Remote sensing of optically complex inland waterbodies is challenging due to the complex nonlinear correlation between water quality parameters and water optical properties. However, integration of deep learning techniques and representative datasets offers the potential to address these challenges effectively. This study aims to develop robust deep learning models, utilizing limited but highly representative dataset of in-situ water quality and radiometrically corrected hyperspectral remote sensing reflectance (R_rs) measurements collected from optically diverse lakes of China, for independent and simultaneous retrieval of Chlorophyll-a (Chl-a), Secchi Disk Depth (SDD), and Total Suspended Solids (TSS) using Sentinel-2 analysis ready products. The GLObal Reflectance community dataset for Imaging and optical sensing of Aquatic environments (GLORIA) provides over 400 such measurements for Chinese lakes, which are simulated to Sentinel-2 R_rs with its spectral response function to build a representative dataset. Using this dataset, Multilayer Perceptron (MLP) based Deep Neural Network (DNN) models are developed and compared with eXtreme Gradient Boosting (XGB), Random Forest (RF), and Support Vector Machine (SVM) algorithms. The DNN models outperformed in effective evaluation of Chl-a (Root Mean Squared Error (RMSE) = 14.18 mg/m³), TSS (RMSE=7.23 g/m³) and SDD (RMSE=0.12 m) on test datasets and Chl-a (RMSE=14.42 mg/m³) and SDD (RMSE=0.07 m) against Sentinel-2A validation dataset of Liangzi lake. Mixed Density Network (MDN) model showed less accuracy for Chl-a (RMSE=16.76 mg/m³) on same validation dataset. Impact of different atmospheric correction processors is also assessed and DNN models achieved their accuracy on Sentinel-2 Atmospheric Correction (Sen2Cor) processor. Finally, water quality maps for various lakes in China are produced showing realistic ranges. These results show the potential of DNN models trained with limited but representative dataset in practical applications for spatial and temporal analysis of water quality.

Accurate global carbon dioxide (CO₂) distribution with high spatial and temporal resolution is essential for understanding its dynamics and impacts on climate change. This study tackles the challenge of data gaps in satellite observations of greenhouse gases, caused by orbital and observational limitations. We reconstructed a comprehensive dataset of Column-averaged CO2 (XCO₂) concentrations by integrating re-analyzed data from the Copernicus Atmosphere Monitoring Service (CAMS) with observations from GOSAT and OCO-3 satellites. Using two advanced data reconstruction methods—Data Interpolating Empirical Orthogonal Functions (DINEOF) and Convolutional Auto-Encoder (DINCAE)—we imputed missing data, preserving spatial and temporal consistency. The combined approach achieved high accuracy, with Pearson correlation values between 0.94 and 0.95 against TCCON measurements, and we also reported root mean square error (RMSE) to assess model performance further. Our results indicate that these techniques generate a daily, high-resolution, gap-free XCO₂ dataset, enabling improved CO₂ monitoring, climate modeling, and policy development.

The Greenhouse Gases Observing Satellite (GOSAT) can help to ascertain the global distribution of carbon dioxide (CO₂) and methane (CH₄), and how the sources and sinks of these gases vary by season, year, and location. However, the data provided by the GOSAT level 2 and 3 products have certain limitations due to their lack of spatial and temporal information; even with the application of the kriging geostatistical method on the level 2 products, the processing algorithms still need further upgrades. In this study, we apply an empirical orthogonal function (EOF)-based method on the GOSAT L3 products (137 images, from January 2010 to May 2021) to estimate the column average of carbon dioxide and methane (XCO₂–XCH₄) within the entire Earth. The reconstructed results are validated against the Total Carbon Column Observing Network (i.e., TCCON), with 31 in situ stations, and GOSAT L4B column-averaged data, using 107 layers. The results show an excellent agreement with the TCCON data and exhibit an R-squared coefficient of 0.95 regarding the CO₂ measurements and 0.86 regarding the CH₄ measurements. Therefore, this methodology can be incorporated into the processing steps used to map global greenhouse gases.

This study introduces a robust model-based framework designed for the verification and validation (V&V) of Attitude Determination and Control Systems (ADCSs) in nanosatellites, focusing on magnetic actuation while still being applicable to larger spacecraft platforms. By employing Model-in-the-Loop (MIL), Software-in-the-Loop (SIL), Processor-in-the-Loop (PIL), and Hardware-in-the-Loop (HIL) methodologies, this framework enables a thorough and systematic approach to testing and validation. The framework facilitates the assessment of long-term maneuvers, addressing challenges such as initial small-attitude errors and restricted 3D movements. Two specific maneuvers are evaluated: detumbling and nadir pointing, utilizing quaternions and a comprehensive suite of sensors, including six sun sensors, a three-axis magnetometer, a three-axis gyroscope, GPS, and three magnetorquers. The methodologies—MIL, SIL, PIL, and HIL—integrate the behaviors of digital sensors, analog signals, and astrodynamic perturbations. Based on an optimized SIL environment, Monte Carlo simulations were performed to optimize control gains for nadir pointing, achieving a mean pointing accuracy of 11.69° (MIL) and 18.22° (PIL), and an angular velocity norm of 0.0022 rad/s for detumbling. The HIL environment demonstrated a mean pointing accuracy of 9.96° and an angular velocity norm of 0.0024 rad/s. This comprehensive framework significantly advances the design and verification processes for nanosatellite ADCSs, enhancing the reliability and performance of nanosatellite missions.

Groundwater level monitoring is crucial for assessing hydrological responses to climate change and human activities, which pose significant threats to the sustainability of semi-arid aquifers and the cultural heritage they sustain. This study presents an integrated remote sensing and transformer-based deep learning framework that combines diverse geospatial datasets to predict spatiotemporal variations across the plain near the Persepolis and Naqsh-e Rustam archaeological complexes—UNESCO World Heritage Sites situated at the plain’s edge. We assemble 432 synthetic aperture radar (SAR) scenes (2015–2022) and derive vertical ground motion rates greater than −180 mm yr⁻¹, which are co-localized with multisource geoinformation, including hydrometeorological indices, biophysical parameters, and terrain attributes, to train transformer models with traditional deep learning methods. A sparse probabilistic transformer (ConvTransformer) trained on 95 gridded variables achieves an out-of-sample R² = 0.83 and RMSE = 6.15 m, outperforming bidirectional deep learning models by >40%. Scenario analysis indicates that, in the absence of intervention, subsidence may exceed 200 mm per year within a decade, threatening irreplaceable Achaemenid stone reliefs. Our results indicate that attention-based networks, when coupled to synergistic geodetic constraints, enable early-warning quantification of groundwater stress over heritage sites and provide a scalable template for sustainable aquifer governance worldwide.