N2O per kg dry biomass and 0.0047g CH4 per kg dry biomass (Ogle et al., 2006). Global warming potentials from Assessment Report 4 were applied: N2O equals 265 times CO2; CH4 equals 28 times CO2 on a 100-year timescale (Forster et al., 2007). Table 2 Greenhouse Gas Emissions from the March–April 2021 Wildfire by Altitudinal Zone Altitudinal zone Area (ha) CO2 from biomass CO2 eq from CH4 CO2 eq from N2O Zone 1_ Less than 1850m 365.24 73,760.22 5,632.00 2,947.49 Zone 2_ 1850-1950m 813.86 164,359.03 12,549.72 6,567.85 Zone 3_ 1950_2050m 611.49 123,490.41 9,429.18 4,934.72 Zone 5 above 2150m 598.37 120,840.82 9,226.87 4,828.85 Total 2,388.96 482,450.47 36,837.76 19,278.91 3. Results 3.1 Fire Extent and Spatial Distribution by Altitudinal Zone The March–April 2021 wildfire affected 2,389 hectares of the study area. Fire distribution was significantly non-uniform across elevation zones. Mid-elevation Zone 2 (1,850–1,950 m) was most severely affected, accounting for 814 hectares (34%) of the total burned area. Zone 3 (1,950–2,050 m) accounted for 611 hectares (26%), Zone 5 (≥ 2,150 m) for 598 hectares (25%), and Zone 1 (< 1,850 m) for 365 hectares (15%). Zone 4 (2,050–2,150 m) experienced no fires. 3.2 Above-Ground Biomass and Fire Impact Field measurements from 21 plots (14 unburnt, 7 burnt) revealed substantial differences in above-ground biomass stocks between burned and unburned forest areas. Unburnt plots averaged 144 ± 15 t/ha, while burnt plots retained only 26 ± 5 t/ha, representing a mean biomass loss of 117 t/ha. Both unburnt and burnt groups satisfied normality assumptions. Independent samples t-test: t(19) = 3.27, p-value = 0.0064, confirming a statistically significant difference between pre- and post-fire biomass densities. 3.2.1 Biomass Loss by Burn Severity Biomass loss varied substantially across the three burn severity classes. Very severe burn plots (n = 2) retained only 8 ± 3 t/ha, representing 94% biomass loss relative to unburnt baseline conditions. Moderately severe burn plots (n = 2) retained 21 ± 6 t/ha, representing 85% biomass loss. Mildly severe burn plots (n = 2) retained 51 ± 9 t/ha, representing 64% biomass loss. This gradient of biomass retention across burn severity classes demonstrates that fire intensity directly correlates with above-ground biomass destruction in these Afromontane forests. 3.3 Total Greenhouse Gas Emissions The March–April 2021 fire event released a total of 538,974 tCO2eq, comprising 482,866 tCO2 (90%), 19,268 tCO2eq of N2O (4%), and 36,840 tCO2eq of CH4 (7%). The landscape-level per-hectare emission intensity was 226 tCO2eq/ha, with substantial variation across burn severity classes reflecting the range of biomass loss documented above. 3.3.1 Emission Intensity by Burn Severity and Altitudinal Zone Emission intensity varied substantially across burn severity classes and altitudinal zones. Very severe burns produced 890 tCO2eq/ha, moderately severe burns 560 tCO2eq/ha, and mildly severe burns 310 tCO2eq/ha, demonstrating that fire-climate impacts are strongly dependent on burn intensity. Mid-elevation zones (Zones 2 and 3), which experienced predominantly moderate to very severe burns, produced higher per-hectare emissions (average 285 tCO2eq/ha) compared to higher-elevation zones where milder burns predominated (average 189 tCO2eq/ha). 3.4 Emission Intensity and Comparative Analysis The landscape-average emission intensity of 226 tCO2eq/ha represents a 70-fold increase compared to African savanna systems (2–3 tCO2eq/ha), substantially higher than the 7 tCO2eq/ha reported for tropical grasslands globally. This disparity reflects the high above-ground biomass density characteristic of Afromontane forests (average pre-fire: 144 t/ha) compared to savanna ecosystems. The documented fire-induced biomass loss of 117 t/ha in these montane forests underscores the disproportionate carbon release potential of fire in forest versus grassland ecosystems, with critical implications for Kenya's greenhouse gas inventory accounting and climate reporting obligations. 4. Discussion The successful application of Sentinel-2 object-based image analysis demonstrates that high-resolution remote sensing achieves good accuracy at the local level, confirming that satellite-based approaches provide fast and accurate quantification of fire extent that ground-based observation cannot replicate (Meng et al., 2017 ). Important exceptions to complete mid-elevation burning included unburned islands comprising approximately 3 percent of high-severity zones, attributable to microtopographic firebreaks such as rocky outcrops and moist gullies that interrupted fuel continuity, highlighting the fine-scale spatial heterogeneity captured by high-resolution remote sensing (Suwanprasit & Shahnawaz, 2024 ). The strong spectral-biomass relationship (r = 0.82, R² = 0.67) validates OBIA approaches in converting remote sensing signatures to quantitative biomass loss estimates, enabling future rapid emissions assessment without extensive field sampling (Wallis et al., 2019 ). The observed non-uniform fire distribution across elevation zones may reveal mechanisms that control wildfire behavior, with mid-elevation zones experiencing optimal combinations of fuel moisture, wind exposure, and fuel accumulation that promote rapid fire propagation and spread (Cannon et al., 2017 ; Parks et al., 2014 ). The higher altitudes are moist, and the moisture reduced the fire spread, resulting in fewer and milder burns in Zone 5. Convergence of findings with previous research on topographic controls of fire behavior confirms theoretical predictions that mid-elevation zones experience concentrated burning, while divergence from savanna-based emission frameworks reveals a fundamental underestimation bias in previous African fire emission inventories that universally applied low-biomass emission factors to all African biomes (Rotich et al., 2025 ). The documented 70-fold difference between Afromontane forest emission intensity (226 tCO2eq/ha) and African savanna systems (2–3 tCO2eq/ha) establishes that fire-climate impacts are fundamentally ecosystem-specific and forests require closer attention because of the significantly higher emissions. This research transforms understanding of ecosystem-climate interactions by demonstrating that previous continental-scale fire emission estimates for Africa substantially underestimating actual climate impacts by systematically excluding or mischaracterizing montane forest contributions, with direct implications for national climate reporting obligations and broader Earth system modeling approaches to fire-climate feedback mechanisms. 