This article describes a multi-sensor dataset collected during the TIRAMISU (Thermal InfraRed Anisotropy Measurements in India and Southern eUrope) campaign at the Nawagam research site in Gujarat, India, during the 2023 monsoon season. The objective was to acquire continuous ground-based optical and thermal measurements over a homogeneous rice canopy across different crop growth stages.The dataset integrates several complementary components. Thermal data were acquired with an Optris longwave infrared camera (8–14 µm) at high temporal resolution, capturing canopy temperature dynamics throughout the diurnal cycle. Optical data were obtained with a Micasense RedEdge-M multispectral sensor, providing imagery in Blue, Green, Red, RedEdge, and Near-Infrared bands with radiometric corrections. An Apogee radiometer supplied reference radiometric temperature. Meteorological measurements included air temperature, humidity, wind speed and direction, and net radiation. Ancillary field measurements comprised Leaf Area Index (LAI), plant height, emissivity sampling, hyperspectral observations, and crop stage information.The datasets are provided with metadata and processing workflows, including calibration procedures for optical reflectance and thermal radiance. Together, these components form a comprehensive record of canopy–atmosphere interactions over a homogeneous rice field. The datasets can support research on optical and thermal directional anisotropy, canopy radiative transfer, emissivity characterization, and crop biophysical parameter estimation. In addition, they are relevant for applications in vegetation monitoring, agricultural water stress assessment, and surface energy balance studies. By combining optical, thermal, and meteorological observations, the resource is suited for multidisciplinary investigations in remote sensing, agronomy, and environmental sciences.

Abstract. Thermal infrared (TIR) satellite remote sensing is essential for monitoring land surface temperature (LST) and surface energy fluxes, supporting applications in hydrology, agriculture, climate, and urban climate. Existing TIR missions—such as LANDSAT, ASTER, ECOSTRESS, and Sentinel-3 SLSTR—offer complementary capabilities but remain constrained by trade-offs between spatial resolution, revisit frequency, and radiometric accuracy, limiting their ability to capture rapidly evolving surface processes at field to regional scales. The TRISHNA (Thermal Infrared Satellite for High-Resolution Natural Resource Assessment) mission, expected in 2027, addresses this gap by providing 60 m TIR imagery over a ~1000 km swath with sub-weekly revisit, enabling systematic monitoring of surface energy processes in natural and managed ecosystems. Its integrated design —including orbit configuration, spectral channels (VNIR, SWIR, TIR), viewing geometry, and calibration strategy— supports accurate retrievals of evapotranspiration, vegetation water stress, and surface temperature dynamics. Synergies with upcoming missions such as ESA’s Land Surface Temperature Mission (LSTM) and NASA’s EAGLE mission (Explorer for Artemis Geology Lunar and Earth) will enhance temporal coverage, cross-calibration, and long-term data continuity.

This article describes a multi-sensor dataset collected during the TIRAMISU (Thermal InfraRed Anisotropy Measurements in India and Southern eUrope) campaign at the Nawagam research site in Gujarat, India, during the 2023 monsoon season. The objective was to acquire continuous ground-based optical and thermal measurements over a homogeneous rice canopy across different crop growth stages. The dataset integrates several complementary components. Thermal data were acquired with an Optris longwave infrared camera (8-14 µm) at high temporal resolution, capturing canopy temperature dynamics throughout the diurnal cycle. Optical data were obtained with a Micasense RedEdge-M multispectral sensor, providing imagery in Blue, Green, Red, RedEdge, and Near-Infrared bands with radiometric corrections. An Apogee radiometer supplied reference radiometric temperature. Meteorological measurements included air temperature, humidity, wind speed and direction, and net radiation. Ancillary field measurements comprised Leaf Area Index (LAI), plant height, emissivity sampling, hyperspectral observations, and crop stage information. The datasets are provided with metadata and processing workflows, including calibration procedures for optical reflectance and thermal radiance. Together, these components form a comprehensive record of canopy-atmosphere interactions over a homogeneous rice field. The datasets can support research on optical and thermal directional anisotropy, canopy radiative transfer, emissivity characterization, and crop biophysical parameter estimation. In addition, they are relevant for applications in vegetation monitoring, agricultural water stress assessment, and surface energy balance studies. By combining optical, thermal, and meteorological observations, the resource is suited for multidisciplinary investigations in remote sensing, agronomy, and environmental sciences.

