Pub Date : 2025-11-17DOI: 10.1016/j.rse.2025.115148
Zhaohui Li , Gabriel Hmimina , Gwendal Latouche , Daniel Berveiller , Abderrahmane Ounis , Yves Goulas , Kamel Soudani
Sun-Induced chlorophyll Fluorescence (SIF) is the most promising remote-sensing proxy of Gross Primary Production (GPP) in terrestrial ecosystems. However, the estimation of GPP using SIF is challenging when plants experience stress, particularly during extreme climatic events whose frequency is projected to increase in the future. Recently, the feasibility of canopy-level active chlorophyll fluorescence measurements (LED-induced chlorophyll fluorescence), which directly measure the apparent fluorescence yield (FyieldLIF), has provided new perspectives on detecting the responses of plants to stress. This study was conducted during the summer 2022 European heat waves in a mixed temperate deciduous broadleaf forest, located in the French Fontainebleau-Barbeau station. Continuous measurements of carbon dioxide (CO2) and energy exchanges, SIF, FyieldLIF, and ancillary environmental variables were acquired. We investigated how heat-wave induced high atmospheric dryness, measured as Vapor Pressure Deficit, affected canopy chlorophyll fluorescence (both SIF and FyieldLIF) and GPP, as well as their relationships. At the half-hourly scale, our results revealed a decrease of the correlation between SIF and GPP (R2 decreased from 0.49 to 0.17) at high atmospheric dryness. In contrast, the correlation between FyieldLIF and GPP increased significantly under high atmospheric dryness (R2 increased from 0.07 to 0.43). However, at the daily scale, the correlations between SIF and GPP and between FyieldLIF and GPP showed an overall increase compared to the half-hourly scale, suggesting a time-scale-dependent response of these relationships to atmospheric dryness. This study also highlighted FyieldLIF's advantage in detecting plant responses to high atmospheric dryness, and emphasized the potential of canopy-level active chlorophyll fluorescence for assessing the chlorophyll fluorescence-photosynthesis relationship under extreme climatic conditions.
{"title":"Atmospheric dryness effects on canopy chlorophyll fluorescence and Gross Primary Production (GPP) in a deciduous forest during heat waves","authors":"Zhaohui Li , Gabriel Hmimina , Gwendal Latouche , Daniel Berveiller , Abderrahmane Ounis , Yves Goulas , Kamel Soudani","doi":"10.1016/j.rse.2025.115148","DOIUrl":"10.1016/j.rse.2025.115148","url":null,"abstract":"<div><div>Sun-Induced chlorophyll Fluorescence (SIF) is the most promising remote-sensing proxy of Gross Primary Production (GPP) in terrestrial ecosystems. However, the estimation of GPP using SIF is challenging when plants experience stress, particularly during extreme climatic events whose frequency is projected to increase in the future. Recently, the feasibility of canopy-level active chlorophyll fluorescence measurements (LED-induced chlorophyll fluorescence), which directly measure the apparent fluorescence yield (F<sub>yieldLIF</sub>), has provided new perspectives on detecting the responses of plants to stress. This study was conducted during the summer 2022 European heat waves in a mixed temperate deciduous broadleaf forest, located in the French Fontainebleau-Barbeau station. Continuous measurements of carbon dioxide (CO<sub>2</sub>) and energy exchanges, SIF, F<sub>yieldLIF</sub>, and ancillary environmental variables were acquired. We investigated how heat-wave induced high atmospheric dryness, measured as Vapor Pressure Deficit, affected canopy chlorophyll fluorescence (both SIF and F<sub>yieldLIF</sub>) and GPP, as well as their relationships. At the half-hourly scale, our results revealed a decrease of the correlation between SIF and GPP (R<sup>2</sup> decreased from 0.49 to 0.17) at high atmospheric dryness. In contrast, the correlation between F<sub>yieldLIF</sub> and GPP increased significantly under high atmospheric dryness (R<sup>2</sup> increased from 0.07 to 0.43). However, at the daily scale, the correlations between SIF and GPP and between F<sub>yieldLIF</sub> and GPP showed an overall increase compared to the half-hourly scale, suggesting a time-scale-dependent response of these relationships to atmospheric dryness. This study also highlighted F<sub>yieldLIF</sub>'s advantage in detecting plant responses to high atmospheric dryness, and emphasized the potential of canopy-level active chlorophyll fluorescence for assessing the chlorophyll fluorescence-photosynthesis relationship under extreme climatic conditions.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"333 ","pages":"Article 115148"},"PeriodicalIF":11.4,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145532048","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The rapid expansion of rice-crayfish farming in China has triggered significant land-use transformations, yet long-term mapping of these patterns remains challenging due to sample limitations and spectral complexities. This study developed a robust classification framework integrating synergistic sample generation and hierarchical classification to address this gap. We first proposed a sample generation method integrating temporal migration with feature-based enlargement strategy, then designed a two-layer stratified classification approach combining machine learning (Random Forest) with phenology-based techniques. Applied to the Jianghan Plain (2013−2022), our framework achieved high accuracy, with overall accuracy higher than 87 % annually and correlation around 0.90 with statistical data. Critical land use dynamics were noticed as follows: (1) Land-use transitions accelerated during 2016–2022, with rice-crayfish expanding predominantly at the expense of traditional rice cultivation (77 % ± 4.76 %) of rice-crayfish fields originated from rice-based cropping). (2) Single-rice areas declined by 24 % ± 3.02 %, while rapeseed-rice and wheat-rice systems decreased by 21 % ± 5.41 % and 26 % ± 5.32 %, respectively. (3) Conversions from dryland and water bodies to rice-crayfish emerged during 2019–2022, a later phase of expansion when the conversion to rice-crayfish became widespread. Overall, this study proposed a robust land use type classification framework for complex regions with limited samples in long-term, providing a transferable solution for monitoring land-system changes under rapid transitions. By revealing the transformative impact of rice-crayfish system expansion on traditional land use patterns, this study highlights its substantial effects on conventional rice cultivation and offers valuable insights for formulating adaptive land management strategies that support ecological sustainability and regional food security.