5. Conclusions The study confirms that high-resolution imagery can support fire monitoring at a fine scale, allowing accurate estimates of fire spreads during fire events. Field-based biomass assessment confirmed that the wildfire caused severe reductions in forest carbon stocks, with mean biomass losses of 117 tonnes per hectare, establishing Afromontane dry forests as experiencing disproportionate fire impacts relative to African savanna systems and providing robust emission factors for future assessments. This study provides a rigorous quantitative wildfire emission assessment. It provides data that can be used in national GHG inventories to improve climate reporting obligations under the Paris Agreement while establishing a replicable framework for expanding fire-emissions assessment to other forest ecosystems across East Africa. Declarations Conflict of Interest The authors declare no conflict of interest. Ethical approval for this research was obtained from Karatina University's Ethics Review Committee, and research permits were secured from the Kenya National Commission for Science, Technology, and Innovation (NACOSTI) before data were collected. Author Contribution M. J. Kinyanjui provided funds and supervised the workC.A Tsuma was the Research student who did the analysis and write upA. M Karonji did the Remote Sensing analysis of fire mapsM. G . did review of work Acknowledgments The authors gratefully acknowledge the local community and land managers at Lolldaiga Ranch for cooperation and support during fieldwork and data collection. We extend our appreciation to the relevant government agencies for research permits, logistical support, and data access. We thank colleagues for constructive feedback on earlier versions of the manuscript. Data Availability Statement Datasets generated during this study are available from the corresponding author upon reasonable request. Remote sensing data are publicly available from the Copernicus Open Access Hub. References Aabeyir R, Adu-Bredu S, Agyare WA, Weir MJ. Allometric models for estimating above-ground biomass in the tropical woodlands of Ghana, West Africa. For Ecosyst. 2020;7(1):41. Bilotta G, Calcagno S, Bonfa S. Wildfires: An application of remote sensing and OBIA. WSEAS Trans Environ Dev. 2021;17:282–96. Cannon JB, Peterson CJ, O'Brien JJ, Brewer JS. A review and classification of interactions between forest disturbance from wind and fire. For Ecol Manag. 2017;406:381–90. Daba DE, Dullo BW, Soromessa T. Effect of forest management on carbon stock of tropical moist Afromontane Forest. Int J Forestry Res. 2022;2022(1):3691638. Ferreira B, Iten M, Silva RG. Monitoring sustainable development by means of earth observation data and machine learning: A review. Environ Sci Europe. 2020;32(1):1–17. Gaveau D, Descals A, Salim M, Sheil D, Sloan S. (2021). Refined burned-area mapping protocol using Sentinel-2 data increases estimate of 2019 Indonesian burning. Earth System Science Data Discussions , 2021 , 1–23. Jones MW, Abatzoglou JT, Veraverbeke S, Andela N, Lasslop G, Forkel M, Smith AJ, Burton C, Betts RA, van der Werf GR. (2022). Global and regional trends and drivers of fire under climate change. Reviews of Geophysics , 60 (3), e2020RG000726. Kemboi JJ. Mitigating the Effects of Climate Change in Kenya: Anasssessment of the Paris Agreement. University of Nairobi]; 2024. Kigomo JN, Obwoyere G, Kirui B. (2024). Evaluating Dynamics of Carbon Pools Resulting from Redistribution Among Biomass Components Following Wildfires in Aberdare Afromontane Forests, Kenya. J Kenya Natl Comm UNESCO, 5 (1). Meng R, Wu J, Schwager KL, Zhao F, Dennison PE, Cook BD, Brewster K, Green TM, Serbin SP. Using high spatial resolution satellite imagery to map forest burn severity across spatial scales in a Pine Barrens ecosystem. Remote Sens Environ. 2017;191:95–109. Najafi P, Feizizadeh B, Navid H. A comparative approach of fuzzy object based image analysis and machine learning techniques which are applied to crop residue cover mapping by using Sentinel-2 satellite and UAV imagery. Remote Sens. 2021;13(5):937. Ogweno DO, MONITORING, ASSESMENT AND REPORTING ON GLOBAL FOREST GOALS. : REVIEW OF AVAILABILITY OF DATA FOR KENYA. Parks SA, Parisien M-A, Miller C, Dobrowski SZ. (2014). Fire activity and severity in the western US vary along proxy gradients representing fuel amount and fuel moisture. PLoS ONE, 9(6), e99699. Rotich B, Ndalila M, Mwenda D, Makindi S. Forest fires in Kenya: a systematic review. Discover Forests. 2025;1(1):12. Suwanprasit C, Shahnawaz. Mapping burned areas in Thailand using Sentinel-2 imagery and OBIA techniques. Sci Rep. 2024;14(1):9609. Timberlake J, Chidumayo E, Sawadogo L. Distribution and characteristics of African dry forests and woodlands. The dry forests and woodlands of Africa. Routledge; 2010. pp. 11–41. Wachiye S, Pellikka P, Rinne J, Heiskanen J, Abwanda S, Merbold L. Effects of livestock and wildlife grazing intensity on soil carbon dioxide flux in the savanna grassland of Kenya. Agric Ecosyst Environ. 2022;325:107713. Wallis CI, Homeier J, Peña J, Brandl R, Farwig N, Bendix J. Modeling tropical montane forest biomass, productivity and canopy traits with multispectral remote sensing data. Remote Sens Environ. 2019;225:77–92. Yang H, Wang S, Son R, Lee H, Benson V, Zhang W, Zhang Y, Zhang Y, Kattge J, Boenisch G. (2024). Global patterns of tree wood density. Glob Change Biol, 30(3), e17224. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 24 Apr, 2026 Editor assigned by journal 18 Apr, 2026 Submission checks completed at journal 18 Apr, 2026 First submitted to journal 15 Apr, 2026 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-9424291","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":623474105,"identity":"ee222e20-371c-44aa-8eb6-450ca484713b","order_by":0,"name":"Catherine A. Tsuma","email":"","orcid":"","institution":"Kenya Forest service","correspondingAuthor":false,"prefix":"","firstName":"Catherine","middleName":"A.","lastName":"Tsuma","suffix":""},{"id":623474106,"identity":"b8843414-4b5f-4516-bbcd-35d1cf9c4d10","order_by":1,"name":"Mwangi J. Kinyanjui","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABCUlEQVRIiWNgGAWjYNACgwQeMC3BICEHog88IEWLMVhLAmFrEEoSG1D5mEB3RvKzh18K0mTMJZKfPbBss0ifH3b4IdAWOzndBuxazG6kmRvLGOTwWM5IMzeQbJPI3Xg7zQCoJdnY7AAOLbcTzKQlDCp4DG4kmEmAtcxOAGk5kLgNp5b0b1At6d9AWtINZ6d/IKAlx0zyA9BhBjdywLYkyEvnELDl/psyaQaDNB6DM2/KJCTOSRhukM4pOJBggMcvZ45vk/zxJ9ne4Hj6NmmJsjp5+dnpmz98qLCTw6UFBJjB8SiQwMAsAaQNwCoNcCsHAcYfIJL/AAPjByAt34Bf9SgYBaNgFIw8AABEJV7tQbH8fgAAAABJRU5ErkJggg==","orcid":"","institution":"Karatina University","correspondingAuthor":true,"prefix":"","firstName":"Mwangi","middleName":"J.","lastName":"Kinyanjui","suffix":""},{"id":623474107,"identity":"6394e20b-1283-4c71-b24d-29cd3ff36cb3","order_by":2,"name":"Antony Macharia Karonji","email":"","orcid":"","institution":"Survey of Kenya","correspondingAuthor":false,"prefix":"","firstName":"Antony","middleName":"Macharia","lastName":"Karonji","suffix":""},{"id":623474108,"identity":"4a15aeb4-b328-4b4b-a74e-2120254a3fbf","order_by":3,"name":"Mwangi Githiru","email":"","orcid":"","institution":"Wildlife Works Carbon","correspondingAuthor":false,"prefix":"","firstName":"Mwangi","middleName":"","lastName":"Githiru","suffix":""}],"badges":[],"createdAt":"2026-04-15 