Pub Date : 2026-01-01DOI: 10.1016/j.rsase.2025.101851
Vincent Haller , Luke Sanford, Timothy Gregoire
Market-based conservation schemes, such as REDD+ (Reducing Emissions from Deforestation and Forest Degradation), rely on the accurate quantification of carbon stocks. However, forest carbon modeling is still uncertain due to the challenging collection of field samples for calibrating remote sensing data. In this research, above-ground biomass (AGB) was measured in 120 inventory plots located on Chiloe Island in southern Chile. A forest carbon model was then calibrated using machine learning algorithms, with spectral, radar, and LiDAR-based canopy height data as predictors. Random Forest models showed the lowest errors, with root-mean-square error (RMSE) ranging from 91.0 to 93.7 tons AGB/ha. The presented methodology allows AGB mapping at the local scale, with no saturation below 350 tons AGB/ha. Significant prediction differences between our AGB maps and global AGB datasets confirm the importance of ground truthing for biomass mapping applications at the local scale. These cost-effective, localized methods could reduce the uncertainties of biomass mapping and further promote REDD+.
基于市场的保护计划，如REDD+（减少毁林和森林退化造成的排放），依赖于碳储量的准确量化。然而，森林碳模型仍然是不确定的，因为收集野外样品校准遥感数据具有挑战性。本研究在智利南部Chiloe岛的120个调查样地测量了地上生物量（AGB）。然后使用机器学习算法校准森林碳模型，以光谱、雷达和基于激光雷达的冠层高度数据作为预测因子。随机森林模型误差最小，均方根误差（RMSE）在91.0 ~ 93.7 t AGB/ha之间。所提出的方法允许在局部尺度上进行AGB测绘，饱和度不低于350吨AGB/ha。我们的AGB地图与全球AGB数据集之间的显著预测差异证实了地面真实性对局部尺度生物质制图应用的重要性。这些具有成本效益的本地化方法可以减少生物量制图的不确定性，进一步促进REDD+。
{"title":"The importance of local calibration in biomass models for REDD+ – a case study in the Chiloe island, Chile","authors":"Vincent Haller , Luke Sanford, Timothy Gregoire","doi":"10.1016/j.rsase.2025.101851","DOIUrl":"10.1016/j.rsase.2025.101851","url":null,"abstract":"<div><div>Market-based conservation schemes, such as REDD+ (Reducing Emissions from Deforestation and Forest Degradation), rely on the accurate quantification of carbon stocks. However, forest carbon modeling is still uncertain due to the challenging collection of field samples for calibrating remote sensing data. In this research, above-ground biomass (AGB) was measured in 120 inventory plots located on Chiloe Island in southern Chile. A forest carbon model was then calibrated using machine learning algorithms, with spectral, radar, and LiDAR-based canopy height data as predictors. Random Forest models showed the lowest errors, with root-mean-square error (RMSE) ranging from 91.0 to 93.7 tons AGB/ha. The presented methodology allows AGB mapping at the local scale, with no saturation below 350 tons AGB/ha. Significant prediction differences between our AGB maps and global AGB datasets confirm the importance of ground truthing for biomass mapping applications at the local scale. These cost-effective, localized methods could reduce the uncertainties of biomass mapping and further promote REDD+.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101851"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146022720","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-01DOI: 10.1016/j.rsase.2026.101870
Abdelhalim Miftah
The Nitrogen dioxide (NO2) is one of the main pollutant gases in the atmosphere with known harmful effects on human health, caused by rapid urbanization, traffic, and industrialization. In the fast-growing urban areas of North Africa, NO2 spatiotemporal variability analysis is hampered by the sparse distribution of ground stations. In neighboring Morocco, particularly in the Casablanca-Settat region, research that utilizes satellite data analysis in combination with geospatial statistical analysis is also limited. The purpose of this study is to fill the scientific gap in spatial variability analysis of NO2 concentrations from 2018 to 2025 in a specific region in Casablanca-Settat in Morocco. The NO2 concentration in the tropospheric region using the Sentinel-5P (TROPOMI) satellite sensor was analyzed using geospatial analysis tools. The hotspots and cold spots of NO2 were determined using the Getis-Ord Gi∗ statistic at a statistically significant confidence level. The annual average NO2 concentration varied between 6.3 μg/m3 in rural areas in southern provinces to 11.3 μg/m3 in Casablanca, showing sharp differences between the urban, industrial areas, and rural areas. The cities of Casablanca, Mediouna, and the Nouaceur province showed higher NO2 concentrations than other cities in any country in the world; they often exceeded the annual average recommended by the World Health Organization at 10 μg/m3. However, southern provinces, such as Settat, Sidi Bennour, and El Jadida, had smaller NO2 concentrations that were less variable. A marked decrease in NO2 was observed during the COVID-19 lockdown period, followed by a stabilized period in 2022–2023 and a decline in NO2 thereafter. The NO2 hotspots were in the Casablanca-Beth Mohammedia-Mediouna-Berrechid region at a 99 % confidence level, which was a hotspot region, but in southern areas, it was a cold spot region.