{"title":"Characterizing land use changes triggered by crop-aquaculture co-cultivation from 2013 to 2022 based on a robust classification framework: Illustration in Jianghan Plain, China","authors":"Yanbing Wei , Wenjuan Li , Peng Zhu , Qiangyi Yu , Wenbin Wu","doi":"10.1016/j.rse.2025.115142","DOIUrl":"10.1016/j.rse.2025.115142","url":null,"abstract":"<div><div>The rapid expansion of rice-crayfish farming in China has triggered significant land-use transformations, yet long-term mapping of these patterns remains challenging due to sample limitations and spectral complexities. This study developed a robust classification framework integrating synergistic sample generation and hierarchical classification to address this gap. We first proposed a sample generation method integrating temporal migration with feature-based enlargement strategy, then designed a two-layer stratified classification approach combining machine learning (Random Forest) with phenology-based techniques. Applied to the Jianghan Plain (2013−2022), our framework achieved high accuracy, with overall accuracy higher than 87 % annually and correlation around 0.90 with statistical data. Critical land use dynamics were noticed as follows: (1) Land-use transitions accelerated during 2016–2022, with rice-crayfish expanding predominantly at the expense of traditional rice cultivation (77 % ± 4.76 %) of rice-crayfish fields originated from rice-based cropping). (2) Single-rice areas declined by 24 % ± 3.02 %, while rapeseed-rice and wheat-rice systems decreased by 21 % ± 5.41 % and 26 % ± 5.32 %, respectively. (3) Conversions from dryland and water bodies to rice-crayfish emerged during 2019–2022, a later phase of expansion when the conversion to rice-crayfish became widespread. Overall, this study proposed a robust land use type classification framework for complex regions with limited samples in long-term, providing a transferable solution for monitoring land-system changes under rapid transitions. By revealing the transformative impact of rice-crayfish system expansion on traditional land use patterns, this study highlights its substantial effects on conventional rice cultivation and offers valuable insights for formulating adaptive land management strategies that support ecological sustainability and regional food security.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"333 ","pages":"Article 115142"},"PeriodicalIF":11.4,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145532050","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-17DOI: 10.1016/j.rse.2025.115143
Aoxiang Yan , Shanzhen Li , Xiaoying Jin , Shuai Huang , Wenhui Wang , Jianjun Tang , Anyuan Li , Ze Zhang , Shengrong Zhang , Jinbang Zhai , Lanzhi Lü , Ruixia He , Xiaoying Li , Wei Shan , Ying Guo , Huijun Jin
This study assesses the stability of the Bei'an–Hei'he Highway (BHH), located near the southern limit of latitudinal permafrost in the Xiao Xing'anling Mountains, Northeast China, where permafrost degradation is intensifying under combined climatic and anthropogenic influences. Freeze-thaw-induced ground deformation and related periglacial hazards remain poorly quantified, limiting regional infrastructure resilience. We developed an integrated framework that fuses multi-source InSAR (ALOS, Sentinel-1, ALOS-2), unmanned aerial vehicle (UAV) photogrammetry, electrical resistivity tomography (ERT), and theoretical modeling to characterize cumulative deformation, evaluate present stability, and project future dynamics. Results reveal long-term deformation rates from −35 to +40 mm/yr within a 1-km buffer on each side of the BHH, with seasonal amplitudes up to 11 mm. Sentinel-1, with its 12-day revisit cycle, demonstrated superior capability for monitoring the Xing'an permafrost. Deformation patterns were primarily controlled by air temperature, while precipitation and the topographic wetness index enhanced spatial heterogeneity through thermo-hydrological coupling. Wavelet analysis identified a 334-day deformation cycle, lagging climate forcing by ∼107 days due to the insulating effects of peat. Early-warning analysis classified 4.99 % of the highway length as high-risk (subsidence <−18.18 mm/yr or frost heave >10.91 mm/yr). The InSAR-based landslide prediction model achieved high accuracy (Area Under the Receiver Operating Characteristic (ROC) Curve, or AUC = 0.9486), validated through field surveys of subsidence, cracking, and slow-moving failures. The proposed ‘past-present-future’ framework demonstrates the potential of multi-sensor integration for permafrost monitoring and provides a transferable approach for assessing infrastructure stability in cold regions.