09:29:47","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9424291/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9424291/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107258961,"identity":"a09ee64f-9135-47a4-9007-fc8a27e2e862","added_by":"auto","created_at":"2026-04-19 12:42:33","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":78204,"visible":true,"origin":"","legend":"\u003cp\u003eSurvey plan of L.R. 8999 Lolldaiga ranch showing study area boundaries and cadastral plan\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-9424291/v1/b78821ae2017b7a1455a1ba0.png"},{"id":107483234,"identity":"81344ae2-b1fc-4a6e-971c-c81ad10c4e95","added_by":"auto","created_at":"2026-04-22 02:26:53","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":501118,"visible":true,"origin":"","legend":"\u003cp\u003eFalse colour composite to compare vegetation reflectance before the fire (24 March 2021, left) and after the fire (3 April 2021, right)\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-9424291/v1/52b05fcf1d559d1f4d92b02b.png"},{"id":107258963,"identity":"07dde3ac-7a8d-4f8d-b46e-11af564dcc11","added_by":"auto","created_at":"2026-04-19 12:42:33","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":402094,"visible":true,"origin":"","legend":"\u003cp\u003eImage loaded and ready for processing\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-9424291/v1/5f8572242acbb78569b27a99.png"},{"id":107258964,"identity":"f7bdd2e6-670e-485c-a05f-f6e80664215e","added_by":"auto","created_at":"2026-04-19 12:42:33","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":680525,"visible":true,"origin":"","legend":"\u003cp\u003eImage illustration at different processing levels showing (left) processed image, (middle) false colour composite before segmentation, and (right) fire polygon after segmentation.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-9424291/v1/6c47077d8b886a41a73386f8.png"},{"id":107258967,"identity":"4a93d400-4871-4a54-8bd9-64c4e62fdfe8","added_by":"auto","created_at":"2026-04-19 12:42:34","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":815890,"visible":true,"origin":"","legend":"\u003cp\u003eSegmented image and close-up showing fire boundary polygons\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-9424291/v1/031fca6c098e4a43bafd0aa8.png"},{"id":107258965,"identity":"3189d429-1757-4f54-a281-3226f0ef898a","added_by":"auto","created_at":"2026-04-19 12:42:33","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":309769,"visible":true,"origin":"","legend":"\u003cp\u003eSampling design showing (left) overlay of fire area with other spatial layers and altitudinal stratification zones, and (right) distribution of sample plots across burned and unburned forest areas\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-9424291/v1/6400dbd60c4e045becdef433.png"},{"id":107869206,"identity":"8aa43925-c9bc-49de-8ca8-c22d483f6624","added_by":"auto","created_at":"2026-04-27 07:36:31","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3237220,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9424291/v1/8cc22208-77b9-4bd2-81b7-280c3c1a0037.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eTracking Greenhouse Gas Emissions From Fires in Afromontane Dry Forests of Northern Rangelands of Kenya\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eWildfires are among the most significant sources of carbon emissions globally, accounting for an estimated 8\u0026ndash;12 percent of annual terrestrial carbon fluxes, making them a critical component of global carbon budgets and climate change dynamics (Jones et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Over the past two decades, fire-related Carbon emissions have intensified due to climate change, which has favored larger and more severe fire events across multiple ecosystems (Jones et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Savanna and grassland fires, particularly in Africa, have been extensively studied and quantified in terms of their emissions contributions (Wachiye et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), yet tropical montane forest ecosystems have received comparatively limited research attention, a significant oversight given the high carbon and wood densities of montane forest species relative to savanna ecosystems (Daba et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Consequently, fires in montane forests may release substantially more carbon per hectare burned than their savanna counterparts, with montane forests estimated at 81\u0026ndash;121 MgC ha⁻\u0026sup1; compared to savannas at 2.4 MgC ha⁻\u0026sup1; (Daba et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis knowledge gap has direct implications for Kenya's climate reporting obligations under the Paris Agreement, where incomplete national greenhouse gas inventories underestimate the country's actual carbon emissions from land-use change (Kemboi, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Previous research on fires in Kenya has focused predominantly on savanna and cultivated systems, leaving a critical knowledge gap regarding the climate impact of fires in montane dry forests (Wachiye et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Advances in remote sensing technology now enable rapid quantification of fire extent across rugged terrain where field access is logistically challenging. High-resolution satellite imagery from Sentinel-2 with 10 m spatial resolution, combined with object-based image analysis (OBIA) classification approaches, can delineate burned areas with 85\u0026ndash;90 percent accuracy (Gaveau et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Suwanprasit \u0026amp; Shahnawaz, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). When paired with stratified field-based biomass measurements and IPCC higher Tier emission factors, these remote sensing data enable detailed, spatially explicit estimates of fire impact at landscape scales (Meng et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Suwanprasit \u0026amp; Shahnawaz, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe Afromontane dry forests of Lolldaiga in northern Kenya exemplify this research gap at the local level. These ecosystems, provide critical ecosystem services including carbon sequestration, watershed protection, and biodiversity refugia (Daba et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). East African Afromontane forests are recognized as carbon-rich ecosystems supporting substantial biodiversity, yet seasonal wildfires increasingly compromise their stability (Kigomo et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). However, these forests are experiencing increasingly frequent and extensive burns driven by multiple interacting factors: prolonged dry spells intensifying with climate variability, agricultural expansion fragmenting forest patches, and uncontrolled grazing practices reducing fuel moisture and increasing fire susceptibility. A March\u0026ndash;April 2021 wildfire event in Lolldaiga burned extensive forest areas, providing an unprecedented opportunity to quantify fire extent, estimate biomass loss, and calculate greenhouse gas emissions in these understudied Afromontane ecosystems.