{"title":"Spatiotemporal dynamics of NO2 concentration data (2018–2025) in Casablanca-Settat region, Morocco: A satellite-based assessment for urban air quality management","authors":"Abdelhalim Miftah","doi":"10.1016/j.rsase.2026.101870","DOIUrl":"10.1016/j.rsase.2026.101870","url":null,"abstract":"<div><div>The Nitrogen dioxide (NO<sub>2</sub>) is one of the main pollutant gases in the atmosphere with known harmful effects on human health, caused by rapid urbanization, traffic, and industrialization. In the fast-growing urban areas of North Africa, NO<sub>2</sub> spatiotemporal variability analysis is hampered by the sparse distribution of ground stations. In neighboring Morocco, particularly in the Casablanca-Settat region, research that utilizes satellite data analysis in combination with geospatial statistical analysis is also limited. The purpose of this study is to fill the scientific gap in spatial variability analysis of NO<sub>2</sub> concentrations from 2018 to 2025 in a specific region in Casablanca-Settat in Morocco. The NO<sub>2</sub> concentration in the tropospheric region using the Sentinel-5P (TROPOMI) satellite sensor was analyzed using geospatial analysis tools. The hotspots and cold spots of NO<sub>2</sub> were determined using the Getis-Ord Gi∗ statistic at a statistically significant confidence level. The annual average NO<sub>2</sub> concentration varied between 6.3 μg/m<sup>3</sup> in rural areas in southern provinces to 11.3 μg/m<sup>3</sup> in Casablanca, showing sharp differences between the urban, industrial areas, and rural areas. The cities of Casablanca, Mediouna, and the Nouaceur province showed higher NO<sub>2</sub> concentrations than other cities in any country in the world; they often exceeded the annual average recommended by the World Health Organization at 10 μg/m<sup>3</sup>. However, southern provinces, such as Settat, Sidi Bennour, and El Jadida, had smaller NO<sub>2</sub> concentrations that were less variable. A marked decrease in NO<sub>2</sub> was observed during the COVID-19 lockdown period, followed by a stabilized period in 2022–2023 and a decline in NO<sub>2</sub> thereafter. The NO<sub>2</sub> hotspots were in the Casablanca-Beth Mohammedia-Mediouna-Berrechid region at a 99 % confidence level, which was a hotspot region, but in southern areas, it was a cold spot region.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101870"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145977874","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Invasive alien plant species (IAPS) are increasingly threatening biodiversity conservation during rapid climate change, especially in vulnerable regions, such as Ethiopia. This study aimed to model current and future habitat suitability for four major IAPS (Prosopis juliflora, Parthenium hysterophorus, Lantana camara, and Acacia spp.) and assess multi-species invasion risks across Ethiopia's 136 protected areas (PAs), using Random Forest (RF)–based species distribution modeling (SDM). The analysis integrated 27 environmental predictors with downscaled climate projections from the Coupled Model Intercomparison Project Phase 6 (CMIP6), under Shared Socioeconomic Pathway (SSP) 2-4.5 (moderate) and SSP5-8.5 (high-emission) scenarios. Findings revealed distinct ecological niches among the four major species, with P. hysterophorus currently exhibiting the largest area of suitable habitat (175,743 km2). All species demonstrated significant expansion potential under climate change, with P. juliflora showing the most dramatic relative increase (up to 106.25 % by 2070 under SSP5-8.5). Niche overlap analysis indicated high environmental similarity between L. camara and P. hysterophorus (Hellinger's I = 0.8826), suggesting potential for co-occurrence in invasion hotspots concentrated in the southwestern highlands and Rift Valley. Most critically, the PA vulnerability assessment revealed that smaller protected areas (<100 km2) face significantly higher potential invasion pressure (mean suitable area: 41.5 %) than larger ones (≥5000 km2: 25.3 %), with P. hysterophorus posing the broadest spatial threat affecting 37 PAs with >50 % habitat suitability. Future projections indicate 206–247 % increases in invasion threats across PAs by mid-century, with eastern and southeastern conservation landscapes facing a disproportionate rise in risk. These findings present the first national, evidence-based spatial prioritization framework for climate-adaptive IAPS management in Ethiopia, demonstrating that conservation planning should integrate dynamic invasion risk assessments to safeguard biodiversity amid global climate change.
