Above-ground biomass density (AGBD) quantification is crucial for understanding carbon dynamics, climate change, and sustainable forest management. This study integrates Global Ecosystem Dynamics Investigation (GEDI) satellite data with multi-spectral, Synthetic Aperture Radar (SAR) based earth observations and soil data for continuous estimation of forest AGBD. Study was focused on the Indian forest. GEDI’s AGBD data from 10,000 points serves as the dependent variable and independent variables are derived from Sentinel-1, Sentinel-2, Digital Elevation Model (DEM), SoilGrids, and Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS). We evaluated Support Vector Regression, Random Forest Regression, and Light Gradient Boosting Machine algorithms for various feature-set scenarios. Hyperparameter tuning employed grid-search-based cross-validation. Results shows that, LightGBM performed well, being computationally efficient and delivering lower RMSE. For the selected LightGBM model, with Sentinel-1, Sentinel-2, DEM, and forest attributes, an RMSE of 68.02 Mg/ha and R2of 0.57 were achieved. Model-generated AGB maps were compared with openly available National Remote Sensing Centre AGBD data at 100m and existing forest AGBD work. Comparison between model predicted AGBD and literature based maps and studies shows that, our model was able to capture the AGBD variations across multiple forest sub-regions from India.

Mapping sugarcane areas is vital for applications like crop monitoring, yield estimation, environmental monitoring, and land use planning. Traditional supervised learning is hindered by the costly and time-intensive collection of ground truth data, whereas, unsupervised methods encounter performance challenges due to parameter initialization. In this study, we propose a hybrid approach that integrates unsupervised learning with domain knowledge of temporal sugarcane crop descriptors, creating reference data for subsequent supervised learning. The main objective of the study was to map sugarcane areas in the absence of ground reference data by combining unsupervised and supervised learning using Sentinel-1 and 2 observations. The study was carried out in two provinces in central Thailand during the sugarcane season of 2021–2022. X means clustering was applied to the raster stack of temporal NDVI and radar backscatter (in VH polarisation). Fifty polygons were then digitized from each cluster and categorized as sugarcane, cassava, field crops, and non-agriculture using the temporal descriptors for each polygon. Polygons not meeting the temporal descriptors based criteria were removed. Further, RF based supervised classification was carried out using bands of Sentinel-1 and 2, and NDVI as the features and labled polygons as the reference. The samples were divided into training (80% data) and testing (20% data). We tuned the RF classifier for a number of trees ranging from 50 to 500. The model with 350 trees performed better on testing data, with an overall accuracy of 87.75% and a Kappa of 0.873. Slight intermixing between sugarcane and cassava was observed mainly due to the planting/sowing window (March–May) and crop duration in the case of the Ratoon sugarcane crop. Further, we implemented the same model in an adjacent province to evaluate the performance of the proposed approach. The overall accuracy of 75.86% was obtained, and precision and recall for sugarcane were 0.79 and 0.76, respectively. This shows that strong temporal descriptors derived from an unsupervised approach can be combined with supervised learning for mapping sugarcane areas in the absence of ground labels.

Vegetation cover plays a crucial role in enriching the soil carbon content. The sequestered CO2gets released due to exposure of soil to the atmosphere by the process of volatilization. Therefore, there is a need to monitor sustainable farm management practises like cover cropping. Remote Sensing coupled with artificial intelligence helps in non-invasive monitoring of vegetation cover. The main objective of this study is to detect the presence of cover crop using the time-series of remote sensing observations. The study was carried out on selected fields from the Europe region. A total of 60 fields from four countries, namely, France, Germany, Poland, and Spain during 2019–2021 were selected. In this study we proposed a two-step approach for cover crop detection. The first step involves separating vegetation period from fallow/bare soil and snow cover. The second step has mainly focused on the vegetation period to initially separate main crop period and subsequent detection of cover crop for the remainder of the vegetation period. Combination of index based thresholding and phenology indicators was used for cover crop detection. The proposed approach was validated using the ground reference data on presence or absence of a cover crop. Results showed an overall accuracy of 91.7% with an F1 score of 91.2%. Moreover, cover crop detection rate was found to be 92.9%. One field was misclassified as cover crop, whereas it had a dense cover of weeds. This was mainly due to higher peak NDVI value of dense weeds than NDVI threshold.

Remote sensing coupled with machine learning is useful for non-invasive monitoring and prediction of Soil Organic Carbon (SOC) and carbon stock. The use of Sentinel-1 and 2 datasets was attempted in this work for SOC, Bulk Density (BD), and carbon stock estimation. The key objective of this study is to estimate Soil Organic Carbon and Bulk Density for carbon stock assessment in croplands using multi-spectral and Synthetic Aperture Radar (SAR) satellite observations. The fields from Colorado and Nebraska states are selected, and assessment is performed after the harvest of the cropping season in 2020. Soil samples were collected during 10-30 Oct 2020. Regression models were developed using bands and indices derived from Sentinel-1 and 2 datasets as independent variables, however SOC or BD were used individually as dependent variables. Models were developed by considering data from individual states and then combining all the data using Random Forest Regression (RFR) and Support Vector Regression (SVR). Results for the SOC estimation showed that RFR performed better than SVR with individual state data. Lowest RMSE achieved using the RFR were 0.114 and 0.159 for fields from Nebraska and Colorado states. However, SVR outperformed over RFR in the case of BD estimation using the combined data from both states (RMSE = 0.24). Further comparison between estimated carbon stock and actual carbon stock at field level shows the good agreement over fields from Colorado compared to Nebraska.