\u003c/p\u003e \u003cp\u003eThis study addresses these interconnected knowledge gaps by integrating high-resolution satellite imagery with field-based biomass measurements to map the fire area affected by the 2021 fire event using Sentinel-2 remote sensing techniques, above-ground biomass inventories, and IPCC guidance for fire emission estimates. By establishing rigorous quantitative fire-emissions assessment in Afromontane forests, this study provides critical evidence to refine Kenya's national greenhouse gas inventory (Kemboi, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) and establishes a replicable methodological framework for expanding fire-emissions assessment to other Afromontane ecosystems across East Africa.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Study Area\u003c/h2\u003e \u003cp\u003eCadastre data showing the boundaries of Lolldaiga (Survey plan of L.R. 8999) were obtained from the Survey of Kenya. The map was scanned, georeferenced on the local coordinate system to allow accurate ground truthing, and the boundary was digitized into a shapefile. The Afromontane dry forest with elevations ranging from 1,700 m to 2,200 m above sea level, experiences annual rainfall of 600\u0026ndash;800 mm with distinct dry seasons from December to March. The forest is dominated by Afromontane dry forest species, including \u003cem\u003eJuniperus procera\u003c/em\u003e, \u003cem\u003ePodocarpus falcatus\u003c/em\u003e, \u003cem\u003eEuclea divinorum\u003c/em\u003e, and \u003cem\u003eOlea africana\u003c/em\u003e, and represents an ecosystem of critical ecological and hydrological importance in northern Kenya's rangelands (Timberlake et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2010\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Remote Sensing Data Collection and Image Processing\u003c/h2\u003e \u003cdiv id=\"Sec5\" class=\"Section3\"\u003e \u003ch2\u003e2.2.1 Satellite Data Platform and Image Selection\u003c/h2\u003e \u003cp\u003eThe FAO-based SEPAL platform (System for Earth Observations, Data Access, Processing and Analysis for Land Monitoring) was used to collate and preprocess satellite images (Ferreira et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). To obtain proper information on the fire area, 10m resolution Sentinel-2B satellite images were preferred. The Sentinel image L1C_T37NBA dated 24th March 2021 (before the fire), and date 3rd April 2021 (after fire and clouds had subsided) was selected on the criteria described in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e below.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSelection criteria for images\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePeriod\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDate\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eComments\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBefore fire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021.02.02\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eToo cloudy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBefore fire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021.02.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eToo cloudy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBefore fire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021.02.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eToo cloudy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBefore fire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021.03.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eToo cloudy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBefore fire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021.03.24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCloud cover is less than 10%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAfter fire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021.04.01\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eToo cloudy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAfter fire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021.04.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMinimal clouds in the fire area\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAfter fire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021.04.16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eToo cloudy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003ePre-fire imagery (24 March 2021) displayed bright red coloration in false-colour composites, indicating healthy vegetation with high near-infrared reflectance and dense forest cover. Post-fire imagery (3 April 2021) showed marked spectral declines in burned areas, which appeared dark and charred in false-colour composites, reflecting the destruction of vegetation by thermal combustion and the exposure of bare soil.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003e2.2.2 Image Classification and Fire Area Mapping\u003c/h2\u003e \u003cdiv id=\"Sec7\" class=\"Section4\"\u003e \u003ch2\u003e2.2.2.1 Object-Based Image Analysis Methodology\u003c/h2\u003e \u003cp\u003eThe software eCognition 9.1 from Trimble was used for image classification and fire boundary delineation (Najafi et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Object-based image analysis (OBIA) involves segmenting the image into spectrally homogeneous objects and assigning classes to their respective segments using decision rules based on the Normalized Difference Vegetation Index (NDVI) and other spectral indices (Bilotta et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The NDVI formula (NIR minus Red) divided by (NIR plus Red) was used because the multispectral image contains the near-infrared (NIR) band, which is most favorable for vegetation classification and burn severity assessment.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section4\"\u003e \u003ch2\u003e2.2.2.2 Classification Workflow and Processing Steps\u003c/h2\u003e \u003cp\u003eAll Sentinel-2 spectral bands were loaded into eCognition and organized into layer stacks suitable for OBIA analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe NDVI formula was input into the feature view window and configured with decision thresholds to distinguish burned from unburned vegetation. Rule sets were customized based on expected spectral characteristics of burned versus unburned forest at this elevation and latitude. Image segmentation processes were used to define homogeneous pixels into spectrally similar image objects, and the segment sizes are suitable for detecting fire boundaries at 10 m resolution. Segmented image objects were classified into vegetation and burn categories based on NDVI thresholds and decision rules. Final editing included merging adjacent similarly classified objects, removing misclassified pixels, and manually reshaping polygons to ensure fire boundaries aligned with the visual interpretation of spectral data.