Timely and reliable crop area estimates are essential for food security assessments. Remote sensing provides scalable, cost-effective solutions, contingent on rigorous uncertainty quantification. When in situ data collection is required, two-stage sampling designs—where primary sampling units (PSUs) partition the study region and secondary sampling units (SSUs) are sampled within selected PSUs—are widely adopted. Although in-season crop type maps (ISCTMs) are increasingly used for stratification in area estimation, their integration into SSU stratification within two-stage frameworks remains limited.This study evaluates the use of ISCTMs to inform SSU stratification in two-stage sampling for multi-crop area estimation. Monte Carlo simulations were run over Ukrainian winter croplands, varying map accuracy (user’s and producer’s accuracy from 0.5 to 0.9), SSU sampling strategy (simple vs. stratified random sampling), estimator type (stratified vs. regression), and design parameters (PSU size and number, total sample size). Performance was assessed for a prevalent crop (winter cereals, 15.8%) and a rare crop (rapeseed, 2.2%). Across 12,600 parameter combinations with 500 repetitions each, stratified SSU sampling combined with a regression estimator consistently minimized estimation variance. User’s accuracy (UA) had a stronger effect than producer’s accuracy in reducing uncertainties, particularly for rare classes. A 5% coefficient of variation was achieved for both crops with UA ≥ 0.8 and PSU sizes ≥10 km using crop-specific stratification. These results demonstrate that ISCTMs enable effective stratification in two-stage designs, supporting simultaneous multi-crop estimation, including rare classes, and substantially improving sampling efficiency and cost-effectiveness of ground surveys.

In this study, we propose a multi-agent explainable AI framework for forest disturbance analysis and scenario-based biomass forecasting using Earth observation data. The system was implemented in an n8n Docker container and integrates four artificial intelligence (AI) agents based on OpenAI models: (1) an Orchestrator Agent for workflow coordination; (2) a Model Designer Agent for ecological model construction and explanation; (3) a Code Agent for automated Python code generation, execution, and preservation within the local file system; and (4) an Explanation Agent for explaining causes of forest disturbance hotspots using news from the State Forest Enterprise “Forests of Ukraine” and annual reports of the State Forest Resources Agency of Ukraine. The proposed framework was evaluated on a series of forest monitoring and ecological forecasting tasks across Ukraine, including biomass change analysis, hotspot detection, explanation of disturbance causes, and simulation of biomass recovery under alternative forest management scenarios. The system successfully identified major disturbance hotspots in war-affected protected areas, as well as underreported hotspots in Western Ukraine, and constructed models of future biomass dynamics. The agents successfully explained the causes of hotspot formation, generated ecological models and code, and preserved all outputs in a local file system for validation and reuse. Across repeated runs, the framework achieved a 93

Accurate cropland area estimation is essential for food security, yet conventional surveys are costly. Satellite-derived area estimates offer a scalable alternative; however, "pixel-counting" from satellite products introduces bias, while probability-based sampling with design-based inference provides unbiased area estimates, but its efficiency depends on map quality. In this study, we introduce an integrated framework linking map accuracy with relative efficiency to optimize stratified sampling designs. We evaluated seven land cover products across six African countries using independent reference data. Our results demonstrate that map selection is a critical determinant of survey costs: the most efficient products (predominantly GLAD and Digital Earth Africa) reduced required sample sizes by 20%-40% compared to less efficient alternatives while maintaining a target 10% coefficient of variation. We found that for rare cropland classes, gains in producers' accuracy improved sampling efficiency more than equivalent gains in users' accuracy. Our sample-based estimates ranged from 1.41 Mha (Rwanda) to 12.66 Mha (Tanzania) and correlated strongly with FAOSTAT arable land statistics (R2 = 0.89). Our framework minimizes the sample size required to achieve a target precision, providing an operational-ready, cost-effective, and statistically rigorous guide for national agricultural monitoring in resource-and data-limited regions.