Agriculture tillage is a fundamental practice in farming that involves preparing the soil for planting crops. It has been an essential technique used by farmers for centuries to improve soil conditions, increase crop yields, and enhance overall agricultural productivity. While tillage information can be acquired through manual field data collection, implementing this approach consistently and systematically over a wide area poses considerable challenges. Instead, remote sensing methods offer a viable option to comprehensively, promptly, and affordably investigate tillage activities. Hence, there is significant value in embracing a remote sensing approach to consistently and methodically monitor tillage practices across various fields. The objective of this research was to determine different types of tillage surfaces by analyzing the radar backscatter response received from the ground. The study used data from the Sentinel-1 satellite, specifically the Interferometric Wide-swath (IW) Ground Range Detected (GRD) dataset, which provided radar measurements in both VV and VH polarizations. To monitor tillage, we utilized supervised classification methods, namely decision tree (DT), random forest (RF), and support vector machine (SVM). Among these classifiers, the RF has the highest test accuracy of 0.86. The obtained results were validated using the ground observation data and found encouraging.

In this study, we used time-series radar backscatter response in VV and VH polarizations (i.e. σ°VVand σ°VH) and ground-based wheat height data to establish the multivariate regression models for wheat height estimate. For generating time-series σ°VVand σ°VH, the C-band Sentinel-1 satellite data were used. The ground height observation was obtained from the farm having Durum wheat variety HI 8759 (also known as Pusa Tejas). We experimented with five different multivariate regression models (MVRM) using combinations of independent variables i.e. σ°VV, σ°VH, σ°VV/σ°VHratio, and days after sowing (DAS). Based on the combination of variables, the five models can be depicted as MVRM1 (σ°VV, σ°VH, and DAS), MVRM2 (σ°VHand DAS), MVRM3 (σ°VVand DAS), MVRM4 (σ°VVand σ°VH) and MVRM5 (σ°VV/σ°VHratio and DAS). Based on statistical observation, it is found that although the model R2for MVRM1 and MVRM2 was about 0.90, the condition number is very high in both cases, which causes multicollinearity. This leads to model overfitting. On the other hand, model R2obtained for MVRM3, MVRM4, and MVRM5 are 0.94, 0.65, and 0.93 respectively and all the models are free from multicollinearity as the condition number is low. Another observation shows that DAS is an important variable in the regression model along with the backscatter response. Overall, based on the statistical parameters, MVRM3 and MVRM5 were found suitable for wheat height estimation. This particular model was validated on an independent wheat farm and found encouraging results.

Crop water stress is one of the major factors limiting crop productivity. Maize (Corn) crop is very sensitive to water stress. Efficient monitoring and detection of water stress is crucial for precision irrigation and sustainable agriculture. The main objective of this study is to detect water stress during grain-fill stage in maize crops using hyperspectral (HS) observations. Maize hybrid field trial plots were selected from Hyderabad region, Telangana, India. HS images of the study area were collected using a hexacopter drone mounted with line scanning HS camera having spectral range of 450–950 nanometer (nm). The line scanning HS drone observations were processed for georeferencing, mosaicing, and ortho-rectification. The study area was divided into two sub-plots, viz. control (i.e., no stress) and water stress during grain-fill stage of the crop. Each plot had 66 sub-plots, each representing a unique variety. Visual spectral analysis was performed to identify the differences and spectral regions suitable for water stress detection. Analysis revealed that the visible region around 680 nm and the Red-Edge (RE) - Near-InfraRed (NIR) region were providing the differences between the no-stress and water stressed plots. The optimal subset of wavelength regions were identified using machine learning (ML) based feature selection. Random Forest (RF) and Support Vector Machine (SVM) classifiers were used to assess the feasibility of the chosen bands. SVM with top 10 bands in the range of 670–780 nm found most effective for detecting grain-fill water stress in maize crop varieties.