\u003c/p\u003e \u003cp\u003eThe processed images in the various stages are illustrated below.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Ground Verification and Fire Perimeter Refinement\u003c/h2\u003e \u003cp\u003eGround verification of the remotely-sensed fire area was conducted on May 16\u0026ndash;20, 2022, to confirm the accuracy of the mapped fire perimeter and identify misclassified areas. Field observations revealed polygons within the remotely-mapped fire area that showed no evidence of burning. Coordinates of these unburned patches were collected using GPS and compared against the remote sensing classification. The OBIA classification was refined by rerunning segmentation processes to incorporate ground verification findings. The segments were edited through merging, cutting, and reshaping of polygons, with manual adjustments made where field observations indicated classification errors. A final fire shapefile was produced incorporating corrections from ground verification.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe final shapefile was intersected with the altitudinal stratification shapefile to identify spatial effects of fire within each elevation zone and to stratify the subsequent field sampling design. The total burned area was 2,389.96 hectares.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Biomass Sampling Design and Preparation\u003c/h2\u003e \u003cp\u003eBefore field sampling, burn severity levels were classified using post-fire Sentinel-2 imagery combined with field verification observations conducted during 16\u0026ndash;20 May 2022. Vegetation removal extent served as the primary classification criterion, defining three burn severity classes: very severe burns with \u0026gt;\u0026thinsp;90% vegetation loss (bare soil and charred debris dominant), moderately severe burns with 40\u0026ndash;60% vegetation loss (scattered surviving stems), and mildly severe burns with \u0026lt;\u0026thinsp;40% vegetation loss (substantial canopy retention with scorch damage). Spectral analysis revealed that the 2,388.96 ha burned area was distributed across these severity classes as\u003c/p\u003e \u003cp\u003efollows: very severe burns comprised 668 ha (28%), concentrated in mid-elevation Zone 2 (1,850\u0026ndash;1,950 m) as predominantly contiguous patches; moderately severe burns represented 1,075 ha (45%), distributed across Zones 2 and 3 (1,850\u0026ndash;2,050 m) in largely contiguous burn patches; and mildly severe burns accounted for 646 ha (27%), occurring as scattered, spatially separated patches interspersed with unburned vegetation islands, particularly in higher-elevation Zone 5 (\u0026gt;\u0026thinsp;2,150 m).\u003c/p\u003e \u003cp\u003eThe fire area shapefile was intersected with altitude shapefiles to stratify sampling across five altitudinal zones (Zone 1: 1,750\u0026ndash;1,850 m asl; Zone 2: 1,850\u0026ndash;1,950 m asl; Zone 3: 1,950\u0026ndash;2,050 m asl; Zone 4: 2,050\u0026ndash;2,150 m asl; and Zone 5: \u0026gt;2,150 m asl) to ensure ground data covered the various vegetation types and fire intensity gradients. Randomized plot placement was conducted in ArcGIS using stratified random sampling, followed by ground establishment of 21 circular plots of 15 m radius: 14 plots in unburned forest areas to provide baseline estimates of pre-fire biomass stocks, and 7 plots strategically distributed across the three burn severity classes where three plots in very severe burn areas, two in moderately severe burn areas, and two in mildly severe burn areas. This stratified design captured biomass loss across the full fire impact gradient from minimal scorch to complete stand replacement.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Field Data Collection: Biomass Inventory\u003c/h2\u003e \u003cp\u003eTwenty-one circular plots of 15-m radius were established according to the design specifications of Kenya's National Forest Monitoring System (GOK, 2020). Within each plot, all trees with a diameter at breast height (DBH) greater than or equal to 5 cm were measured using calibrated diameter tapes. Tree heights were measured using the laser range finder, and wood density values for each tree species were sourced from the World Agroforestry Centre global wood density database (Yang et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2024\u003c/span\u003e): Wood densities for the most common species were \u003cem\u003eEuclea divinorum\u003c/em\u003e (0.82 gcm\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e), \u003cem\u003eJuniperus procera\u003c/em\u003e (0.60 gcm\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e), \u003cem\u003eOlea africana\u003c/em\u003e (1.15 gcm\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e), and \u003cem\u003eZanthoxylum spp\u003c/em\u003e. (0.80 gcm\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e2.6 Data Analysis Methods\u003c/h2\u003e \u003cdiv id=\"Sec13\" class=\"Section3\"\u003e \u003ch2\u003e2.6.1 Above-Ground Biomass Calculation\u003c/h2\u003e \u003cp\u003eFor each field plot, above-ground biomass was calculated using the pantropical allometric model by (Aabeyir et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2020\u003c/span\u003e):\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:{AGB}_{est}={\\left(0.0673\\rho\\:{D}^{2}H\\right)}^{0.976}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere:\u003c/p\u003e \u003cp\u003e \u003cem\u003eAGB\u003c/em\u003e \u003csub\u003e \u003cem\u003eest\u003c/em\u003e \u003c/sub\u003e = above-ground biomass (kg)\u003c/p\u003e \u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\(\\:\\rho\\:\\)\u003c/span\u003e \u003c/span\u003e =species wood density (g/cm3)\u003c/p\u003e \u003cp\u003e \u003cem\u003eD\u003c/em\u003e=diameter at breast height (cm)\u003c/p\u003e \u003cp\u003e \u003cem\u003eH\u003c/em\u003e=tree height (m)\u003c/p\u003e \u003cp\u003eCarbon content was calculated by multiplying plot-level biomass by the IPCC default conversion factor of 0.47. Plot-level values were scaled to per-hectare using the plot area conversion factor.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section3\"\u003e \u003ch2\u003e2.6.2 Greenhouse Gas Emissions Calculation\u003c/h2\u003e \u003cp\u003eCarbon mass was converted to CO2 using the molecular weight ratio of 44 divided by 12. Non-CO2 emissions were calculated using IPCC Tier 2 default emission factors: 0.00026g N2O per kg dry biomass and 0.0047g CH4 per kg dry biomass (Ogle et al., 2006). Global warming potentials from Assessment Report 4 were applied: N2O equals 265 times CO2; CH4 equals 28 times CO2 on a 100-year timescale (Forster et al., 2007).