Cloud masking is an essential task for satellite-based Earth monitoring, and the quality of cloud masks can directly impact the solutions of the downstream Earth monitoring tasks. While significant progress has been made especially for data with desired bands (e.g., thermal bands in Landsat-8), the masking quality on small satellites with higher resolution but fewer spectral bands is still unreliable at high latitudes, where confusion with snow and ice makes the task significantly more challenging. We propose a novel learning-enabled cross-platform ability transfer paradigm that offers a scalable and effective solution to tackle this challenge through a case study using PlanetScope images in the Arctic. A unique characteristic of the new paradigm is that it does not require manual annotations to be collected for PlanetScope images, which is often the bottleneck and the most time-consuming part of machine learning-based cloud masking, especially given the similarity between clouds and snow/ice. To realize this, our approach first designs and creates a new training dataset, Co-Clouds, which contains around 45,000 coincident pairs of PlanetScope and Landsat-8 image patches collected within a nearly simultaneous temporal window. This coincident dataset offers a way to generate large volumes of training data and builds a bridge to transfer Landsat-8’s stronger cloud masking skills in the Arctic to PlanetScope images via data-driven learning. We also show the feasibility of the ability transfer from spectral signatures (e.g., thermal bands) to spatial signatures (e.g., textures). Using our Co-Clouds dataset, we train several deep learning models including both regular-size deep learning models and large foundation models. To validate the quality of the masks, we further create a manually labeled cloud mask dataset for PlanetScope images in the Arctic. Both the quantitative and qualitative results show significant improvements over the current operational cloud masks by PlanetScope. For example, the large foundation models such as SegFormer achieve approximately 20 % higher overall accuracy and 28 % higher producer’s accuracy than the operational cloud masks, while maintaining comparable or better user’s accuracy exceeding 90 %. The new approach is also very easy to implement and extend to other platforms, opening new opportunities for broadcasting advanced skills from one platform to others.

Timely and transparent agricultural statistics are essential for safeguarding global food security. When the war in Ukraine disrupted agricultural reporting - particularly in Russian-held territories - complementary and/or alternative approaches were needed to produce reliable statistics. We developed a fully remote sensing-based framework to estimate areas of two major crops, wheat and rapeseed, from 2022 to 2025. Using Planet and Sentinel-1/2 imagery, we applied clustering techniques to generate in-season crop type maps that supported stratified random sampling in sample-based area estimation, using remotely interpreted reference data. Our estimates closely matched official statistics in Government-controlled areas (RMSE = 0.138 and 0.32 million hectares (Mha) for wheat and rapeseed, respectively), while filling critical data gaps in Russian-controlled regions. Between 2022 and 2025, wheat area declined from 5.14 +/- 0.52 to 4.84 +/- 0.25 Mha in Government-controlled areas and from 2.06 +/- 0.16 to 1.55 +/- 0.09 Mha in Russian-held regions. Rapeseed expanded from 1.16 +/- 0.17 to 1.47 +/- 0.23 Mha (2022-2025) in Government-controlled territories but collapsed in Russian-held areas, from 0.17 +/- 0.01 to 0.05 +/- 0.01 Mha. Our findings underscore the critical role of remote sensing in providing timely, transparent, and independent agricultural statistics to support informed food security and market-stabilizing decision-making.

The full-scale invasion of Ukraine on 24 February 2022 resulted in widespread disruption to its agricultural system. As winter crops were already planted in late 2021, this led to uncertainty regarding whether all the planted fields would be harvested. Monitoring the harvest status was therefore essential for reliable production estimates. As ground-based assessments were no longer feasible in conflict-affected areas, we relied on remote sensing techniques. We developed a method to monitor crop harvest status in-season with the capability to detect fields that were not-harvested.We monitored harvest from 13 June 2022 until 19 September 2022 and found that 94.1% and 87.5% of planted winter crops were harvested in government controlled and temporarily occupied regions, respectively. The highest intensity of not-harvested fields was observed along the occupation boundary. Validation using visually interpreted high-temporal-frequency Planet imagery yielded an overall accuracy of 85%, with an F1-score of 90% for the harvested class and 73% for the not-harvested class.