{"title":"Multi-source assessment of permafrost deformation along the Bei'an–Hei'he highway in Northeast China","authors":"Aoxiang Yan , Shanzhen Li , Xiaoying Jin , Shuai Huang , Wenhui Wang , Jianjun Tang , Anyuan Li , Ze Zhang , Shengrong Zhang , Jinbang Zhai , Lanzhi Lü , Ruixia He , Xiaoying Li , Wei Shan , Ying Guo , Huijun Jin","doi":"10.1016/j.rse.2025.115143","DOIUrl":"10.1016/j.rse.2025.115143","url":null,"abstract":"<div><div>This study assesses the stability of the Bei'an–Hei'he Highway (BHH), located near the southern limit of latitudinal permafrost in the Xiao Xing'anling Mountains, Northeast China, where permafrost degradation is intensifying under combined climatic and anthropogenic influences. Freeze-thaw-induced ground deformation and related periglacial hazards remain poorly quantified, limiting regional infrastructure resilience. We developed an integrated framework that fuses multi-source InSAR (ALOS, Sentinel-1, ALOS-2), unmanned aerial vehicle (UAV) photogrammetry, electrical resistivity tomography (ERT), and theoretical modeling to characterize cumulative deformation, evaluate present stability, and project future dynamics. Results reveal long-term deformation rates from −35 to +40 mm/yr within a 1-km buffer on each side of the BHH, with seasonal amplitudes up to 11 mm. Sentinel-1, with its 12-day revisit cycle, demonstrated superior capability for monitoring the Xing'an permafrost. Deformation patterns were primarily controlled by air temperature, while precipitation and the topographic wetness index enhanced spatial heterogeneity through thermo-hydrological coupling. Wavelet analysis identified a 334-day deformation cycle, lagging climate forcing by ∼107 days due to the insulating effects of peat. Early-warning analysis classified 4.99 % of the highway length as high-risk (subsidence <−18.18 mm/yr or frost heave >10.91 mm/yr). The InSAR-based landslide prediction model achieved high accuracy (Area Under the Receiver Operating Characteristic (ROC) Curve, or AUC = 0.9486), validated through field surveys of subsidence, cracking, and slow-moving failures. The proposed ‘past-present-future’ framework demonstrates the potential of multi-sensor integration for permafrost monitoring and provides a transferable approach for assessing infrastructure stability in cold regions.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"333 ","pages":"Article 115143"},"PeriodicalIF":11.4,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145532051","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-15DOI: 10.1016/j.rse.2025.115131
Almudena García-García , Pietro Stradiotti , Federico Di Paolo , Paolo Filippucci , Milan Fischer , Matěj Orság , Luca Brocca , Jian Peng , Wouter Dorigo , Alexander Gruber , Bram Droppers , Niko Wanders , Arjen Haag , Albrecht Weerts , Ehsan Modiri , Oldrich Rakovec , Félix Francés , Matteo Dall’Amico , Martha Anderson , Christopher Hain , Luis Samaniego
The increasing frequency and severity of hydrological extremes demand the development of early warning systems and effective adaptation and mitigation strategies. Such systems and strategies require spatially detailed hydrological predictions, mostly provided by hydrological models. However, current state-of-the-art hydrological predictions remain limited in their spatial resolution. A proposed solution is the integration of high-resolution (1 km) Earth observation (EO) products in hydrological modelling in order to reach hyper-resolution (approximately 1 km2). Nonetheless, proper use of these data in hydrological modelling requires a comprehensive characterization of their uncertainties. Here, we evaluate the performance of high-resolution EO products of four hydrological variables (7 precipitation products, 5 snow cover area products, 6 surface soil moisture products, and 6 actual evapotranspiration products) against observational references. Two merged EO precipitation products at 1 km resolution (merged IMERG-SM2A and merged ERA5-IMERG-SM2A) reached correlation coefficients 0.5 with the benchmark reference over most areas and are recommended for hyper-resolution hydrological modelling over Europe. The MODIS (resolution of 250 m) and Sentinel-2/Landsat-8 (resolution of 20/30 m) snow cover products showed the highest classification accuracy and were selected as the best choice for the use of snow cover area products in hyper-resolution hydrological modelling. For surface soil moisture, the NSIDC SMAP product at 1 km resolution yielded correlation coefficients 0.6 at most stations and is recommended for hyper-resolution hydrological modelling. Finally, evapotranspiration products showed similar performances at the selected flux sites (correlations coefficients 0.8). While the MODIS-Terra/Aqua evapotranspiration products (MOD16A2/MYD16A2) offer higher spatial resolution (500 m), making them potentially advantageous for hyper-resolution hydrological modelling, their temporal resolution is coarser (8-day intervals). In contrast, products like ETMonitor (1 km), ALEXI, and HOLAPS (5 km) provide daily estimates, albeit at lower spatial resolution. The assimilation of the proposed high-resolution products in models individually or in combination could lead us to hyper-resolution hydrological modelling. Still, developing integration workflows is required to overcome difficulties related to scale mismatches and data-gaps.