Accurate multinational soil prediction remains a fundamental challenge due to pedogenic heterogeneity, management-induced variability, and sample distribution imbalance that constrain the transferability of conventional global calibration models. To address these limitations, this study introduces the Mixture-of-Algorithmic-Experts (MoAE) framework, an architectural approach that decouples global coverage from local specialization by combining algorithmically specialised expert ensembles with a learned routing mechanism for conditional computation. The framework was evaluated using proximal sensor data (visible near infrared spectroscopy and portable X-ray fluorescence spectrometry) from four pedoclimatically distinct countries (Brazil, France, India, and the USA), MoAE was evaluated across 17 physicochemical soil properties encompassing carbon fractions, particle-size distribution, cation exchange capacity, electrical conductivity, and macro- and micronutrients. In the primary hold-out validation, the framework achieved high validation performance for structurally stable properties, including total carbon (Coefficient of Correlation (R2) = 0.99), total nitrogen (R2 = 0.99), and texture fractions (R2 = 0.95–0.98), with negligible systematic bias, substantially exceeding typical global spectroscopic calibration benchmarks reported in the literature. In contrast, lower performance for management-sensitive nutrients such as available S and P reflected signal limitations inherent to the sensing-feature space rather than architectural failure. Shapley Additive Explanations-based interpretability confirmed that predictions were driven by geochemically meaningful elemental proxies and diagnostically relevant spectral wavelengths, while routing-weight distributions provided an internal confidence indicator for deployment-aware decision support. Unlike monolithic calibration strategies, MoAE enables context-aware computation, dynamically weighting expert contributions according to sample-specific feature representations. By integrating conditional modelling, interpretability, and multi-sensor complementarity within a unified framework, MoAE offers a scalable and responsible foundation for multinational digital soil mapping and next-generation operational soil intelligence systems.

Accurate and scalable estimation of soil organic carbon (SOC) is essential for sustainable land management and climate mitigation, yet conventional laboratory-based methods remain costly and impractical for large-area monitoring. This study presents a novel two-stage framework that integrates low-cost proximal red, green, and blue (RGB) soil imaging, deep learning, and machine learning-based digital soil mapping (DSM) to enable regional SOC prediction at 100 m resolution. In the first stage, SOC was estimated from RGB images of 405 surface soil samples collected across multiple agro-climatic zones of West Bengal, India, using a random forest model based on color features and a fine-tuned Visual Geometry Group 16-layer (VGG16) convolutional neural network. The deep learning model achieved superior performance, with a 37.5 % higher validation R² and 13.1 % lower mean absolute error (MAE) compared to the color feature-based model. In the second stage, RGB-derived SOC predictions were integrated with multi-source environmental covariates within a DSM framework, yielding regional SOC maps that slightly outperformed a benchmark wet chemistry-based DSM with 8.8 % higher test R² and 32.1 % lower MAE. The maps revealed strong agro-climatic control on SOC distribution, with higher SOC in the northern Terai Zone (1.1–2.0 %) and lower SOC in the Red and Laterite Zone (<0.8 %), although both models showed limited sensitivity for low SOC (<0.5 %) soils. A benefit–cost analysis based on the Analytic Hierarchy Process demonstrated that the proposed RGB–DSM framework is more than twice as efficient as conventional wet chemistry-based mapping in terms of cost, time, and environmental safety. The study establishes a scalable and operationally viable pathway for SOC monitoring using consumer-grade imaging and deep learning, with strong potential for deployment in resource-limited regions and digital measurement, reporting, and verification systems.