{"title":"Random forest-based species distribution modeling reveals intensifying multi-species invasion risks of alien plants in Ethiopia under climate change","authors":"Kalid Hassen Yasin , Diriba Tulu , Tadele Bedo Gelete , Beyan Ahmed Yuya , Anteneh Derribew Iguala , Kiya Adare Tadesse , Erana Kebede","doi":"10.1016/j.rsase.2026.101869","DOIUrl":"10.1016/j.rsase.2026.101869","url":null,"abstract":"<div><div>Invasive alien plant species (IAPS) are increasingly threatening biodiversity conservation during rapid climate change, especially in vulnerable regions, such as Ethiopia. This study aimed to model current and future habitat suitability for four major IAPS (<em>Prosopis juliflora</em>, <em>Parthenium hysterophorus</em>, <em>Lantana camara</em>, and <em>Acacia</em> spp.) and assess multi-species invasion risks across Ethiopia's 136 protected areas (PAs), using Random Forest (RF)–based species distribution modeling (SDM). The analysis integrated 27 environmental predictors with downscaled climate projections from the Coupled Model Intercomparison Project Phase 6 (CMIP6), under Shared Socioeconomic Pathway (SSP) 2-4.5 (moderate) and SSP5-8.5 (high-emission) scenarios. Findings revealed distinct ecological niches among the four major species, with <em>P. hysterophorus</em> currently exhibiting the largest area of suitable habitat (175,743 km<sup>2</sup>). All species demonstrated significant expansion potential under climate change, with <em>P. juliflora</em> showing the most dramatic relative increase (up to 106.25 % by 2070 under SSP5-8.5). Niche overlap analysis indicated high environmental similarity between <em>L. camara</em> and <em>P. hysterophorus</em> (Hellinger's <em>I</em> = 0.8826), suggesting potential for co-occurrence in invasion hotspots concentrated in the southwestern highlands and Rift Valley. Most critically, the PA vulnerability assessment revealed that smaller protected areas (<100 km<sup>2</sup>) face significantly higher potential invasion pressure (mean suitable area: 41.5 %) than larger ones (≥5000 km<sup>2</sup>: 25.3 %), with <em>P. hysterophorus</em> posing the broadest spatial threat affecting 37 PAs with >50 % habitat suitability. Future projections indicate 206–247 % increases in invasion threats across PAs by mid-century, with eastern and southeastern conservation landscapes facing a disproportionate rise in risk. These findings present the first national, evidence-based spatial prioritization framework for climate-adaptive IAPS management in Ethiopia, demonstrating that conservation planning should integrate dynamic invasion risk assessments to safeguard biodiversity amid global climate change.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101869"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926032","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-01DOI: 10.1016/j.rsase.2026.101881
Laura L. Bourgeau-Chavez , Chelene Hanes , Michael Billmire , Karl Bosse , Michael J. Battaglia , Andreas Colliander
In boreal and Arctic regions, where organic soils act as wildfire fuel, NASA’s SMAP soil moisture products offer strong potential to improve drought and fuel (organic soil) moisture assessment beyond point-based weather station fire danger models. Since SMAP is not calibrated for organic soils, we evaluated its suitability using a network of fuel moisture stations we established across three SMAP grid cells, as well as fire weather station data across the North American boreal and Arctic. Comparison of SMAP products, brightness temperature, and reflectivity with in situ fuel moisture measurements revealed SMAP products to be dry-biased with low dynamic range (r = −0.03 to 0.40). In contrast, SMAP reflectivity showed good relationships to in situ fuel moisture at 6 cm depth in the Alaska tundra site (r = 0.62), and 10–18 cm depth for the Alberta (r = 0.46) and Ontario (r = 0.62) boreal sites. SMAP soil moisture products were then used to develop a statistical model to predict Drought Code (DC), a weather-based index of fuel availability in the deeper (10–20 cm) organic soil layers. The model, created using hundreds of weather stations across boreal and Arctic regions, explained 63 % of overall deviance (range 28–86 %). Additionally, incorporating SMAP retrievals flagged for dense vegetation increased spatial coverage without compromising model performance. These results indicate that an operational SMAP-derived deep organic fuel moisture (e.g. DC) product is feasible if future retrievals account for soil organic content. This would enhance fire danger monitoring and decision support across boreal and Arctic regions.