Super-resolution (SR) techniques have made major advances in reconstructing high-resolution images from low-resolution inputs. The increased resolution provides visual enhancement and utility for monitoring tasks. In particular, SR has been increasingly developed for satellite-based Earth observation, with applications in urban planning, agriculture, ecology, and disaster response. However, existing SR studies and benchmarks typically use fidelity metrics such as PSNR or SSIM, whereas the true utility of super-resolved images lies in supporting downstream tasks such as land cover classification, biomass estimation, and change detection. To bridge this gap, we introduce GeoSR-Bench, a downstream task-integrated SR benchmark dataset to evaluate SR models beyond fidelity metrics. GeoSR-Bench comprises spatially co-located, temporally aligned, and quality-controlled image pairs from about 36,000 locations across diverse land covers, spanning resolutions from 500m to 0.6m. To the best of our knowledge, GeoSR-Bench is the first SR benchmark that directly connects improved image resolution from SR models with downstream Earth monitoring tasks, including land cover segmentation, infrastructure mapping, and biophysical variable estimation. Using GeoSR-Bench, we benchmark GAN, transformer, neural operator, and diffusion-based SR models on perceptual quality and downstream task performance. We conduct experiments with 270 settings, covering 2 cross-platform SR tasks, 9 SR models, 3 downstream task models, and 5 downstream tasks for each SR task. The results show that improvements in traditional SR metrics often do not correlate with gains in task performance, and the correlations can be negative, indicating that these metrics provide limited guidance for selecting superior models for downstream tasks. This reveals the need to integrate downstream tasks into SR model development and evaluation.

The ongoing war in Ukraine has drawn international attention to the impact of conflict on global agricultural production and food security. On June 6, 2023, the Kakhovka Dam was destroyed. The loss of this critical agricultural and energy infrastructure has left most farmers on the East bank of the Dnipro River without irrigation in the historically arid Kherson and Zaporizhzhia oblasts.Based on satellite analysis, pre-war irrigated areas between 2019 to 2021 were on average 187.94 ± 35.80 kHa, approximately 4% of all cropland and 8% of summer season cropland in Kherson and Zaporizhzhia oblasts. These numbers decreased in 2022 to 140.04 ± 17.55 kHa with the expanded occupation of these territories by Russia, seeing an additional decline in 2023 to 74.97 ± 17.84 kHa. Fields in 2023 also exhibited variability in irrigation, ranging interrupted irrigation owing to the destruction of the Kakhovka reservoir, to sufficiently irrigated. Irrigated areas diminished further in 2024 to 15.95± 9.97 kHa, a 90% decline from pre-war averages. Formerly irrigated fields before 2023 were either abandoned or converted to winter cropland by 2024–a pattern that continued into 2025 (total irrigated area 5.76± 0.87 kHa). These results of irrigation loss demonstrate a significant change in cropping patterns in response to the loss of irrigation as a result of conflict.

Accurate delineation of agricultural field boundaries from satellite imagery is essential for land management and crop monitoring, yet existing methods often produce incomplete boundaries, merge adjacent fields, and struggle to scale. We present the Delineate Anything Flow (DelAnyFlow) methodology, a resolution-agnostic approach for large-scale field boundary mapping. DelAnyFlow combines the DelAny instance segmentation model, based on a YOLOv11 backbone and trained on the large-scale Field Boundary Instance Segmentation-22M (FBIS 22M) dataset, with a structured post-processing, merging, and vectorization sequence to generate topologically consistent vector boundaries. FBIS 22M, the largest dataset of its kind, contains 672,909 multi-resolution image patches (0.25-10m) and 22.9million validated field instances. The DelAny model delivers state-of-the-art accuracy with over 100

One of the core applications of satellite-based classification maps is area estimation. Regardless of the algorithms used, maps will always contain errors stemming from imperfect input and training/calibration data, incomplete data coverage, and spectral and/or temporal confusion between land cover and land use classes. Because of omission and commission errors, the pixel-counting area estimator will be a biased estimator for area estimation. Therefore, the remote sensing research and application communities have developed a framework and recommended practices to address this problem. One such approach is a stratified random sampling design, in which classification maps could be used for stratification in the sampling design, and areas are estimated from the sample data, which represent reference data or reference class labels. As such, the quality of the map, i.e., producer's (PA) and user's accuracy (UA), will not affect the bias of the estimator, as the bias depends on the sampling design and the choice of estimator. However, map quality will impact the efficiency of stratification: a more accurate map will require a smaller sample size to reach the target variance of the estimate, or it will yield improved precision if the sample size is fixed. This study aims to provide a quantitative assessment of the impact of map accuracies on area estimation within the stratified random sampling design. The relative bias of the pixelcounting estimator is expressed using class-specific PA and UA, and shown to be (PA)(UA)- 1. Furthermore, for the case of binary classification, elements of the confusion matrix, as well as the sample size, variance of the area estimator, and relative efficiency of stratification (the ratio of the products of variance and sample size for the two sampling approaches) are expressed using PA and UA. Numerical simulations demonstrate how relative efficiency depends on area estimation objectives, target area proportion, and the map's performance metrics (PA and UA). Such dependence is nonlinear, and the impact of those parameters varies. For example, when the target class is minor or rare (i.e., its true proportion is <<0.5), the impact of PA outweighs that of UA. As the target area proportion increases, the impact of accuracies converges, and UA has a greater impact on efficiency than PA. There are multiple values in the PA/UA space, though constrained, to reach the same objectives, e.g., in terms of relative efficiency, sample size, and target variance. Overall, this study offers map producers a criterion that can be used to benchmark algorithm performance for map generation when area estimation is the primary objective of the classification maps.