The optical satellite observations are affected by cloud cover. Vegetation Optical Depth (VOD) has the potential to provide insights into plant water and vegetation structure. The main objective of this study is to spatially downscale VOD using a Moderate Resolution Imaging Spectrometer (MODIS) and Shuttle Radar Topographic Mission (SRTM) observations. The study considered India geography and post-monsoon cropping season (locally called Rabi season). Data for three different Rabi crop seasons, i.e., 2017–18, 2018–19, and 2019–20, was used in the analysis. The VOD estimates at 25 km scale derived from the Advanced Microwave Scanning Radiometer for EOS (AMSR-E) and Advanced Microwave Scanning Radiometer 2 (AMSR2) were used for spatial down-scaling. Three day VOD composites were created to cover the study area. MODIS products such as Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), albedo (black and white sky), and Land Surface Temperature (LST) were similarly acquired at a 3-day interval by constructing a 3-day composite. In addition, SRTM digital elevation model with a spatial resolution of 90 m was used in this study. We carried out regression modeling where VOD was used as a dependent variable, with NDVI, NDWI, Albedo (black and white sky), LST, and elevation as independent variables. We compared three regression algorithms, viz., Linear Regression (LR), Random Forest Regression (RFR), and Support Vector Regression (SVR), using the R-square (R2) as the assessment metric. A comparison between the various regression techniques showed that the SVR outperformed monthly and seasonal models. Further more, a comparison of monthly and seasonal models revealed that the model generated with January data performed best, with an R2of 0.85, followed by R2of 0.82, 0.80, and 0.78 for March, December, and February, respectively. The R2for the seasonal model was 0.83. Finally, for the wheat crop time series of down-scaled VOD and Sentinel 2 based NDVI was compared to gain insights on seasonal variations in VOD. We found that down-scaled VOD and NDVI have a significant agreement.

Nitrogen (N) is one of the key nutrient element needed for optimum crop growth and production. Deficiency of N leads to a decrease in crop production and excess results in poor root growth and leaching into groundwater thereby causing environmental issues. Hence the optimum application of N is needed which is possible by exactly estimating the available quantities of N in the plant. In this study, an attempt has been made to estimate N in Soybean leaves using the hyperspectral and simulated Sentinel-2 observations. Spectral observations of fifteen soybean leaf samples were collected using the EKO MS-720 Spectroradiometer. The instrument operates in the spectral range of 350–1050 nm. and collects data in contiguous 213 bands. Support Vector Regression-based models were evaluated using three feature selection methods, 1) individual hyperspectral bands, 2) Normalized band ratio's and 3) simulated Sentinel-2 bands and indices. Model performance was evaluated using R2. Analysis carried out using the individual hyperspectral bands showed that bands from the red and red-edge region are performing best with R2between 0.872 and 0.876. However, NBR's estimated from band combinations in the red-edge region are performing best with R2between 0.938 - 0.956. Further, we identified a subset of wavelengths to simulate Sentinel-2 spectral bands, results showed that red-edge and narrow NIR bands provide the highest R2between 0.878 and 0.893. We observed that indices such as Canopy Chlorophyll Content Index (CCCI) and Chlorophyll Index Red Edge (CIRE) are performing better for N estimation with R2of 0.946, 0.923, respectively. Based on the observations we can conclude that red, red-edge and narrow NIR region is useful for Soybean N estimation.

In the present study, InSAR coherence products extracted from Sentinel-1 were used for the quantification of the agricultural region affected by normal to large excess rainfall events that occurred in the districts of Gujrat, India in July and August 2020. In this analysis, the coherence values retrieved during co-event (i.e. during rainfall) were comparatively found lower than the pre-event (i.e. before rainfall) for all the sub-divisions of the districts. Here, the pre-event and co-event coherence histograms were drawn and their point of intersection was used to determine an optimal threshold value for coherence below which agriculture area was considered as affected due to rainfall. More than 80 % of the total study area was found affected due to rainfall events. The obtained outcomes were crossexamined with the Landsat-8 images obtained for the study duration and the results were found encouraging.

The main objective of this paper is mapping of Soybean and Cotton crop area using the integration of Sentinel-1 and 2 observations and Deep Neural Networks (DNN). The present research also attempted the identification of early and late sowing of Soybean and Cotton using time-series of Normalized Difference Vegetation Index (NDVI) and VH backscatter. The study was carried out in Wardha district of Maharashtra, India during Kharif 2019. The Sentinel-1 observations available during 15 Jun. to 30 Nov. 2019 and Sentinel-2 maximum NDVI and Normalized Difference Water Index (NDWI) composites during Aug. to Sept. and Oct. to Nov. 2019 were used for Cotton and Soybean area mapping. We evaluated the performance of tuned Random Forest (RF) and DNN for classification of Soybean and Cotton. Results showed that DNN performs better with an overall accuracy of 89.15% and the F-score of 0.856. Further, identification of early and late sown Soybean and Cotton was performed using short time-series of VH backscatter and NDVI. Local minima along with subsequent increase/decrease based approach was developed for the separation of early and late sown crop. Visual comparison with geo-tagged crop growth stage images shows good agreement between estimated and actual sowing time window.

To monitor changes in the landscapes at local to regional scales, a large amount of multi-temporal remote sensing products are required for the analysis. Due to the frequent need for remote sensing-based results, research as well as commercial organizations required automated, faster and more efficient methods for downloading and subsequently, processing the enormous amount of such remote sensing datasets. In this study, we have designed and developed geospatial data processing frameworks enriched with automated data search, download, and processing of Sentinel-1,-2 data. The analysis-ready outputs i.e. Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI) and radar backscatter (σ°) etc. are essential in operational remote sensing use cases such as crop area, crop health, crop loss mapping. The overall framework for Sentinel-1,-2 has been successfully implemented in the Python programming environment.