Desert Locust (DL) infestations pose a significant threat to food security in arid and semi-arid regions, particularly in East Africa, Central Asia, and the Indian subcontinent. In 2020, during the COVID-19 pandemic, India witnessed an unprecedented upsurge of DL activity during the summer (zaid) season (April-June), severely impacting Rajasthan, Gujarat, and neighbouring states. This study investigates the environmental drivers of the DL outbreak and assesses crop damage using geospatial datasets, reanalysis products, and numerical weather models. Fifteen grid cells (100 km x 100 km) along the DL-prone corridor from East Africa to India were analyzed for environmental suitability, with seasonal Spearman correlation analysis applied to identify significant factors influencing locust activity. In winter, locust activity was significantly positively correlated with rainfall (rho = 0.47, p = 0.021), dew point temperature (rho = 0.76, p = 0.01), and soil moisture (rho = 0.50, p = 0.05), highlighting the importance of moisture and temperature conditions in facilitating locust presence. In spring, significant positive correlations were observed with air temperature (rho = 0.56, p = 0.027), soil temperature 1 (rho = 0.65, p = 0.01), and a very strong correlation with soil temperature 2 (rho = 0.73, p = 0.002). These findings showed the crucial role of temperature and moisture during the winter and spring seasons as key drivers of locust behaviour. The Linear Discriminant Analysis (LDA) model shows potential in locust presence prediction, though challenges remain due to data limitations. Crop damage was quantified using Normalized Difference Vegetative Index (NDVI), showing severe vegetation loss in affected areas (NDVI <0.3) and degradation due to locust feeding. The study further integrates weather forecast wind patterns, MODIS Leaf Area Index (LAI), and soil moisture from SMAP to track locust migration. Wind patterns, particularly westerly and south-westerly winds, guided the locusts' entry into western India. Despite moderate LAI values, the vegetation cover in central and western India provided sufficient sustenance for the locusts. Soil moisture from SMAP consistently supported locust dispersal across northern Rajasthan, central India, and parts of Uttar Pradesh. The integration of these environmental factors offers a comprehensive understanding of DL behaviour, enhancing early warning and control efforts.

Soil organic carbon (SOC) is a key indicator of soil health, yet conventional laboratory assays are labor-intensive and costly. This study investigates a rapid and low-cost alternative by using a handheld Nix Spectro 2 Color Sensor, which captured high-resolution color data from air-dried soil samples. These color parameters were used to predict SOC with four data-driven prediction engines: Random Forest (RF), Gradient Boosting Regression (GBR), Extreme Gradient Boosting (XGBoost), and an Artificial Neural Network (ANN) and further strengthened them with synthetic data augmentation techniques. A total of 641 surface soil samples collected from six districts in West Bengal, India, were divided into 70% calibration and 30% validation subsets. Synthetic samples were produced using a combination of generative artificial intelligence (AI) techniques [generative adversarial networks (GANs) and Gaussian mixture models (GMM)] and non-parametric/statistical data augmentation methods [k-nearest neighbors (KNN) and bootstrapping] to fill critical gaps in the SOC range (3-14%). Among the baseline models using raw Nix color data, RF achieved the best validation accuracy (R² = 0.71, RMSE = 0.93%). After augmenting the calibration set with 44 GMM-generated samples (3-7% SOC), RF performance rose to R² = 0.77 and RMSE = 0.84%, while bias dropped and coverage across the SOC distribution improved markedly. The incorporation of synthetic data mitigated model bias and enhanced predictive accuracy despite Levene's test revealing significant variance differences between calibration and validation datasets. The enhanced generalization of the model was attributed to better coverage of the SOC distribution, reducing underrepresented gaps in the dataset. The study highlighted the potential of AI-driven soil monitoring techniques in precision agriculture, demonstrating that integrating the Nix color sensor with synthetic data augmentation, provides a rapid and cost-effective solution for on-site soil assessments. Future research should expand these methodologies to multi-parameter soil assessments, digital soil mapping, and broader applications in sustainable soil management and climate change mitigation.

Soil organic carbon (SOC) plays a key role in soil health and ecosystem services. This study introduces Deep Carbon, a modelling framework that integrates static and time-series environmental covariates for high-resolution SOC prediction at the field scale. Time-series data were encoded using a stacked long short-term memory (LSTM) neural network to extract temporal patterns of dynamic features. These encoded time-series representations were combined with static covariates and used as inputs to train machine learning models at multiple spatial resolutions (5 km to 10 m). Individual predictions at each scale were then fused using a partial least squares regression (PLSR) model to generate SOC maps at 10 m resolution. The best accuracy was observed at 5 km scale (R2 = 0.75; RMSE = 0.30% in log scale), while the fused 10 m prediction yielded a testing R2 of 0.58 and RMSE of 0.44%. Fusion modelling identified 30 and 250 m resolutions as the most influential predictors. The approach successfully captured both high- and low-frequency SOC variations and demonstrated good transferability when tested on new observations from 2022. This multi-scale feature-time fusion approach uses legacy ground samples and satellite data to enable scalable and accurate digital SOC mapping.

Efficient discrimination of diverse kharif crops, remains crucial for crop monitoring and production forecasting, and plays a pivotal role in decision -making for food security in India. This study aims to harness temporal backscatter data from EOS-04 C -band synthetic aperture data (SAR) payload to achieve precise discrimination among six short -duration (cereal, oilseeds, fibre) and long -duration (fibre, pulses) kharif crops. The study integrates limited ground -truth polygons and a Random Forest machine learning approach for analysing EOS-04 time -series data. The classification accuracies were found to be higher than 75% across all kharif crops, with cereals exhibiting the highest accuracy, succeeded by fibre, oilseed and pulse crops. A key focus lies in identifying optimal polarization combinations for effective discrimination among diverse kharif crop types. The study reveals that the synergistic utilization of dual polarizations outperforms individual co- or cross -polarizations, notably benefiting discrimination of cotton, soybean and groundnut crops. Horizontal-vertical polarizations are found to be most effective for achieving peak accuracies in rice and red gram crops. Furthermore, the analysis indicates a promising potential for early crop assessment, presenting an opportunity to furnish precise crop estimates at least one and a half months before the harvest.

Soil fertility, specifically phosphorus (P) availability, is critical for agricultural productivity and environmental health. Traditional methods for measuring soil available P are typically lab-based and expensive. This study explored the potential of a smartphone-integrated imaging device combined with a digital soil mapping (DSM) approach to estimate and spatially map soil available P in six districts of West Bengal, India. A total of 482 surface soil samples were collected and analyzed using both conventional spectrophotometry (UV) and the developed smartphone-based method (NP). The results showed a strong correlation (R² = 0.94) between the two methods, with no significant difference in P estimation, validating the device's field applicability. A DSM model was developed using environmental covariates and random forest model to predict soil P distribution. Additionally, synthetic data were generated using three generative models (Triplet-based Variational Autoencoder, Conditional Tabular Generative Adversarial Network, and Gaussian Copula) to enhance prediction accuracy, particularly for high P values. The Gaussian Copula model, combined with real data, provided the highest test accuracy (R² = 0.69, RMSE = 21.15 kg ha−1 for UV and R² = 0.73, RMSE = 12.91 kg ha−1 for NP). Spatial maps revealed high P availability in alluvial soils of Nadia and East Medinipur, and low P in red and lateritic soils of Birbhum and Jhargram, reflecting the influence of soil type and climatic conditions. The smartphone-based device, coupled with DSM, offers a cost-effective, accurate, and practical tool for soil P assessment and mapping. This technology can significantly aid farmers in resource-constrained regions by providing precise nutrient management recommendations, enhancing sustainable agricultural practices, and mitigating environmental impacts. Future work will focus on further validation with diverse soil types and continuous improvement of the DSM models to address dynamic soil nutrient variability.

This study investigated the use of portable X-ray fluorescence (PXRF) spectrometry and soil image analysis for rapid soil fertility assessment, with a focus on key indicators such as available boron (B), organic carbon (OC), available manganese (Mn), available sulfur (S), and the sulfur availability index (SAI). A total of 1,133 soil samples from diverse agro-climatic zones in Eastern India were analyzed. The research integrated color and texture features from microscopic soil images, PXRF data, and auxiliary soil variables (AVs) using a Random Forest model. Results showed that combining image features (IFs) with AVs significantly improved prediction accuracy for available B (R2 = 0.80) and OC (R2 = 0.88). A data fusion approach, incorporating IFs, AVs, and PXRF data, further enhanced predictions for available Mn and SAI, with R2 values of 0.72 and 0.70, respectively. The study highlights the potential of integrating these technologies to offer rapid, cost-effective soil testing methods, paving the way for more advanced predictive models and a deeper understanding of soil fertility. Future work should explore the application of deep learning models on a larger dataset, incorporating soils from a wider range of agro-climatic zones under field conditions.