{"title":"Assessing SMAP for enhanced wildfire danger prediction in boreal-Arctic ecosystems","authors":"Laura L. Bourgeau-Chavez , Chelene Hanes , Michael Billmire , Karl Bosse , Michael J. Battaglia , Andreas Colliander","doi":"10.1016/j.rsase.2026.101881","DOIUrl":"10.1016/j.rsase.2026.101881","url":null,"abstract":"<div><div>In boreal and Arctic regions, where organic soils act as wildfire fuel, NASA’s SMAP soil moisture products offer strong potential to improve drought and fuel (organic soil) moisture assessment beyond point-based weather station fire danger models. Since SMAP is not calibrated for organic soils, we evaluated its suitability using a network of fuel moisture stations we established across three SMAP grid cells, as well as fire weather station data across the North American boreal and Arctic. Comparison of SMAP products, brightness temperature, and reflectivity with <em>in situ</em> fuel moisture measurements revealed SMAP products to be dry-biased with low dynamic range (r = −0.03 to 0.40). In contrast, SMAP reflectivity showed good relationships to <em>in situ</em> fuel moisture at 6 cm depth in the Alaska tundra site (r = 0.62), and 10–18 cm depth for the Alberta (r = 0.46) and Ontario (r = 0.62) boreal sites. SMAP soil moisture products were then used to develop a statistical model to predict Drought Code (DC), a weather-based index of fuel availability in the deeper (10–20 cm) organic soil layers. The model, created using hundreds of weather stations across boreal and Arctic regions, explained 63 % of overall deviance (range 28–86 %). Additionally, incorporating SMAP retrievals flagged for dense vegetation increased spatial coverage without compromising model performance. These results indicate that an operational SMAP-derived deep organic fuel moisture (e.g. DC) product is feasible if future retrievals account for soil organic content. This would enhance fire danger monitoring and decision support across boreal and Arctic regions.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101881"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077860","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Winter wheat constitutes a key component of Armenia's agricultural economy and national food security. Accurate and timely identification and mapping of winter wheat fields can support informed policymaking, infrastructure planning, resource allocation, and crop monitoring. However, large-scale crop mapping using traditional field-based surveys is labor-intensive, spatially limited, and often impractical. In this study, we present a framework for large-scale winter wheat field mapping in Armenia under limited ground truth availability using multi-source satellite images. We evaluate the possibility of cross-region generalization through image-based semantic segmentation models (3D Unet and SegNet) and pointwise classification models (Random Forest, one-dimensional convolutional neural network, and long short-term memory (LSTM)) on multi-temporal data using Sentinel-2 and PlanetScope images. We train these models using large, well-labeled winter wheat datasets from United States counties and a limited set of known winter wheat fields in Armenia, and subsequently transfer and test across five independent test regions in Armenia. Models trained on Sentinel-2 imagery generalize well within the United States, achieving test accuracy and F1-scores of 0.96 and 0.86, respectively. However, their performance degrades when transferred to fragmented agricultural landscapes in Armenia, particularly for image-based semantic segmentation models. In contrast, the LSTM-based temporal model demonstrates superior transferability, achieving accuracy and F1-scores of 0.96 and 0.96, respectively, while effectively suppressing non-wheat features and identifying both known and previously unmapped winter wheat fields. Using the best-performing model, we generate provincial-scale winter wheat maps for Shirak Province for the years 2023, 2024, and 2025, estimating cultivated areas of 25,641 ha, 21,088 ha (approximately 3 % higher than the USDA Foreign Agricultural Service estimate), and 28,909 ha, respectively. These results highlight the potential of temporal machine learning models for scalable winter wheat mapping in data-scarce regions, offering a practical pathway toward nationwide crop inventory generation and agricultural decision support.
{"title":"Transferability of spatial and temporal learning models for winter wheat mapping in data-scarce environments: A case study in Armenia","authors":"Ashutosh Tiwari , Lei Zhao , Sayantan Sarkar , Vardan Urutyan , Uday Santhosh Raju Vysyaraju , Sejeong Moon , Benjamin Ghansah , Garnik Sevoyan , Juan Landivar , Yuri Calil , Mahendra Bhandari","doi":"10.1016/j.rsase.2026.101885","DOIUrl":"10.1016/j.rsase.2026.101885","url":null,"abstract":"<div><div>Winter wheat constitutes a key component of Armenia's agricultural economy and national food security. Accurate and timely identification and mapping of winter wheat fields can support informed policymaking, infrastructure planning, resource allocation, and crop monitoring. However, large-scale crop mapping using traditional field-based surveys is labor-intensive, spatially limited, and often impractical. In this study, we present a framework for large-scale winter wheat field mapping in Armenia under limited ground truth availability using multi-source satellite images. We evaluate the possibility of cross-region generalization through image-based semantic segmentation models (3D Unet and SegNet) and pointwise classification models (Random Forest, one-dimensional convolutional neural network, and long short-term memory (LSTM)) on multi-temporal data using Sentinel-2 and PlanetScope images. We train these models using large, well-labeled winter wheat datasets from United States counties and a limited set of known winter wheat fields in Armenia, and subsequently transfer and test across five independent test regions in Armenia. Models trained on Sentinel-2 imagery generalize well within the United States, achieving test accuracy and F1-scores of 0.96 and 0.86, respectively. However, their performance degrades when transferred to fragmented agricultural landscapes in Armenia, particularly for image-based semantic segmentation models. In contrast, the LSTM-based temporal model demonstrates superior transferability, achieving accuracy and F1-scores of 0.96 and 0.96, respectively, while effectively suppressing non-wheat features and identifying both known and previously unmapped winter wheat fields. Using the best-performing model, we generate provincial-scale winter wheat maps for Shirak Province for the years 2023, 2024, and 2025, estimating cultivated areas of 25,641 ha, 21,088 ha (approximately 3 % higher than the USDA Foreign Agricultural Service estimate), and 28,909 ha, respectively. These results highlight the potential of temporal machine learning models for scalable winter wheat mapping in data-scarce regions, offering a practical pathway toward nationwide crop inventory generation and agricultural decision support.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101885"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146022715","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-01DOI: 10.1016/j.rsase.2025.101854