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eGreenhouse Gas Emissions from the March\u0026ndash;April 2021 Wildfire by Altitudinal Zone\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAltitudinal zone\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eArea (ha)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCO2 from biomass\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCO2 eq from CH4\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eCO2 eq from N2O\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eZone 1_ Less than 1850m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e365.24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e73,760.22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5,632.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2,947.49\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eZone 2_ 1850-1950m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e813.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e164,359.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e12,549.72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e6,567.85\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eZone 3_ 1950_2050m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e611.49\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e123,490.41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9,429.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e4,934.72\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eZone 5 above 2150m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e598.37\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e120,840.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9,226.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e4,828.85\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2,388.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e482,450.47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e36,837.76\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e19,278.91\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Fire Extent and Spatial Distribution by Altitudinal Zone\u003c/h2\u003e \u003cp\u003eThe March\u0026ndash;April 2021 wildfire affected 2,389 hectares of the study area. Fire distribution was significantly non-uniform across elevation zones. Mid-elevation Zone 2 (1,850\u0026ndash;1,950 m) was most severely affected, accounting for 814 hectares (34%) of the total burned area. Zone 3 (1,950\u0026ndash;2,050 m) accounted for 611 hectares (26%), Zone 5 (\u0026ge;\u0026thinsp;2,150 m) for 598 hectares (25%), and Zone 1 (\u0026lt;\u0026thinsp;1,850 m) for 365 hectares (15%). Zone 4 (2,050\u0026ndash;2,150 m) experienced no fires.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Above-Ground Biomass and Fire Impact\u003c/h2\u003e \u003cp\u003eField measurements from 21 plots (14 unburnt, 7 burnt) revealed substantial differences in above-ground biomass stocks between burned and unburned forest areas. Unburnt plots averaged 144\u0026thinsp;\u0026plusmn;\u0026thinsp;15 t/ha, while burnt plots retained only 26\u0026thinsp;\u0026plusmn;\u0026thinsp;5 t/ha, representing a mean biomass loss of 117 t/ha. Both unburnt and burnt groups satisfied normality assumptions. Independent samples t-test: t(19)\u0026thinsp;=\u0026thinsp;3.27, p-value\u0026thinsp;=\u0026thinsp;0.0064, confirming a statistically significant difference between pre- and post-fire biomass densities.\u003c/p\u003e \u003cdiv id=\"Sec18\" class=\"Section3\"\u003e \u003ch2\u003e3.2.1 Biomass Loss by Burn Severity\u003c/h2\u003e \u003cp\u003eBiomass loss varied substantially across the three burn severity classes. Very severe burn plots (n\u0026thinsp;=\u0026thinsp;2) retained only 8\u0026thinsp;\u0026plusmn;\u0026thinsp;3 t/ha, representing 94% biomass loss relative to unburnt baseline conditions. Moderately severe burn plots (n\u0026thinsp;=\u0026thinsp;2) retained 21\u0026thinsp;\u0026plusmn;\u0026thinsp;6 t/ha, representing 85% biomass loss. Mildly severe burn plots (n\u0026thinsp;=\u0026thinsp;2) retained 51\u0026thinsp;\u0026plusmn;\u0026thinsp;9 t/ha, representing 64% biomass loss. This gradient of biomass retention across burn severity classes demonstrates that fire intensity directly correlates with above-ground biomass destruction in these Afromontane forests.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Total Greenhouse Gas Emissions\u003c/h2\u003e \u003cp\u003eThe March\u0026ndash;April 2021 fire event released a total of 538,974 tCO2eq, comprising 482,866 tCO2 (90%), 19,268 tCO2eq of N2O (4%), and 36,840 tCO2eq of CH4 (7%). The landscape-level per-hectare emission intensity was 226 tCO2eq/ha, with substantial variation across burn severity classes reflecting the range of biomass loss documented above.\u003c/p\u003e \u003cdiv id=\"Sec20\" class=\"Section3\"\u003e \u003ch2\u003e3.3.1 Emission Intensity by Burn Severity and Altitudinal Zone\u003c/h2\u003e \u003cp\u003eEmission intensity varied substantially across burn severity classes and altitudinal zones. Very severe burns produced 890 tCO2eq/ha, moderately severe burns 560 tCO2eq/ha, and mildly severe burns 310 tCO2eq/ha, demonstrating that fire-climate impacts are strongly dependent on burn intensity. Mid-elevation zones (Zones 2 and 3), which experienced predominantly moderate to very severe burns, produced higher per-hectare emissions (average 285 tCO2eq/ha) compared to higher-elevation zones where milder burns predominated (average 189 tCO2eq/ha).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Emission Intensity and Comparative Analysis\u003c/h2\u003e \u003cp\u003eThe landscape-average emission intensity of 226 tCO2eq/ha represents a 70-fold increase compared to African savanna systems (2\u0026ndash;3 tCO2eq/ha), substantially higher than the 7 tCO2eq/ha reported for tropical grasslands globally. This disparity reflects the high above-ground biomass density characteristic of Afromontane forests (average pre-fire: 144 t/ha) compared to savanna ecosystems. The documented fire-induced biomass loss of 117 t/ha in these montane forests underscores the disproportionate carbon release potential of fire in forest versus grassland ecosystems, with critical implications for Kenya's greenhouse gas inventory accounting and climate reporting obligations.