Three years of sustained shelling, mining, and active combat have caused major cropland abandonment in Ukraine, particularly along frontlines. Existing estimates of abandoned areas vary up to fourfold, due to inconsistent definitions, baselines and biased estimators of area. Many studies classify fallow land – temporarily unused but managed – as abandoned. In contrast, abandoned lands (neither cultivated nor managed) are often contaminated by unexploded ordnance, mines, or chemicals, requiring clearance before recultivation. Disentangling fallow from abandoned cropland is therefore crucial for post-war recovery planning and for determining tax relief for farmers unable to access their fields. We applied a two-level stratified random sampling design and unbiased estimators of area to quantify the extent of cultivated, fallow, and abandoned cropland. Four regions with distinct conflict dynamics were delineated using the Armed Conflict Location and Event Dataset, and within each region stratified random samples were drawn from a Planet-based cultivation status map. Between 2021 and 2024, full Ukraine's cultivated area declined by 2.5 Mha (−8.5 %). By 2024, 7 % of cropland (2.213 ± 0.256 Mha) was abandoned, of which 1.121 ± 0.148 Mha to 1.671 ± 0.169 Mha may be permanently lost to cultivation, primarily along frontlines. Fallow areas increased nationally, especially in territories reclaimed from occupation. In 2024 alone, up to 0.472 ± 0.143 Mha were recultivated, by far exceeding official land clearance figures and suggesting widespread reliance on informal or self-organized demining. These results establish a replicable framework for monitoring land-use dynamics in conflict zones, supporting evidence-based recovery, landmine clearance prioritization, and agricultural policy planning in post-war Ukraine.

The issue of cloud coverage over agricultural areas impedes the use of optical Earth observation data for consistent in-season monitoring of cropland including area estimations. Sentinel-1 SAR has been proven to be a critical dataset in both complementing optical data and on its own in mapping crop-types and thus can overcome the issue of cloud coverage. However, there is still a lack of research on the efficacy of Sentinel-1 SAR for mapping winter cropland in major wheat producing countries such as Russia, and in particular at varying stages of the winter crop growing season. This gap highlights the need to provide research on how SAR data can address the challenge of cloud coverage that often obscures optical data, thereby improving timely and accurate cropland area estimations. In this study, we assessed the performance of mapping winter cropland at successive 15-day intervals, starting in the autumn planting period through to the end of the harvest period, using solely Sentinel-1 SAR VV (vertical-vertical) and VH (vertical-horizontal) bands in both ascending and descending orbits. Three models were utilized being a multi-layer perceptron model, a long short-term memory model, and a random forest model and their efficacy was assessed. Performance was assessed at each in-season time period in terms of user (UA) and producer accuracy (PA) in addition to area estimation comparisons with final year recorded winter cropland statistics at the sub-national level. Results indicate that in mid-November (planting year) (DOY -40), the UA was 0.68 and PA was 0.47; in early spring (DOY 100), the UA was 0.70 and PA was 0.78; and by early August, the UA reached 0.84 and PA 0.93. Notably, the LSTM model achieved even better results than the MLP model for estimating areas in the late autumn season (DOY -40) when comparing raw area differences (pixel count) from final reported statistics.

Military operations in Ukraine have led to significant changes in land use, especially in the southern regions, where irrigation plays a critical role in agriculture. The destruction of infrastructure and the occupation of territories have caused a significant reduction in agricultural production, affecting food security in the region and the world. Land use changes were analyzed based on the analysis of the time series of satellite data for 2016-2024. Using the developed reference mask of cropland, the area of abandoned land increased to 17.4% (5.1 million hectares) of the total area (29.1 million hectares) detected. In the occupied territories, 46.3% of agricultural land (2.7 million hectares out of 5.9 million hectares) was not cultivated in 2024. An analysis of historically irrigated lands in the Kherson region revealed changes in the crop structure, with the largest reduction in moisture-dependent crops: soybean areas decreased by 95.8%, and maize by 52.3%. The study's results quantitatively characterize the war's impact on the agricultural sector and its consequences for food security.