Stephen Biliyitorb Liwur , Abdul Rashid Adam , Jacob Nchagmado Tagnan
Bush-burning plays a key role in shaping ecosystems and supporting livelihoods, yet its intensification under human pressures has transformed the practice from a regenerative agent into a major environmental hazard. In Ghana's savannah zones, bushfire is used for land clearing, pasture management, and pest control, but also drives deforestation, soil degradation, and air pollution. However, extant studies mainly focused on its ecological and agricultural impacts, with limited attention to atmospheric and public health implications. This study addresses this gap by examining the spatiotemporal dynamics of bushfires, vegetation change, and air quality. We employ multiscale geographic weighted regression (MGWR), land use and land cover classification (LULC), normalized burn ratio (NBR), and normalized difference vegetation index (NDVI) to examine the spatiotemporal dynamics of bushfires, vegetation health, and air quality across Ghana's savannah zones (2019/2020 and 2024/2025). The findings point out that burn severity affected more than 60 % of the landscape, peaking between December and February. Correspondingly, NDVI declined with over 70 % showing vegetation stress. Similarly, air pollution intensified: PM2.5 from 28 to 35 μg/m3, CO from 0.4 to 0.6 μg/m3, and SO2 from 3 to 5 ppbv. MGWR revealed weakening NDVI–NBR associations (median β = 0.62; Adj. R2 = 0.51) and spatial clustering in pollutant emissions (Residual Moran's I ≈ −0.03-0.04; p < 0.01). These findings emphasize that recurrent bush burning is intensifying vegetation degradation and air pollution--a trajectory of deteriorating ecological health with significant public health implications. Therefore, there is a critical need for spatial heterogeneity-vulnerability mapping for early-warning systems, land-use planning, and public health interventions.
{"title":"After the flames and smoke, then what? Spatial analysis and heterogeneity modeling of bushfire effects on vegetation health and air quality in Ghana","authors":"Stephen Biliyitorb Liwur , Abdul Rashid Adam , Jacob Nchagmado Tagnan","doi":"10.1016/j.rsase.2025.101854","DOIUrl":"10.1016/j.rsase.2025.101854","url":null,"abstract":"<div><div>Bush-burning plays a key role in shaping ecosystems and supporting livelihoods, yet its intensification under human pressures has transformed the practice from a regenerative agent into a major environmental hazard. In Ghana's savannah zones, bushfire is used for land clearing, pasture management, and pest control, but also drives deforestation, soil degradation, and air pollution. However, extant studies mainly focused on its ecological and agricultural impacts, with limited attention to atmospheric and public health implications. This study addresses this gap by examining the spatiotemporal dynamics of bushfires, vegetation change, and air quality. We employ multiscale geographic weighted regression (MGWR), land use and land cover classification (LULC), normalized burn ratio (NBR), and normalized difference vegetation index (NDVI) to examine the spatiotemporal dynamics of bushfires, vegetation health, and air quality across Ghana's savannah zones (2019/2020 and 2024/2025). The findings point out that burn severity affected more than 60 % of the landscape, peaking between December and February. Correspondingly, NDVI declined with over 70 % showing vegetation stress. Similarly, air pollution intensified: PM<sub>2.5</sub> from 28 to 35 μg/m<sup>3</sup>, CO from 0.4 to 0.6 μg/m<sup>3</sup>, and SO<sub>2</sub> from 3 to 5 ppbv. MGWR revealed weakening NDVI–NBR associations (median β = 0.62; Adj. R<sup>2</sup> = 0.51) and spatial clustering in pollutant emissions (Residual Moran's I ≈ −0.03-0.04; p < 0.01). These findings emphasize that recurrent bush burning is intensifying vegetation degradation and air pollution--a trajectory of deteriorating ecological health with significant public health implications. Therefore, there is a critical need for spatial heterogeneity-vulnerability mapping for early-warning systems, land-use planning, and public health interventions.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101854"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145884535","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-01DOI: 10.1016/j.rsase.2025.101858
Sze Ping Hui , Yu Liu , Pir Mohammad , Qihao Weng
Urban expansion and the resulting three-dimensional building patterns have significant ecological impacts, particularly the urban heat island (UHI) effect. However, the influence of urban morphology on the thermal environment of a dense city is still not clear. This study investigates the diurnal and seasonal relationships between two-dimensional (2D) and three-dimensional (3D) urban morphology indicators with the remotely sensed land surface temperature (LST) in Hong Kong, a city characterized by extreme urban density and significant vertical development. Using a boosted regression tree (BRT) model, we analyze the spatial variation of LST during both hot and cold months, in daytime and nighttime, and evaluate the marginal impacts of selected 2D/3D indicators on LST. The results reveal pronounced spatial heterogeneity in LST, with high daytime temperatures concentrated in industrial areas and nighttime peaks shifting to commercial and residential zones. The relationships between urban form indicators and LST are complex and nonlinear, differing significantly between day and night. Among all indicators, the sky view factor (SVF) is identified as the most influential, with its marginal effect on LST shifting from positive to negative after a threshold of 0.75 in January daytime and 0.76 in January nighttime. Whereas in July, the SVF threshold decreases to 0.71 in daytime and 0.62 in nighttime. The marginal effect of building height shows a similar trend with 30 m threshold in January in both daytime and nighttime, while 21.57 m in daytime and 20.31 m in nighttime in July, proving a significant cooling benefit in daytime. This study provides urban planners with evidence-based guidelines for optimizing building design and layout to mitigate UHI, potentially through manipulating factors like SVF and building height. These insights can inform sustainable urban development strategies in highly dense cities worldwide.
{"title":"Assessing the diurnal relationship between urban morphology and land surface temperature in a highly dense city: A case study in Hong Kong","authors":"Sze Ping Hui , Yu Liu , Pir Mohammad , Qihao Weng","doi":"10.1016/j.rsase.2025.101858","DOIUrl":"10.1016/j.rsase.2025.101858","url":null,"abstract":"<div><div>Urban expansion and the resulting three-dimensional building patterns have significant ecological impacts, particularly the urban heat island (UHI) effect. However, the influence of urban morphology on the thermal environment of a dense city is still not clear. This study investigates the diurnal and seasonal relationships between two-dimensional (2D) and three-dimensional (3D) urban morphology indicators with the remotely sensed land surface temperature (LST) in Hong Kong, a city characterized by extreme urban density and significant vertical development. Using a boosted regression tree (BRT) model, we analyze the spatial variation of LST during both hot and cold months, in daytime and nighttime, and evaluate the marginal impacts of selected 2D/3D indicators on LST. The results reveal pronounced spatial heterogeneity in LST, with high daytime temperatures concentrated in industrial areas and nighttime peaks shifting to commercial and residential zones. The relationships between urban form indicators and LST are complex and nonlinear, differing significantly between day and night. Among all indicators, the sky view factor (SVF) is identified as the most influential, with its marginal effect on LST shifting from positive to negative after a threshold of 0.75 in January daytime and 0.76 in January nighttime. Whereas in July, the SVF threshold decreases to 0.71 in daytime and 0.62 in nighttime. The marginal effect of building height shows a similar trend with 30 m threshold in January in both daytime and nighttime, while 21.57 m in daytime and 20.31 m in nighttime in July, proving a significant cooling benefit in daytime. This study provides urban planners with evidence-based guidelines for optimizing building design and layout to mitigate UHI, potentially through manipulating factors like SVF and building height. These insights can inform sustainable urban development strategies in highly dense cities worldwide.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101858"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926178","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-01DOI: 10.1016/j.rsase.2026.101883
Hui Gao , Xianjun Du
Cloud detection is a key step in remote sensing image preprocessing, but existing methods face challenges with fragmented and thin clouds. Fragmented clouds have dispersed, multi-scale features, while thin clouds are optically similar to non-cloud regions and sparsely distributed, making them hard to distinguish with local features alone. Therefore, jointly modeling global and local features is crucial. To address these issues while maintaining linear complexity, this paper proposes a Wavelet-Enhanced Multi-Scale Mamba-like Linear Attention Decoder (WMSMLAD) method. WMSMLAD consists of three components: the WMSMLA block, the Weighted Fusion (WF) module, and the Cloud Head (CH). The WMSMLA block cascades a Feature Decomposition (FD) module to extract spatial feature distributions and uses a Multi-Layer Perceptron (MLP) to handle channel interactions. The FD module suppresses irrelevant features, while the WMSMLA enhances global information through Haar wavelet transform and refines feature boundaries. It combines multi-scale convolution with a Mamba-like linear attention (MLLA) mechanism to capture multi-scale local features and global dependencies. The WF module dynamically adjusts the weight between encoding and decoding features, with the cloud mask output through the CH. Experimental results demonstrate that WMSMLAD achieves competitive performance among the compared methods, with an mIoU of 91.53 % on the MODIS dataset and a MAE of 0.0599 on the CHLandsat dataset.
{"title":"Wavelet-enhanced Mamba-like multi-scale linear attention decoding for remote sensing cloud detection","authors":"Hui Gao , Xianjun Du","doi":"10.1016/j.rsase.2026.101883","DOIUrl":"10.1016/j.rsase.2026.101883","url":null,"abstract":"<div><div>Cloud detection is a key step in remote sensing image preprocessing, but existing methods face challenges with fragmented and thin clouds. Fragmented clouds have dispersed, multi-scale features, while thin clouds are optically similar to non-cloud regions and sparsely distributed, making them hard to distinguish with local features alone. Therefore, jointly modeling global and local features is crucial. To address these issues while maintaining linear complexity, this paper proposes a Wavelet-Enhanced Multi-Scale Mamba-like Linear Attention Decoder (WMSMLAD) method. WMSMLAD consists of three components: the WMSMLA block, the Weighted Fusion (WF) module, and the Cloud Head (CH). The WMSMLA block cascades a Feature Decomposition (FD) module to extract spatial feature distributions and uses a Multi-Layer Perceptron (MLP) to handle channel interactions. The FD module suppresses irrelevant features, while the WMSMLA enhances global information through Haar wavelet transform and refines feature boundaries. It combines multi-scale convolution with a Mamba-like linear attention (MLLA) mechanism to capture multi-scale local features and global dependencies. The WF module dynamically adjusts the weight between encoding and decoding features, with the cloud mask output through the CH. Experimental results demonstrate that WMSMLAD achieves competitive performance among the compared methods, with an mIoU of 91.53 % on the MODIS dataset and a MAE of 0.0599 on the CHLandsat dataset.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101883"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145977875","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-01DOI: 10.1016/j.rsase.2026.101892
Caique Machado e Silva , Victor Silva Signorini , Henrique Caletti Mezzomo , João Paulo Oliveira Ribeiro , Gabriel Wolter Lima , Marcelo Fagundes Portes , Lucas de Paula Corrêdo , Tiago Olivoto , Gota Morota , Maicon Nardino
Unoccupied aerial vehicles (UAV) coupled with spectral sensors allow the collection of a wide range of spectral information suitable for indirect selection of primary target traits. However, most studies using such strategies analyze spectral variables derived from longitudinal high-throughput phenotyping based on single-time-point approaches, neglecting the temporal trajectory of such variables. In addition, the cause-and-effect relationship between the target and secondary variables is often overlooked. To guide appropriate decisions regarding indirect selection, this study explored a linear mixed model-based repeatability framework to analyze longitudinal UAV-based spectral variables, investigate the associations between target agronomic and spectral variables, and evaluate the efficiency of indirect selection strategies in a diverse tropical wheat panel. A total of 49 genotypes were evaluated in a 7 × 7 lattice design with two replications for grain yield and days to heading. Five spectral bands and seven vegetation indices were also extracted from images of a field trial collected in ten flights. A linear mixed model was used to estimate variance components and predict genotypic values. Genotypic correlations between variables were decomposed into direct and indirect effects. The spectral variable with the largest direct effect on the target agronomic variables was used in an indirect selection strategy. The genotypic variance was significant for all traits, indicating the presence of suitable genetic variability that can be exploited in the selection process. Broad sense heritability estimates were low to moderate for most traits, highlighting the challenges of phenotypic selection. Inferred genetic and non-genetic parameters allowed us to estimate the optimal number of flights to obtain reliable spectral information. The red wavelength was shown to have the largest direct effect on the target agronomic variables and to be efficient in indirect selection, especially for days to heading. This study will serve as a guide for the implementation of indirect selection pipelines in tropical wheat breeding programs to optimize selection decisions.