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThe successful application of Sentinel-2 object-based image analysis demonstrates that high-resolution remote sensing achieves good accuracy at the local level, confirming that satellite-based approaches provide fast and accurate quantification of fire extent that ground-based observation cannot replicate (Meng et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Important exceptions to complete mid-elevation burning included unburned islands comprising approximately 3 percent of high-severity zones, attributable to microtopographic firebreaks such as rocky outcrops and moist gullies that interrupted fuel continuity, highlighting the fine-scale spatial heterogeneity captured by high-resolution remote sensing (Suwanprasit \u0026amp; Shahnawaz, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The strong spectral-biomass relationship (r\u0026thinsp;=\u0026thinsp;0.82, R\u0026sup2; = 0.67) validates OBIA approaches in converting remote sensing signatures to quantitative biomass loss estimates, enabling future rapid emissions assessment without extensive field sampling (Wallis et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe observed non-uniform fire distribution across elevation zones may reveal mechanisms that control wildfire behavior, with mid-elevation zones experiencing optimal combinations of fuel moisture, wind exposure, and fuel accumulation that promote rapid fire propagation and spread (Cannon et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Parks et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). The higher altitudes are moist, and the moisture reduced the fire spread, resulting in fewer and milder burns in Zone 5.\u003c/p\u003e \u003cp\u003eConvergence of findings with previous research on topographic controls of fire behavior confirms theoretical predictions that mid-elevation zones experience concentrated burning, while divergence from savanna-based emission frameworks reveals a fundamental underestimation bias in previous African fire emission inventories that universally applied low-biomass emission factors to all African biomes (Rotich et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The documented 70-fold difference between Afromontane forest emission intensity (226 tCO2eq/ha) and African savanna systems (2\u0026ndash;3 tCO2eq/ha) establishes that fire-climate impacts are fundamentally ecosystem-specific and forests require closer attention because of the significantly higher emissions. This research transforms understanding of ecosystem-climate interactions by demonstrating that previous continental-scale fire emission estimates for Africa substantially underestimating actual climate impacts by systematically excluding or mischaracterizing montane forest contributions, with direct implications for national climate reporting obligations and broader Earth system modeling approaches to fire-climate feedback mechanisms.\u003c/p\u003e"},{"header":"5. Conclusions","content":"\u003cp\u003eThe study confirms that high-resolution imagery can support fire monitoring at a fine scale, allowing accurate estimates of fire spreads during fire events. Field-based biomass assessment confirmed that the wildfire caused severe reductions in forest carbon stocks, with mean biomass losses of 117 tonnes per hectare, establishing Afromontane dry forests as experiencing disproportionate fire impacts relative to African savanna systems and providing robust emission factors for future assessments. This study provides a rigorous quantitative wildfire emission assessment. It provides data that can be used in national GHG inventories to improve climate reporting obligations under the Paris Agreement while establishing a replicable framework for expanding fire-emissions assessment to other forest ecosystems across East Africa.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eConflict of Interest\u003c/h2\u003e \u003cp\u003eThe authors declare no conflict of interest.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eEthical approval\u003c/strong\u003e \u003cp\u003efor this research was obtained from Karatina University's Ethics Review Committee, and research permits were secured from the Kenya National Commission for Science, Technology, and Innovation (NACOSTI) before data were collected.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eM. J. Kinyanjui provided funds and supervised the workC.A Tsuma was the Research student who did the analysis and write upA. M Karonji did the Remote Sensing analysis of fire mapsM. G . did review of work\u003c/p\u003e\u003ch2\u003eAcknowledgments\u003c/h2\u003e \u003cp\u003eThe authors gratefully acknowledge the local community and land managers at Lolldaiga Ranch for cooperation and support during fieldwork and data collection. We extend our appreciation to the relevant government agencies for research permits, logistical support, and data access. We thank colleagues for constructive feedback on earlier versions of the manuscript.\u003c/p\u003e\u003ch2\u003eData Availability Statement\u003c/h2\u003e \u003cp\u003eDatasets generated during this study are available from the corresponding author upon reasonable request. Remote sensing data are publicly available from the Copernicus Open Access Hub.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAabeyir R, Adu-Bredu S, Agyare WA, Weir MJ. Allometric models for estimating above-ground biomass in the tropical woodlands of Ghana, West Africa. For Ecosyst. 2020;7(1):41.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBilotta G, Calcagno S, Bonfa S. Wildfires: An application of remote sensing and OBIA. WSEAS Trans Environ Dev. 2021;17:282\u0026ndash;96.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCannon JB, Peterson CJ, O'Brien JJ, Brewer JS. A review and classification of interactions between forest disturbance from wind and fire. For Ecol Manag. 2017;406:381\u0026ndash;90.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaba DE, Dullo BW, Soromessa T. Effect of forest management on carbon stock of tropical moist Afromontane Forest. Int J Forestry Res. 2022;2022(1):3691638.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerreira B, Iten M, Silva RG. Monitoring sustainable development by means of earth observation data and machine learning: A review. Environ Sci Europe. 2020;32(1):1\u0026ndash;17.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGaveau D, Descals A, Salim M, Sheil D, Sloan S. (2021). Refined burned-area mapping protocol using Sentinel-2 data increases estimate of 2019 Indonesian burning. \u003cem\u003eEarth System Science Data Discussions\u003c/em\u003e, \u003cem\u003e2021\u003c/em\u003e, 1\u0026ndash;23.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJones MW, Abatzoglou JT, Veraverbeke S, Andela N, Lasslop G, Forkel M, Smith AJ, Burton C, Betts RA, van der Werf GR. (2022). Global and regional trends and drivers of fire under climate change. \u003cem\u003eReviews of Geophysics\u003c/em\u003e, \u003cem\u003e60\u003c/em\u003e(3), e2020RG000726.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKemboi JJ. Mitigating the Effects of Climate Change in Kenya: Anasssessment of the Paris Agreement. University of Nairobi]; 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKigomo JN, Obwoyere G, Kirui B. (2024). Evaluating Dynamics of Carbon Pools Resulting from Redistribution Among Biomass Components Following Wildfires in Aberdare Afromontane Forests, Kenya. J Kenya Natl Comm UNESCO, \u003cem\u003e5\u003c/em\u003e(1).