The NASA-ISRO Synthetic Aperture Radar (NISAR) is an Indo-US collaborative mission planned for launch in early 2025. Once deployed, NISAR will be a powerful and unique Earth-orbiting radar instrument that will provide L and S band dual-frequency SAR data with high repeat cycle (12 days exact repeat orbit), high resolution (3-10 meters range resolution) and large swath (> 240 km), with capability of acquiring full-polarimetric and repeat-pass interferometric data [1], [2]. NISAR systematic observation at L-band and Sband over Indian region will provide very valuable timeseries data for the ecosystem sciences addressing the critical issues of forest carbon stock estimation and monitoring carbon fluxes from vegetation disturbances; agriculture crop monitoring and changing cropping patterns; spatio-temporal distribution of field-scale soil moisture and inundation dynamics of wetlands. Accordingly, several ecosystems science products have been planned from NISAR as value added products or NISAR Level-4 science products to be used as input for climate change models, various land-based applications and decision support systems. In order to ensure that the products meet the desired accuracy levels, extensive in situ measurements from a network of ground validation sites has been planned for calibration and validation of these products. This paper provides an overview of the ground measurement sites network established in India for calibration and validation of NISAR Ecosystems Science products being developed for the Indian region.

Non-destructive estimation of chlorophyll-a (Chl-a) and chlorophyll-b (Chl-b), is required for assessment and understanding of vegetation structural and functional dynamics. We have developed an empirical model for estimating Chl-a and Chl-b for different crops using Airborne Visible-Infrared Imaging Spectrometer Next Generation (AVIRIS-NG) data. Second derivative reflectance at 672 nm and 587 nm has shown maximum sensitivity towards Chl-a and Chl-b content, respectively. Separate linear models within a bootstrapped resampling framework were developed. During calibration, mean R2 of 0.46 and 0.51 were obtained for Chl-a and -b respectively. While during model validation, in both the cases mean R2 was found to be 0.3. This work shows the capability of airborne hyperspectral imaging data in segregating chlorophyll types using a simplistic modeling framework. (c) 2022 COSPAR. Published by Elsevier B.V. All rights reserved.

Pre-harvest estimate of sugarcane production is required by sugar mill officials for proper planning about intra or inter-regional trading of sugarcane if expected production is more or less than mill's crushable capacity. Integration of optical and synthetic aperture radar (SAR) remote sensing has shown to improve biomass prediction accuracy of a perennial crop like sugarcane, particularly when optical data is unavailable due to presence of clouds. This study aims at estimating sugarcane yield using optical data from Sentinel-2 and SAR data from Sentinel-1 at mill catchment level of four sugar mills in Gujarat and Maharashtra, India. A variety of machine learning (ML) algorithms, including those based on Bayesian inference, as well as ensemble methods like bagging or boosting, were utilized to predict biomass. Additionally, a specific type of ensemble technique known as model stacking was also employed for predicting cane biomass. Fusing optical and SAR based yield driving variables from different active growth phases of sugarcane in an ensemble modeling framework explained about 63–70% variations of pixel-level above ground biomass during model training and 44–60% during testing stage. Aggregation of pixel level sugarcane yield at micro-zone level (cluster of villages) showed good prediction accuracies in Gujarat (NRMSE of 18%) and Maharashtra (NRMSE of 32%) at least 1–2 months before harvesting.

A Very Severe Cyclonic Storm ‘Yaas’ developed over the Bay of Bengal (BoB) on 23 May 2021 and crossed over the Odisha coast on 26 May with maximum sustained wind speed of 75 kts. Herein, a pathway has been developed and exemplified for ‘Yaas’ through three-stage cyclone-induced hazard tracking. Days before the cyclone formation, cyclone genesis potential parameter, sea surface temperature (SST) (> 30 °C) and tropical cyclone heat potential (anomaly of 40–80 kJ/cm 2 ) indicated a strong possibility of cyclogenesis in the BoB. A Lagrangian advection model used for its track prediction with 24-h lead-time provided an accuracy of ~ 19 km and ~ 6 h in its landfall location and time. Further, intensity prediction was done using numerical weather prediction model. Geostationary satellites, INSAT-3D/3DR, were used to visualize cyclone structure. Passing of cyclone had its reverbarations in oceans, which are observed in SST drop of ~ 3 °C, salinity and density increase by ~ 1 psu and ~ 2 kg/m 3 , respectively. During the period, 23–26 May 2021, the Ekman suction velocity and chlorophyll concentration were found significantly high at ~ 5 m/day and > 0.5 mg/m 3 , respectively. Forecast of storm surge was found to be between 3.5 and 4 m at coastal locations. Significant wave height was found to be 5.5–9.2 m. The coastal inundation forecast for 24 May 2021 provided its quantitative maximum inland extent. Finally, loss of the crop, fishery and forest areas by strong winds and inundation/ingress of saline water associated with storm surge were examined using SAR and optical data.

Agricultural production in India is highly vulnerable to climate change. Transformational change to farming systems is required to cope with this changing climate to maintain food security, and ensure farming to remain economically viable. The south Asian rice-fallow systems occupying 22.3 million ha with about 88% in India, mostly (82%) concentrated in the eastern states, are under threat. These systems currently provide economic and food security for about 11 million people, but only achieve 50% of their yield potential. Improvement in productivity is possible through efficient utilization of these fallow lands. The relatively low production occurs because of sub-optimal water and nutrient management strategies. HHaJathrough Historically, the Agro-met advisory service has assisted farmers and disseminated information at a district-level for all the states. In some instances, Agro-met delivers advice at the block level also, but in general, farmers use to follow the district level advice and develop an appropriate management plan like land preparation, sowing, irrigation timing, harvesting etc. The advisories are generated through the District Agrometeorology Unit (DAMU) and Krishi Vigyan Kendra (KVK) network, that consider medium-range weather forecast. Unfortunately, these forecasts advisories are general and broad in nature for a given district and do not scale down to the individual field or farm. Farmers must make complex crop management decisions with limited or generalised information. The lack of fine scale information creates uncertainty for farmers, who then develop risk-averse management strategies that reduce productivity. It is unrealistic to expect the Agro-met advisory service to deliver bespoke information to every farmer and to every field simply with the help of Kilometre-scale weather forecast. New technologies must be embraced to address the emerging crises in food security and economic prosperity. Despite these problems, Agro-met has been successful. New digital technologies have emerged though, and these digital technologies should become part of the Agro-met arsenal to deliver valuable information directly to the farmers at the field scale. The Agro-met service is poised to embrace and deliver new interventions through technology cross-sections such as satellite remote sensing, drone-based survey, mobile based data collection systems, IoT based sensors, using insights derived from a hybridisation of crop and AIML (Artificial Intelligence and Machine Learning) models. These technological advancements will generate fine-scale static and dynamic Agro-met information on cultivated lands, that can be delivered through Application Programming Interface (APIs) and farmers facing applications. We believe investment in this technology, that delivers information directly to the farmers, can reverse the yield gap, and address the negative impacts of a changing climate.

Mapping of soil micronutrient variability is critical for improving agronomic biofortification. This study used 1778 surface soil samples collected from four agro-climatic regions of the Indo-Gangetic Plain of India to produce digital soil maps of available Zn, Cu, Fe, and Mn using 52 environmental covariates at a resolution of 150 m. The micronutrient prediction accuracy was compared for 14 machine learning approaches and their ensemble model. The hybrid ensemble model outperformed all 14 base learners and was subsequently used for producing micronutrient maps. All four micronutrients exhibited sufficient spatial variability. Both available Zn and Fe maps exhibited lower prediction uncertainties. Moreover, the inter-relationship between micronutrient con-centration in soil and rice grain was explored to understand the Zn and Fe biofortification potential. The linear regression models revealed moderate agreement between soil available and grain micronutrient concentrations, with R2 values of 0.52-0.63 for Zn and Fe, respectively. The developed models were used to predict grain Zn and Fe content from their respective soil concentrations, indicating the potential of the tested approach to identify specific pockets where rice varieties with biofortification potential can be planted. In the future, the digital soil mapping approach tested herein can help policymakers with regional decision-making, encouraging nutrient -based subsidy and investment opportunities and sustainable micronutrient recommendations toward micronutrient-enriched food. Further research is needed to develop a digital soil intelligence platform using micronutrient DSM products in resource-poor countries.

Soil salinization is one of the major land degradation processes spread over millions of hectares of global land. Hyperspectral Remote Sensing (HRS) coupled with modern data mining approaches help in real-time and cost-effective assessment or monitoring of salt-affected soils. This study aimed at predicting soil salinity across five sites in India using the Airborne Visible-Infrared Imaging Spectrometer - Next Generation (AVIRIS-NG) data in low to moderately salt-affected cropland soils. We have identified four unique spectral absorption features having sensitivity towards soil salinity through a hybrid feature selection algorithm. Soil electrical conductivity (EC) was estimated using different machine learning (ML) based models such as random forest (RF), gradient boosting machines (GBM), and deep learning (DL). An ensemble of RF and DL models showed the best performance with the coefficient of determination (R2) of 0.89 and 0.55 and normalized root-mean-squared error of 0.15 and 0.16 in training and test datasets, respectively. We also proposed a new hyperspectral soil salinity index using Shannon entropy-based aggregation of selected absorption features. The newly proposed index outperformed other majorly used remote sensing-based salinity indices. It also showed a strong correlation with measured EC (r = 0.68) and ML-predicted soil EC (r = 0.78), both being significant at 1% level of significance. The index was effective in classifying HRS images into six distinct salinity classes. We also assessed the feasibility of applying the proposed salinity index for future hyperspectral missions through the simulation of various spectral-spatial resampling scenarios and estimated the optimal spectral and spatial resolution for salinity prediction. The hyperspectral salinity index can be directly estimated from HRS data without the need for time-consuming and expensive field samplings and used as a proxy to evaluate soil salinity status under field conditions.

Estimating sugarcane ( Saccharum officinarum L. ) production at micro-scale prior to harvest is required for fixing of Fair and Remunerative Price (FRP) payable by sugar factories, levy price of sugar and its supply for public distribution systems and regulating supply of free-sale sugar. This may also help the sugar mill owners to plan for crushing the expected cane biomass, estimate the production of sugar in each mill and look for opportunities to sell or buy from nearest sugar mills if expected production is more or less than factory’s crushable capacity. A pilot-scale study was carried out in four sugar mills of Gujarat and Maharashtra states during 2017–2019 period. Multi-date multispectral data from LISS IV, LISS III of Resourcesat-2&2A, GPS and mobile-based ground truth data and Crop Cutting Experiment data (CCE) were used. Crop discrimination in the form of fresh and ratoon, field-scale crop health assessment, yield-model development and mill-level crop acreage and production estimation were carried out. LISS IV data along with error-free GPS-based polygons could lead to discrimination with 95% accuracy and between 88–91% with mobile-based point locations. The mill-level production was found to have less than 10% deviation from reported production. The field-scale assessment and enumeration could lead to mill-level crushable cane production forecast 2 months before harvest. Future efforts are needed to utilize agro-met products and SAR-based metrics to improve the production forecasting.

Supervised classification of time series image classification through state-of-the-art Machine Learning algorithms such as Random Forest demands good quality training data for achieving good classification accuracy. A pilot study has been carried out to find out the effect of training data quality on discriminability of kharif maize crop from competing crops using multi-date C-band Synthetic Aperture Radar (SAR) data. The study was done in five districts of Telangana and four districts each of Madhya Pradesh and Maharashtra. Fairly good discrimination of kharif maize using C-band SAR with average classification accuracy of 85% have been obtained where at least 30 good quality ground truth (GT) polygons and similar number of GT for competing crops were available.