{"title":"Optimizing indirect selection of tropical wheat genotypes using high-throughput longitudinal phenotyping and trait relationships","authors":"Caique Machado e Silva , Victor Silva Signorini , Henrique Caletti Mezzomo , João Paulo Oliveira Ribeiro , Gabriel Wolter Lima , Marcelo Fagundes Portes , Lucas de Paula Corrêdo , Tiago Olivoto , Gota Morota , Maicon Nardino","doi":"10.1016/j.rsase.2026.101892","DOIUrl":"10.1016/j.rsase.2026.101892","url":null,"abstract":"<div><div>Unoccupied aerial vehicles (UAV) coupled with spectral sensors allow the collection of a wide range of spectral information suitable for indirect selection of primary target traits. However, most studies using such strategies analyze spectral variables derived from longitudinal high-throughput phenotyping based on single-time-point approaches, neglecting the temporal trajectory of such variables. In addition, the cause-and-effect relationship between the target and secondary variables is often overlooked. To guide appropriate decisions regarding indirect selection, this study explored a linear mixed model-based repeatability framework to analyze longitudinal UAV-based spectral variables, investigate the associations between target agronomic and spectral variables, and evaluate the efficiency of indirect selection strategies in a diverse tropical wheat panel. A total of 49 genotypes were evaluated in a 7 × 7 lattice design with two replications for grain yield and days to heading. Five spectral bands and seven vegetation indices were also extracted from images of a field trial collected in ten flights. A linear mixed model was used to estimate variance components and predict genotypic values. Genotypic correlations between variables were decomposed into direct and indirect effects. The spectral variable with the largest direct effect on the target agronomic variables was used in an indirect selection strategy. The genotypic variance was significant for all traits, indicating the presence of suitable genetic variability that can be exploited in the selection process. Broad sense heritability estimates were low to moderate for most traits, highlighting the challenges of phenotypic selection. Inferred genetic and non-genetic parameters allowed us to estimate the optimal number of flights to obtain reliable spectral information. The red wavelength was shown to have the largest direct effect on the target agronomic variables and to be efficient in indirect selection, especially for days to heading. This study will serve as a guide for the implementation of indirect selection pipelines in tropical wheat breeding programs to optimize selection decisions.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101892"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077730","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-01DOI: 10.1016/j.rsase.2026.101891
Rytis Maskeliūnas, Sarmad Maqsood
Accurate and efficient semantic segmentation of point cloud data is essential for a wide range of remote sensing applications, from urban mapping to environmental monitoring. However, challenges such as data sparsity, class imbalance, and high computational complexity often lead to poor performance of existing segmentation methods. In this study, we propose a novel hybrid segmentation framework that integrates Point Transformer v3 (PTv3) and Squeeze-Excitation (SE) attention mechanisms to enhance feature extraction and improve segmentation accuracy. The preprocessing pipeline incorporates voxel grid downsampling to reduce redundancy, fixed-size point cloud preparation (1024 points per sample) for computational consistency, and Synthetic Minority Over-Sampling Technique (SMOTE) to address class imbalance by generating synthetic data for underrepresented classes. The proposed PTv3-SE model is built on hierarchical attention mechanisms and channel-wise recalibration to efficiently capture both local and global features in sparse and noisy point clouds. The approach incorporates compression, based on Octrees and Truncated Pyramid subdivision, as well as internal ”data” transmission optimization, with feature prioritization and adaptive streaming strategy. The model was evaluated using the SemanticKITTI and ShapeNet datasets, demonstrating robust state-of-the-art level segmentation performance, with good performance of 93.4% accuracy, 95.03% precision, 93.44% recall, and an F1 score of 93.98%, while also maintaining high computational efficiency of 2s per frame, almost a second faster than the closest competitor model. Compared to existing methods, our approach measurably improves segmentation accuracy and efficiency. On the SemanticKITTI dataset, the framework achieves a mIoU of 87.5%, which surpasses PointNet++ (74.1%) and DGCNN (72.5%) by a significant margin. Similarly, on the ShapeNet dataset, the framework shows good object-level segmentation with an mIoU of 89.4%, reflecting its robustness in capturing fine-grained details.
{"title":"Hybrid attention-based PTv3-SE model for efficient point cloud segmentation","authors":"Rytis Maskeliūnas, Sarmad Maqsood","doi":"10.1016/j.rsase.2026.101891","DOIUrl":"10.1016/j.rsase.2026.101891","url":null,"abstract":"<div><div>Accurate and efficient semantic segmentation of point cloud data is essential for a wide range of remote sensing applications, from urban mapping to environmental monitoring. However, challenges such as data sparsity, class imbalance, and high computational complexity often lead to poor performance of existing segmentation methods. In this study, we propose a novel hybrid segmentation framework that integrates Point Transformer v3 (PTv3) and Squeeze-Excitation (SE) attention mechanisms to enhance feature extraction and improve segmentation accuracy. The preprocessing pipeline incorporates voxel grid downsampling to reduce redundancy, fixed-size point cloud preparation (1024 points per sample) for computational consistency, and Synthetic Minority Over-Sampling Technique (SMOTE) to address class imbalance by generating synthetic data for underrepresented classes. The proposed PTv3-SE model is built on hierarchical attention mechanisms and channel-wise recalibration to efficiently capture both local and global features in sparse and noisy point clouds. The approach incorporates compression, based on Octrees and Truncated Pyramid subdivision, as well as internal ”data” transmission optimization, with feature prioritization and adaptive streaming strategy. The model was evaluated using the SemanticKITTI and ShapeNet datasets, demonstrating robust state-of-the-art level segmentation performance, with good performance of 93.4% accuracy, 95.03% precision, 93.44% recall, and an F1 score of 93.98%, while also maintaining high computational efficiency of 2s per frame, almost a second faster than the closest competitor model. Compared to existing methods, our approach measurably improves segmentation accuracy and efficiency. On the SemanticKITTI dataset, the framework achieves a mIoU of 87.5%, which surpasses PointNet++ (74.1%) and DGCNN (72.5%) by a significant margin. Similarly, on the ShapeNet dataset, the framework shows good object-level segmentation with an mIoU of 89.4%, reflecting its robustness in capturing fine-grained details.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101891"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077803","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号