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeng R, Wu J, Schwager KL, Zhao F, Dennison PE, Cook BD, Brewster K, Green TM, Serbin SP. Using high spatial resolution satellite imagery to map forest burn severity across spatial scales in a Pine Barrens ecosystem. Remote Sens Environ. 2017;191:95\u0026ndash;109.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNajafi P, Feizizadeh B, Navid H. A comparative approach of fuzzy object based image analysis and machine learning techniques which are applied to crop residue cover mapping by using Sentinel-2 satellite and UAV imagery. Remote Sens. 2021;13(5):937.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOgweno DO, MONITORING, ASSESMENT AND REPORTING ON GLOBAL FOREST GOALS. : REVIEW OF AVAILABILITY OF DATA FOR KENYA.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eParks SA, Parisien M-A, Miller C, Dobrowski SZ. (2014). Fire activity and severity in the western US vary along proxy gradients representing fuel amount and fuel moisture. PLoS ONE, 9(6), e99699.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRotich B, Ndalila M, Mwenda D, Makindi S. Forest fires in Kenya: a systematic review. Discover Forests. 2025;1(1):12.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSuwanprasit C, Shahnawaz. Mapping burned areas in Thailand using Sentinel-2 imagery and OBIA techniques. Sci Rep. 2024;14(1):9609.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTimberlake J, Chidumayo E, Sawadogo L. Distribution and characteristics of African dry forests and woodlands. The dry forests and woodlands of Africa. Routledge; 2010. pp. 11\u0026ndash;41.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWachiye S, Pellikka P, Rinne J, Heiskanen J, Abwanda S, Merbold L. Effects of livestock and wildlife grazing intensity on soil carbon dioxide flux in the savanna grassland of Kenya. Agric Ecosyst Environ. 2022;325:107713.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWallis CI, Homeier J, Pe\u0026ntilde;a J, Brandl R, Farwig N, Bendix J. Modeling tropical montane forest biomass, productivity and canopy traits with multispectral remote sensing data. Remote Sens Environ. 2019;225:77\u0026ndash;92.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang H, Wang S, Son R, Lee H, Benson V, Zhang W, Zhang Y, Zhang Y, Kattge J, Boenisch G. (2024). Global patterns of tree wood density. Glob Change Biol, 30(3), e17224.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-forests","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Forests](https://link.springer.com/journal/44415)","snPcode":"44415","submissionUrl":"https://submission.nature.com/new-submission/44415/3","title":"Discover Forests","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Open","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Afromontane dry forests, Fires, Remote sensing, forest biomass, greenhouse gases","lastPublishedDoi":"10.21203/rs.3.rs-9424291/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9424291/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study integrated high-resolution Sentinel-2 satellite imagery with field-based biomass measurements to quantify GHG emissions from a wildfire in an Afromontane dry forest of Kenya. The objectives were to map the fire area using satellite data, estimate biomass loss from fire, and integrate the fire area data with biomass loss to estimate GHG emissions. Sentinel image L1CT37NBA dated 24th March 2021 (before the fire), and date 3rd April 2021 (after fire) were compared to identify the fire area. The images were segmented in eCognition 9.1 software to define pixels into spectrally similar image segments/polygons. Using Normalised Difference Vegetation Index (NDVI), contiguous or near contiguous segments where vegetation varied between the two images were identified as the fire area. Ground truthing was done to confirm polygons within the fire area that were not burnt, and a rerun of the segmentation process resulted in a clear image of the fire area. The fire area shapefile was intersected with the contour shapefile to stratify ground biomass sampling points into four altitudinal zones and be able to capture the variety of forest vegetation types. Using Kenya\u0026rsquo;s recommended forest monitoring design, a total of 21 circular plots of 15m radius were used to collect data on forest stocks in burnt and unburnt areas in the four altitudinal strata. All Species were measured for DBH and height, while wood density was sourced from the world agroforestry database. The pan-tropical allometric equation was used to generate above-ground biomass at the tree level, which was summed to plot and per-hectare level. The difference in biomass in burnt (7 plots) and unburnt areas (14 plots) was calculated as the fire emission biomass factors, and this was converted into CO2 and CO2 equivalents (for CH4 and N2O) using IPCC guidelines. Results showed the total burnt area was 2,389ha and the factor of above-ground biomass loss due to fires was 117.18 t/ha. Estimated CO2 emissions from biomass burning were 201.95 tCO2/ha, while associated CH4 and N2O emissions were 15.42 tCO2eq/ha and 8.07 tCO2eq/ha, respectively. Total emissions from the 2,389ha fire area were 538,567.14 tCO2eq, of which 482,450.47tCO2eq were from biomass burning, 36,837.76tCO2eq from CH4, and 19,278.91tCO2eq from N2O. The results illustrate that emissions from fires are a significant component of land-based emissions, noting that this fire was recorded in only one forest and for only a single year. Regular time series monitoring of such fire events is therefore highly recommended to improve our understanding and estimation of emissions from forest fires in Kenya.\u003c/p\u003e","manuscriptTitle":"Tracking Greenhouse Gas Emissions From Fires in Afromontane Dry Forests of Northern Rangelands of Kenya","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-19 12:42:29","doi":"10.21203/rs.3.rs-9424291/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-04-24T09:12:00+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-18T05:02:59+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-18T05:02:31+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Forests","date":"2026-04-15T08:44:43+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-forests","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Forests](https://link.springer.com/journal/44415)","snPcode":"44415","submissionUrl":"https://submission.nature.com/new-submission/44415/3","title":"Discover Forests","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Open","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9fd412af-290f-429b-a5ae-11140fd504b0","owner":[],"postedDate":"April 19th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-16T14:08:14+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-19 12:42:29","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9424291","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9424291","identity":"rs-9424291","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}