{"title":"Intercomparison of Earth Observation products for hyper-resolution hydrological modelling over Europe","authors":"Almudena García-García , Pietro Stradiotti , Federico Di Paolo , Paolo Filippucci , Milan Fischer , Matěj Orság , Luca Brocca , Jian Peng , Wouter Dorigo , Alexander Gruber , Bram Droppers , Niko Wanders , Arjen Haag , Albrecht Weerts , Ehsan Modiri , Oldrich Rakovec , Félix Francés , Matteo Dall’Amico , Martha Anderson , Christopher Hain , Luis Samaniego","doi":"10.1016/j.rse.2025.115131","DOIUrl":"10.1016/j.rse.2025.115131","url":null,"abstract":"<div><div>The increasing frequency and severity of hydrological extremes demand the development of early warning systems and effective adaptation and mitigation strategies. Such systems and strategies require spatially detailed hydrological predictions, mostly provided by hydrological models. However, current state-of-the-art hydrological predictions remain limited in their spatial resolution. A proposed solution is the integration of high-resolution (<span><math><mo><</mo></math></span>1 km) Earth observation (EO) products in hydrological modelling in order to reach hyper-resolution (approximately 1<!--> <!-->km<sup>2</sup>). Nonetheless, proper use of these data in hydrological modelling requires a comprehensive characterization of their uncertainties. Here, we evaluate the performance of high-resolution EO products of four hydrological variables (7 precipitation products, 5 snow cover area products, 6 surface soil moisture products, and 6 actual evapotranspiration products) against observational references. Two merged EO precipitation products at 1<!--> <!-->km resolution (merged IMERG-SM2A and merged ERA5-IMERG-SM2A) reached correlation coefficients <span><math><mo>></mo></math></span>0.5 with the benchmark reference over most areas and are recommended for hyper-resolution hydrological modelling over Europe. The MODIS (resolution of 250 m) and Sentinel-2/Landsat-8 (resolution of 20/30 m) snow cover products showed the highest classification accuracy and were selected as the best choice for the use of snow cover area products in hyper-resolution hydrological modelling. For surface soil moisture, the NSIDC SMAP product at 1<!--> <!-->km resolution yielded correlation coefficients <span><math><mo>></mo></math></span>0.6 at most stations and is recommended for hyper-resolution hydrological modelling. Finally, evapotranspiration products showed similar performances at the selected flux sites (correlations coefficients <span><math><mo>></mo></math></span> 0.8). While the MODIS-Terra/Aqua evapotranspiration products (MOD16A2/MYD16A2) offer higher spatial resolution (500 m), making them potentially advantageous for hyper-resolution hydrological modelling, their temporal resolution is coarser (8-day intervals). In contrast, products like ETMonitor (1 km), ALEXI, and HOLAPS (5 km) provide daily estimates, albeit at lower spatial resolution. The assimilation of the proposed high-resolution products in models individually or in combination could lead us to hyper-resolution hydrological modelling. Still, developing integration workflows is required to overcome difficulties related to scale mismatches and data-gaps.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"333 ","pages":"Article 115131"},"PeriodicalIF":11.4,"publicationDate":"2025-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145516140","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-14DOI: 10.1016/j.rse.2025.115134
Zhiqiang Xiong , Guangcai Feng , Zefa Yang , Hua Gao , Wenbin Xu , Lei Zhang , Xiaohua Xu , Zhong Lu , Jun Hu , Zhiwei Li , Jianjun Zhu
Land subsidence (LS) threatens public safety and sustainable socioeconomic development across the United States (US). Wide-area LS mapping plays a crucial role in improving the characterization of LS, thus contributing to a better understanding of its impacts beyond national borders. However, a nationwide LS map for the US is still lacking, and mapping LS at this scale faces challenges such as atmospheric artifacts, computational limitations, and data calibration issues. We develop a strategy for mapping and identifying LS on a nationwide scale. This strategy includes average deformation calculation, mosaicking of Interferometric Synthetic Aperture Radar (InSAR) results, and LS areas extraction using deep learning. Using this strategy and extensive SAR data, we present the first high-resolution, nationwide LS map for the contiguous U. S. (CONUS), derived from 23,391 ascending Sentinel-1 images acquired between 2019 and 2021. Our analysis reveals that LS affects all 48 states and the District of Columbia, covering ∼45,666 km2. Approximately 73 % of LS occurs in cultivated lands, 10 % in wetlands, with groundwater overexploitation being the predominant driver of subsidence. These subsiding areas expose 6.26 million people, 160,071 buildings, and critical infrastructure, including thousands of kilometers of railways and roads. Although some measures have been implemented to mitigate LS, existing cases suggest that long-term groundwater management is essential, and significant challenges remain in addressing severe LS. This national-scale assessment provides critical data for targeted management and informs future monitoring and mitigation strategies.
{"title":"Nationwide mapping and characterization of land subsidence in the United States using InSAR","authors":"Zhiqiang Xiong , Guangcai Feng , Zefa Yang , Hua Gao , Wenbin Xu , Lei Zhang , Xiaohua Xu , Zhong Lu , Jun Hu , Zhiwei Li , Jianjun Zhu","doi":"10.1016/j.rse.2025.115134","DOIUrl":"10.1016/j.rse.2025.115134","url":null,"abstract":"<div><div>Land subsidence (LS) threatens public safety and sustainable socioeconomic development across the United States (US). Wide-area LS mapping plays a crucial role in improving the characterization of LS, thus contributing to a better understanding of its impacts beyond national borders. However, a nationwide LS map for the US is still lacking, and mapping LS at this scale faces challenges such as atmospheric artifacts, computational limitations, and data calibration issues. We develop a strategy for mapping and identifying LS on a nationwide scale. This strategy includes average deformation calculation, mosaicking of Interferometric Synthetic Aperture Radar (InSAR) results, and LS areas extraction using deep learning. Using this strategy and extensive SAR data, we present the first high-resolution, nationwide LS map for the contiguous U. S. (CONUS), derived from 23,391 ascending Sentinel-1 images acquired between 2019 and 2021. Our analysis reveals that LS affects all 48 states and the District of Columbia, covering ∼45,666 km<sup>2</sup>. Approximately 73 % of LS occurs in cultivated lands, 10 % in wetlands, with groundwater overexploitation being the predominant driver of subsidence. These subsiding areas expose 6.26 million people, 160,071 buildings, and critical infrastructure, including thousands of kilometers of railways and roads. Although some measures have been implemented to mitigate LS, existing cases suggest that long-term groundwater management is essential, and significant challenges remain in addressing severe LS. This national-scale assessment provides critical data for targeted management and informs future monitoring and mitigation strategies.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"333 ","pages":"Article 115134"},"PeriodicalIF":11.4,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145509262","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-14DOI: 10.1016/j.rse.2025.115133
Africa I. Flores-Anderson , Jeffrey A. Cardille , Josef Kellndorfer , Franz J. Meyer , Pontus Olofsson
Operational national forest monitoring requires frequent, reliable observations for timely detection of deforestation and other forest changes. However, tropical forests, which encompass all current major deforestation fronts, are often hampered by persistent cloud cover. While radar remote sensing offers a compelling complement to optical data, a significant research gap remains in distinguishing the strengths and limitations of different radar sensors to the multiple stages of forest disturbance. Given the impending rapid expansion of available data outside the C band, it is imperative to assess the ability of these sensors to rapidly and accurately detect changes of high importance to the remote sensing community.
To address this gap, we investigated the sensitivity of two freely available radar datasets — JAXA’s ALOS-2 PALSAR-2 (L-band) and ESA’s Sentinel-1 (C-band) — to distinct stages of tropical forest loss. With a particular focus on detecting the earliest stage of deforestation, we compared backscatter values, SAR indices, and statistical metrics against time series data from 92 locations over three years in the Amazon. In these locations, early deforestation, biomass burning, and vegetation regrowth were occurring in different stages of the full conversion process. Our analysis revealed that the L-band-derived Radar Forest Degradation Index (RFDI) is highly sensitive to early deforestation, even when biomass remains on the ground soon after cutting. In contrast, C-band information showed limited ability to sense this critical initial stage of change, but was much stronger at detecting later deforestation stages. Our results point the way toward combining information from the upcoming L-Band NISAR mission with the existing C-band information to produce a multi-component system that can accurately detect deforestation in all its temporal stages.
{"title":"On the sensitivity of SAR C- and L-band dual-polarized data for detection of early deforestation in the tropics","authors":"Africa I. Flores-Anderson , Jeffrey A. Cardille , Josef Kellndorfer , Franz J. Meyer , Pontus Olofsson","doi":"10.1016/j.rse.2025.115133","DOIUrl":"10.1016/j.rse.2025.115133","url":null,"abstract":"<div><div>Operational national forest monitoring requires frequent, reliable observations for timely detection of deforestation and other forest changes. However, tropical forests, which encompass all current major deforestation fronts, are often hampered by persistent cloud cover. While radar remote sensing offers a compelling complement to optical data, a significant research gap remains in distinguishing the strengths and limitations of different radar sensors to the multiple stages of forest disturbance. Given the impending rapid expansion of available data outside the C band, it is imperative to assess the ability of these sensors to rapidly and accurately detect changes of high importance to the remote sensing community.</div><div>To address this gap, we investigated the sensitivity of two freely available radar datasets — JAXA’s ALOS-2 PALSAR-2 (L-band) and ESA’s Sentinel-1 (C-band) — to distinct stages of tropical forest loss. With a particular focus on detecting the earliest stage of deforestation, we compared backscatter values, SAR indices, and statistical metrics against time series data from 92 locations over three years in the Amazon. In these locations, early deforestation, biomass burning, and vegetation regrowth were occurring in different stages of the full conversion process. Our analysis revealed that the L-band-derived Radar Forest Degradation Index (RFDI) is highly sensitive to early deforestation, even when biomass remains on the ground soon after cutting. In contrast, C-band information showed limited ability to sense this critical initial stage of change, but was much stronger at detecting later deforestation stages. Our results point the way toward combining information from the upcoming L-Band NISAR mission with the existing C-band information to produce a multi-component system that can accurately detect deforestation in all its temporal stages.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"333 ","pages":"Article 115133"},"PeriodicalIF":11.4,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145509943","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-14DOI: 10.1016/j.rse.2025.115102
George Worrall, Jasmeet Judge
Many commodity crops have growth stages during which they are particularly vulnerable to stress-induced yield loss. In-season crop progress information is useful for quantifying crop risk, and satellite remote sensing (RS) can be used to track progress at regional scales. At present, all existing RS-based crop progress estimation (CPE) methods which target crop-specific stages rely on ground truth data for training/calibration. Such data are collected via field trials or surveys. This reliance on ground survey data confines CPE methods to surveyed regions, limiting their utility. In this study, a new method is developed for conducting RS-based in-season CPE in unsurveyed regions by combining data from surveyed regions with synthetic crop progress data generated for an unsurveyed region of interest. Corn-growing zones in Argentina were used as surrogate ‘unsurveyed’ regions. These zones have climates and dual planting systems which differ from the single planting system in the US Midwest – the surveyed region in this study. Existing weather generation, crop growth, and optical radiative transfer models were linked to produce synthetic weather, crop progress, and canopy reflectance data. These data mimic weather, cultivars, and cropping practices in the unsurveyed region. A neural network (NN) method based upon bi-directional Long Short-Term Memory was trained separately on surveyed data, synthetic data, and two different combinations of surveyed and synthetic data. In the absence of real validation data in unsurveyed regions, a stopping criterion was developed which uses the weighted divergence of surveyed and synthetic data validation loss. F score was modified to measure CPE accuracy when the NN was trained on each data combination, with scores based on over- and under-estimates of crop progress throughout the season. Including synthetic data during training improved performance in 9 out of 11 corn-growing zones in Argentina. Net F scores across all crop progress stages increased by 8.7% when trained on a combination of surveyed region and synthetic data, and overall performance was only 21% lower than when the NN was trained on surveyed data and applied in the US Midwest. Performance gain from synthetic data was greatest in zones with dual planting windows, while the inclusion of surveyed region data from the US Midwest helped mitigate NN sensitivity to noise in NDVI data. Overall results suggest in-season CPE in other unsurveyed regions may be possible with increased quantity and variety of synthetic crop progress data.
