Cloud-Based Satellite Analytics for Sustainable Agriculture: A Denitrification Assessment Framework with Applications in Bangladesh

Excessive nitrogen fertilisation in croplands creates a significant threat to environmental sustainability, contributing to soil degradation, water pollution and greenhouse gas (N₂O) emissions. To address this challenge, a precise, scalable, and accessible tool is needed to monitor nitrogen dynamics. The present research introduces a novel cloud-based satellite analytics framework to estimate the potential denitrification rate (PDR). PDR is a key biogeochemical indicator of nitrogen retention and loss in croplands. Using Earth Observation data, the framework enables rapid assessment of nitrogen removal potential, as represented by PDR, in agricultural landscapes. To estimate PDR, different linear mixed models (LMMs) were developed using Sentinel-2 data, establishing the relationships between the Normalised Difference Vegetation Index (NDVI) and vegetation cover (%), and between Sentinel-2 band 11 and soil moisture (%), which reasonably described PDR. In addition, LULC was used as a random variable in the models to predict PDR. Among LMMs, the best-fit model was the random-intercept model with a variance structure, which was selected using the Akaike Information Criterion (AIC). To evaluate field-based and satellite-derived PDR predictions, type II regression was used, which showed no significant difference between the two estimates.  To apply the research results, a user-friendly decision-support system was developed in R, leveraging Shiny, Leaflet, and Terra for interactive geospatial processing. The system was successfully deployed on the Posit Cloud, enabling seamless web-based access without requiring local software installation. Users can define areas of interest, perform PDR calculations using the best-performing linear mixed model, and visualise or download spatially explicit results. The framework was tested in the wetland-dominated agricultural regions of Arial Beel and Chalan Beel, in Bangladesh. The result demonstrates its capacity to differentiate between high-PDR wetland zones and low-PDR vegetated areas. This cloud-native approach not only supports localised nitrogen management but also offers a replicable, scalable model for agricultural regions worldwide. By translating complex satellite data into an actionable tool, this system can help researchers, policymakers, and farmers make decisions to mitigate nitrogen overuse, enhancing crop sustainability and protecting aquatic ecosystems. This work aligns with next-generation Earth observation goals by democratising access to high-level geospatial analytics and fostering sustainable agricultural practices worldwide.

Keywords: Cloud Computing; Denitrification; Earth Observation; Nitrogen Management; Remote Sensing; R; Sustainable Agriculture; Wetland Conservation

*Author of Correspondence: gmdataul@bot.jnu.ac.bd