Grasslands are vital for global food security, making reliable monitoring of forage mass (FM) essential for sustainable pasture management. The availability and quality of FM are key factors in determining the profitability of pasture-based farms. This study presents a replicable methodology for estimating FM using multi-sensor satellite data and an agrometeorological modeling framework. Conducted at the Brazilian Agricultural Research Corporation Southeast Livestock Center (Embrapa Pecuaria Sudeste) in Sao Carlos, Brazil, the research integrates NASA's Harmonized Landsat and Sentinel-2 (HLS) imagery with climate data processed through the Simple Algorithm for Evapotranspiration Retrieving (SAFER) and Monteith's Light Use Efficiency (LUE) models. The SAFER model explained over 67 % of FM variability in three pasture-based livestock systems. A key factor in achieving accurate FM estimates was the differentiation between field green matter (GM) and total dry matter, as GM represents the most nutritious and consumable forage component. The model performed best in extensive systems, where minimal management intervention resulted in stable forage conditions. In integrated croplivestock systems, the accuracy remained high, though fertilization and crop residue decomposition influenced FM estimates. In intensive systems, model performance was slightly lower due to higher management variability. This study contributes to the development of automated, scalable FM assessment methods, enabling systematic pasture monitoring and data-driven grazing management. The SAFER model allowed simultaneous processing of satellite imagery and climate data, increasing the accuracy of FM estimations. Future research should explore the use of higher-resolution imagery (e.g., CBERS-4A, PlanetScope) to better capture within-field variability and consider increasing the frequency of field sampling frequency (from 32 days to 15 or even 7 days) to further improve FM estimation accuracy, particularly in intensive systems.

Spatial variability has been one of the major challenges for large-area crop monitoring and classification with remote sensing. Recent works on deep learning have introduced spatial transformation methods to automatically partition a heterogeneous region into multiple homogeneous sub-regions during the training process. However, the framework is only designed for deep learning and is not available for other models, e.g., decision tree and random forest, which are frequently the models of choice in many crop mapping products. This paper develops a geo-aware random forest (Geo-RF) model to enable new capabilities to automatically recognize spatial variability during training, partition the space, and learn local models. Specifically, Geo-RF can capture spatial partitions with flexible shapes via an efficient bi-partitioning optimization algorithm. GeoRF also automatically determines the number of partitions needed in a hierarchical manner via statistical tests and builds local RF models along the partitioning process to explicitly address spatial variability and improve classification quality. We used both synthetic and real-world data to evaluate the effectiveness of Geo-RF. First, through the controlled synthetic experiment, Geo-RF demonstrated the ability to capture the artificially-inserted true partition where a different relationship between the inputs and outputs is used. Second, we showed the improvements from Geo-RF using crop classification for five major crops over the contiguous US. The results demonstrated that Geo-RF is able to significantly improve classification performance in sub-regions that are otherwise compromised in a single RF model. For example, the partition around downstream Mississippi for soybean classification led to major improvements for about 0.10-0.25 in F1 scores in the area, and the score increased from 0.57 to 0.82 at certain locations. Similarly, for rice classification, the partition in Arkansas led to F1 scores increasing from 0.59 to 0.88 in local areas. In addition, we evaluated the models under different parameter settings, and the results showed that Geo-RF led to improvements over RF in the vast majority of scenarios (e.g., varying model complexity and training sizes). Computationally, Geo-RF took about one to three times more training time while its execution time during testing was similar to that of RF. Overall, Geo-RF showed the ability to automatically address spatial variability via partitioning optimization, which is an important skill for improving crop classification over heterogeneous geographic areas at large scale. Future research can explore the use of Geo-RF for other geographic regions and applications, interpretable methods to understand the data-driven partitioning, and new designs to further enhance the computational efficiency.