{"title":"In-season crop progress in unsurveyed regions using networks trained on synthetic data","authors":"George Worrall, Jasmeet Judge","doi":"10.1016/j.rse.2025.115102","DOIUrl":"10.1016/j.rse.2025.115102","url":null,"abstract":"<div><div>Many commodity crops have growth stages during which they are particularly vulnerable to stress-induced yield loss. In-season crop progress information is useful for quantifying crop risk, and satellite remote sensing (RS) can be used to track progress at regional scales. At present, all existing RS-based crop progress estimation (CPE) methods which target crop-specific stages rely on ground truth data for training/calibration. Such data are collected via field trials or surveys. This reliance on ground survey data confines CPE methods to surveyed regions, limiting their utility. In this study, a new method is developed for conducting RS-based in-season CPE in unsurveyed regions by combining data from surveyed regions with synthetic crop progress data generated for an unsurveyed region of interest. Corn-growing zones in Argentina were used as surrogate ‘unsurveyed’ regions. These zones have climates and dual planting systems which differ from the single planting system in the US Midwest – the surveyed region in this study. Existing weather generation, crop growth, and optical radiative transfer models were linked to produce synthetic weather, crop progress, and canopy reflectance data. These data mimic weather, cultivars, and cropping practices in the unsurveyed region. A neural network (NN) method based upon bi-directional Long Short-Term Memory was trained separately on surveyed data, synthetic data, and two different combinations of surveyed and synthetic data. In the absence of real validation data in unsurveyed regions, a stopping criterion was developed which uses the weighted divergence of surveyed and synthetic data validation loss. F<span><math><msub><mrow></mrow><mrow><mn>1</mn></mrow></msub></math></span> score was modified to measure CPE accuracy when the NN was trained on each data combination, with scores based on over- and under-estimates of crop progress throughout the season. Including synthetic data during training improved performance in 9 out of 11 corn-growing zones in Argentina. Net F<span><math><msub><mrow></mrow><mrow><mn>1</mn></mrow></msub></math></span> scores across all crop progress stages increased by 8.7% when trained on a combination of surveyed region and synthetic data, and overall performance was only 21% lower than when the NN was trained on surveyed data and applied in the US Midwest. Performance gain from synthetic data was greatest in zones with dual planting windows, while the inclusion of surveyed region data from the US Midwest helped mitigate NN sensitivity to noise in NDVI data. Overall results suggest in-season CPE in other unsurveyed regions may be possible with increased quantity and variety of synthetic crop progress data.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"333 ","pages":"Article 115102"},"PeriodicalIF":11.4,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145516024","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-13DOI: 10.1016/j.rse.2025.115135
Jianglei Xu , Shunlin Liang , Han Ma , Yongzhe Chen , Wenyuan Li , Yichuan Ma , Xiang Zhao , Bo Jiang , Xiaotong Zhang , Shikang Guan
Accurate characterization of the surface radiation budget (SRB) is essential for understanding the Earth's climate. Satellite-based SRB products are typically generated through estimating different components separately using various algorithms, thereby resulting in varying uncertainties and poor conservation. This study developed the SRB conservation constraint multi-task learning densely connected convolutional neural network models to jointly estimate global daily SRB at 1 km spatial resolution from MODIS observations spanning 2000–2023. These observations include reflectance from bands 1–5, 7, and 19; thermal radiance from bands 28–29 and 31–34; and ancillary information such as elevation, solar-viewing geometry, and GLASS-MODIS surface longwave radiation. Validation results against 224 sites over three years showed that the RMSEs of daily estimates for downward, upward, and net shortwave radiation were 29.38, 20.73, and 23.14 Wm−2, respectively; for downward, upward, and net longwave radiation, they were 19.98, 15.69, and 14.70 Wm−2; and for net radiation, it was 24.28 Wm−2. This method improves the underestimations of downward and net shortwave radiation in the MCD18A1, GLASS-MODIS, and BESS products; the accuracy of retrievals for downward and net longwave radiation is better than those from GLASS-MODIS and CERES-SYN. The estimates also reduce the non-conservation by 26.69 % compared to GLASS-MODIS. These improvements lie in the method's ability to enhance retrieval accuracy by utilizing cross-domain features from all components and to address non-conservation issues in SRB retrievals through the constraint of SRB conservation in the training. The SRB estimates exhibit great spatiotemporal consistency with other SRB products, except for regional reflected solar radiation and net radiation. The high accuracy and great conservation facilitate understanding of the coordinated variation of the SRB components, and these new SRB products would benefit a variety of fields, such as climate, ecology, and hydrology. These products are freely accessed at www.glass.hku.hk.
{"title":"Joint estimation of global daily 1 km surface radiation budget components from MODIS observations (2000−2023) using conservation-constrained deep neural networks","authors":"Jianglei Xu , Shunlin Liang , Han Ma , Yongzhe Chen , Wenyuan Li , Yichuan Ma , Xiang Zhao , Bo Jiang , Xiaotong Zhang , Shikang Guan","doi":"10.1016/j.rse.2025.115135","DOIUrl":"10.1016/j.rse.2025.115135","url":null,"abstract":"<div><div>Accurate characterization of the surface radiation budget (SRB) is essential for understanding the Earth's climate. Satellite-based SRB products are typically generated through estimating different components separately using various algorithms, thereby resulting in varying uncertainties and poor conservation. This study developed the SRB conservation constraint multi-task learning densely connected convolutional neural network models to jointly estimate global daily SRB at 1 km spatial resolution from MODIS observations spanning 2000–2023. These observations include reflectance from bands 1–5, 7, and 19; thermal radiance from bands 28–29 and 31–34; and ancillary information such as elevation, solar-viewing geometry, and GLASS-MODIS surface longwave radiation. Validation results against 224 sites over three years showed that the RMSEs of daily estimates for downward, upward, and net shortwave radiation were 29.38, 20.73, and 23.14 Wm<sup>−2</sup>, respectively; for downward, upward, and net longwave radiation, they were 19.98, 15.69, and 14.70 Wm<sup>−2</sup>; and for net radiation, it was 24.28 Wm<sup>−2</sup>. This method improves the underestimations of downward and net shortwave radiation in the MCD18A1, GLASS-MODIS, and BESS products; the accuracy of retrievals for downward and net longwave radiation is better than those from GLASS-MODIS and CERES-SYN. The estimates also reduce the non-conservation by 26.69 % compared to GLASS-MODIS. These improvements lie in the method's ability to enhance retrieval accuracy by utilizing cross-domain features from all components and to address non-conservation issues in SRB retrievals through the constraint of SRB conservation in the training. The SRB estimates exhibit great spatiotemporal consistency with other SRB products, except for regional reflected solar radiation and net radiation. The high accuracy and great conservation facilitate understanding of the coordinated variation of the SRB components, and these new SRB products would benefit a variety of fields, such as climate, ecology, and hydrology. These products are freely accessed at <span><span>www.glass.hku.hk</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"333 ","pages":"Article 115135"},"PeriodicalIF":11.4,"publicationDate":"2025-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145499163","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-12DOI: 10.1016/j.rse.2025.115136
Yoojin Kang , Jaese Lee , Jungho Im
Satellite remote sensing has provided valuable information on fire radiative power (FRP), which is an important indicator for estimating biomass burning emissions and understanding fire dynamics. However, FRP evaluation faces challenges due to limited field measurements and the high variability of satellite sensor characteristics. Previous studies have tried to quantify the uncertainty of FRP by intercomparison, but they are limited to small numbers of cases and are not free from the detection capacity. To address this issue, we presented a comprehensive global evaluation of FRP from three satellite sensors—Moderate Resolution Imaging Spectroradiometer (MODIS), Visible Infrared Imaging Radiometer Suite (VIIRS), and Sea and Land Surface Temperature Radiometer (SLSTR)—by integrating all possible fire clusters after minimizing differences in fire occurrence detection while preserving inherent differences in detection extent across sensors. Furthermore, we applied extended triple collocation analysis (ETC) to evaluate the consistency of FRP without relying on true reference data for the first time. Intercomparative results highlight robust consistency when the fire clusters overlap well across all three sensors, even under varying sample selection criteria. Notably, SLSTR and VIIRS have slightly higher FRP than MODIS, even after aligning detected fire events, due to the superiority of observing small or weak fires at the edge of fire clusters. ETC revealed high consistency in boreal forests, where large-scale, strong fire clusters are well matched. In contrast, uncertainties remain in South Africa because of highly variable fire dynamics in that area. This study contributes to understanding the regional characteristics of FRP and provides a robust framework for global-scale FRP assessments as new satellite datasets become available.
{"title":"Comprehensive global fire radiative power evaluation by minimizing detection bias with intercomparison and extended triple collocation analysis","authors":"Yoojin Kang , Jaese Lee , Jungho Im","doi":"10.1016/j.rse.2025.115136","DOIUrl":"10.1016/j.rse.2025.115136","url":null,"abstract":"<div><div>Satellite remote sensing has provided valuable information on fire radiative power (FRP), which is an important indicator for estimating biomass burning emissions and understanding fire dynamics. However, FRP evaluation faces challenges due to limited field measurements and the high variability of satellite sensor characteristics. Previous studies have tried to quantify the uncertainty of FRP by intercomparison, but they are limited to small numbers of cases and are not free from the detection capacity. To address this issue, we presented a comprehensive global evaluation of FRP from three satellite sensors—Moderate Resolution Imaging Spectroradiometer (MODIS), Visible Infrared Imaging Radiometer Suite (VIIRS), and Sea and Land Surface Temperature Radiometer (SLSTR)—by integrating all possible fire clusters after minimizing differences in fire occurrence detection while preserving inherent differences in detection extent across sensors. Furthermore, we applied extended triple collocation analysis (ETC) to evaluate the consistency of FRP without relying on true reference data for the first time. Intercomparative results highlight robust consistency when the fire clusters overlap well across all three sensors, even under varying sample selection criteria. Notably, SLSTR and VIIRS have slightly higher FRP than MODIS, even after aligning detected fire events, due to the superiority of observing small or weak fires at the edge of fire clusters. ETC revealed high consistency in boreal forests, where large-scale, strong fire clusters are well matched. In contrast, uncertainties remain in South Africa because of highly variable fire dynamics in that area. This study contributes to understanding the regional characteristics of FRP and provides a robust framework for global-scale FRP assessments as new satellite datasets become available.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"333 ","pages":"Article 115136"},"PeriodicalIF":11.4,"publicationDate":"2025-11-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145499139","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-12DOI: 10.1016/j.rse.2025.115091
Yunzhe Zhu, Dan J. Dixon, Yufang Jin
The encroachment of human settlements into wildland areas has expanded the Wildland-Urban Interface (WUI), disrupting ecosystems and increasing wildfire risks. A systematic mapping of woody vegetation and building structures across the wildland-urban continuum is critical for fire risk and ecosystem service assessment, fuel management, conservation, and land use planning. The highly heterogeneous and dynamic nature of this landscape requires a fine-scale monitoring approach. Although machine learning and computer vision have advanced land cover classification using very high resolution imagery (VHR), most studies either focus on identifying only a particular object or a limited study area. We here developed a unified semantic segmentation deep learning model (U-Net) to map trees, shrubs, and building footprints at 0.6 m resolution using the open-source VHR multispectral imagery from the National Aerial Imaging Program (NAIP) for the state of California. A semi-automatic labeling process was adapted to generate a large set of labels for model training and testing with available aerial LiDAR surveys. The validation showed robust performance of the trained U-Net model in predicting canopy-scale trees and shrubs as well as buildings, achieving an overall accuracy of 87.1% and F1 scores of 83.1% for trees and 78.9% for shrubs across different years. Further evaluation demonstrated that this approach captured the fine-scale spatial arrangement of woody vegetation and buildings, and the temporal vegetation dynamics from selective logging, regrowth, and tree mortality following wildfire. The model’s scalability was also shown for county and statewide mapping. Given the availability of open-access NAIP imagery every 2–3 years over the continental US, our scalable approach provides a sub-meter monitoring tool and data to improve ecological and building assessment and fire simulation. These capabilities support adaptive land management to mitigate WUI fire risk and ultimately promote fire-safe communities and resilient ecosystems along the WUI-wildland gradient.
{"title":"Scalable sub-meter mapping of woody vegetation and structures across California’s heterogeneous landscape using deep learning","authors":"Yunzhe Zhu, Dan J. Dixon, Yufang Jin","doi":"10.1016/j.rse.2025.115091","DOIUrl":"10.1016/j.rse.2025.115091","url":null,"abstract":"<div><div>The encroachment of human settlements into wildland areas has expanded the Wildland-Urban Interface (WUI), disrupting ecosystems and increasing wildfire risks. A systematic mapping of woody vegetation and building structures across the wildland-urban continuum is critical for fire risk and ecosystem service assessment, fuel management, conservation, and land use planning. The highly heterogeneous and dynamic nature of this landscape requires a fine-scale monitoring approach. Although machine learning and computer vision have advanced land cover classification using very high resolution imagery (VHR), most studies either focus on identifying only a particular object or a limited study area. We here developed a unified semantic segmentation deep learning model (U-Net) to map trees, shrubs, and building footprints at 0.6 m resolution using the open-source VHR multispectral imagery from the National Aerial Imaging Program (NAIP) for the state of California. A semi-automatic labeling process was adapted to generate a large set of labels for model training and testing with available aerial LiDAR surveys. The validation showed robust performance of the trained U-Net model in predicting canopy-scale trees and shrubs as well as buildings, achieving an overall accuracy of 87.1% and F1 scores of 83.1% for trees and 78.9% for shrubs across different years. Further evaluation demonstrated that this approach captured the fine-scale spatial arrangement of woody vegetation and buildings, and the temporal vegetation dynamics from selective logging, regrowth, and tree mortality following wildfire. The model’s scalability was also shown for county and statewide mapping. Given the availability of open-access NAIP imagery every 2–3 years over the continental US, our scalable approach provides a sub-meter monitoring tool and data to improve ecological and building assessment and fire simulation. These capabilities support adaptive land management to mitigate WUI fire risk and ultimately promote fire-safe communities and resilient ecosystems along the WUI-wildland gradient.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"333 ","pages":"Article 115091"},"PeriodicalIF":11.4,"publicationDate":"2025-11-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145492683","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","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