This study aims to assess the impact of the Kakhovka Dam destruction in Ukraine that occurred on 6 June 2023, on cropland irrigation using satellite remote sensing data. The main goal of this study is to assess flooded areas and the impact on irrigated area before and after Kakhovka Dam destruction. In particular, we analyzed flooded areas in 2023, and the changes in irrigated areas before and after the dam destruction (in 2019 and 2024) were also assessed. Maps of water bodies were generated before and after the flood using Sentinel-1, Sentinel-2, and Landsat-9 images. The random forest classifier was used for flooded area mapping, while the multilayer perceptron and the U-shaped network classifier were used for irrigated land identification. The main findings are as follows. 1) As of 9 June 2023, the total area of flooding under the Kakhovka Dam was 47.33 thousand hectares (th. ha) (473 km2), affecting 1.67 th. ha of cropland, 0.97 th. ha of forests, 12.3 th. ha of grasslands, 1.85 th. ha of settlements, and 29.4 th. ha of wetlands. 2) The analysis of irrigated area shows a decrease in irrigated cropland—from 351 th. ha in 2019 to 38 th. ha. in 2024. 3) The classification accuracy for 2019 irrigation mapping achieved 90.4% overall accuracy with F1-scores of 90.4% for both irrigated and nonirrigated classes based on ground truth data. 4) The complete disappearance of water in irrigation canals was documented, indicating the systematic destruction of agricultural infrastructure with far-reaching consequences for regional food security. The flood also affected areas along the Ingulets River, which led to the flooding of agricultural land located near the river banks and affected water quality. The disappearance of water in canals used to irrigate cropland is also analyzed, indicating the disruption of irrigation systems and possible far-reaching consequences for agriculture. Thus, this study shows the utility of satellite remote sensing and machine learning approaches for rapid monitoring and quantification of flood-related natural disaster impacts and the analysis of irrigated areas in conflict-affected regions.

Crop yield forecasting is an essential component of crop production assessment, impacting people at the global scale down to the level of individual farms. Until now, yield forecasting has predominantly relied on optical data, particularly the maximum value of vegetation indexes. However, this approach only presents a short forecasting window, and it is essential to obtain yield estimates as early as possible in the growing season and then further improve forecasting even after the vegetation index has reached its peak. So far, optical satellite data at high-temporal resolution (1–3 days) has been actively used for real time crop yield monitoring, whereas fewer operational models make a use of synthetic aperture radar (SAR). In this study, we explore whether SAR data can capture distinct aspects of crop dynamics, providing new insights for yield estimation depending on the crop's phenological stage. We assess the efficiency of dual- (Sentinel-1) and quad-polarimetric (UAVSAR, RADARSAT-2) data to explain inter-field crop yield variability for corn, soybean, and rice over a test area in Arkansas, US (258 fields, 2019). We used optical imagery acquired by Planet/Dove-Classic, Sentinel-2, and Landsat 8, to establish a baseline performance of satellite-based indicators to explain yield variability and assess dual- and quad-polarimetric SAR data for crop yield assessment. In terms of polarimetric indexes, the results showed that in general the results for rice were mostly stable and better than the other crops (R2adj ∼ 0.4 on average). The best results were obtained for the Sentinel-1 VHasc with R2adj = 0.47 and RADARSAT-2 phase difference with R2adj = 0.45. The results for corn performed the least with an R2adj <0.35 for all the indexes. The results for soybeans were more variable and were highly correlated with certain indicators such as RADARSAT-2 HV, RADARSAT-2 Volume, and RADARSAT-2 Pauli HV with R2adj>0.4. We also investigated the day of year (DOY) with the maximum correlation between optical and SAR-derived features and the final yields for corn, soybean, and rice. The maximum correlation for optical features occurs over a short time between DOY 155 (June 4) and 185 (July 5) for corn and rice, and DOY 190 (July 9) and DOY 211 (July 30) for soybean, with these results being consistent across various optical-based sensors. On the contrary, the maximum correlation for SAR-derived features varied significantly and was between DOY 120 (April 30) to DOY 225 (August 13). A study of the time series parameters cross-correlation showed that the optical parameters were highly correlated, but the SAR parameters showed strong temporal decorrelation. We conducted a comparison between C-band and L-band to assess their sensitivity at each stage of growth. In this experiment, we determined that for low vegetation, the C band will be more useful at the beginning of the growth cycle, while the L band provides more information in later stages of growth. Using a random forest regression model combining SAR parameters with common difference vegetation index (DVI), we improved the error by 50% in comparison to the error using the (DVI) for corn, soybean, and rice.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI