Temporal changes in soil moisture (SM) and green vegetation affecting the spectral reflectance can heavily reduce the accuracy of topsoil properties estimation from satellite imaging. To minimize these effects on the soil organic carbon (SOC), sand, silt and clay estimations, an external parameter orthogonalization (EPO) model developed using laboratory based measured spectra was tested on PRISMA hyperspectral satellite data. The estimation of soil properties was performed using different machine learning algorithms. The results show that as compared to the uncorrected spectra, removing the effects of both green vegetation and SM (EPOSM+GV) from the reflectance spectra leads to 18%, 13%, 10%, and 24% improvement in the R2for clay, silt, sand and SOC retrieval, respectively. The Gaussian Process Regression (GPR) algorithm provides the best results for all of the soil properties with an RMSE of 9.5%, 14.2%, 6.9% and 0.68% for clay, silt, sand and SOC retrievals, respectively.

The project "EO AFRICA EXPLORERS—PRISMA 4 AFRICA", funded by the ESA, aims at combining products derived from hyperspectral (e.g., PRISMA, EnMAP, DESIS) and thermal data (e.g., ECOSTRESS, Landsat) to detect vegetation anomalies and identify whether they may be related to biotic or abiotic stress factors. In this context, water stress as a main abiotic stress was considered. At this aim we investigated the possibility to exploit PRISMA combined with the ECOSTRESS data to derive evapotranspiration (ET). Even though LST products are available and appear of good quality, the lack of ancillary data prevents ET products to be generated timely. 30-m ET map was produced by combining of ECOSTRESS and PRISMA data and used for water stress estimation. The resulting products proved to be sufficiently accurate to describe the crop water stress at the field scale.

The increasing availability of hyperspectral satellite-based systems has emphasized the importance of hyperspectral pansharpening. This technique, which fuses a high-resolution panchromatic image with a lower resolution hyperspectral cube, aims to produce an enhanced hyperspectral cube from a spatial point of view while preserving the spectral quality. In this paper we first recall the Guided Full Resolution Frame-work, our novel deep learning training approach, to optimize PRISMA data pansharpening, leveraging both synthetic and real PRISMA data. Then, we introduce an innovative model design and a convolutional attention module as part of architectural enhancement, further improving the performance with respect to the previous work. The results reveal that the minimal computational overhead from this module improves distortion metrics, highlighting its effectiveness in advancing hyperspectral pansharpening techniques.

The project "Topsoil properties Estimation from Hyperspectral Remote sensing for Agriculture" (TEHRA), funded by the Italian Space Agency (ASI), aims at developing methods and algorithms for the estimation of soil properties of agronomic and environmental interest from PRISMA satellite hyperspectral data, that could support: 1) the adoption of more sustainable and climate-smart farming practices, e.g. through the implementation of precision agriculture applications; 2) monitoring in support of agricultural and environmental policies, e.g. related to climate change and for the encouragement of the adoption of practices preserving soil health.In this paper, some results of the first year of the project are illustrated. They concern: 1) a scenario definition study; 2) studies on the confounding effect of soil moisture and crop residues; 3) exploitation of multi-temporal PRISMA data and 4) data fusion with proximal soil sensing.

The unique spectral content provided by PRISMA's hyperspectral sensor gives the possibility to study the Earth's surface and environment from space in unprecedented detail. In this respect, our work presents the results of an Italian Space Agency-funded project aiming to develop eight prototypes for providing Value Added products based on such data. Prototypes focus on vegetation, urban areas, water quality, material detection, and natural hazards, combining multiple state-of-the-art techniques based on Machine Learning, physical models, and index-based algorithms. This is particularly relevant given the increasing demand for accurate information to address sustainable policies and support decision-making processes. Through a series of case studies, we highlight the versatility and utility of PRISMA's hyperspectral data for various scientific and operational applications.

The increasing availability of huge amounts of satellite data, together with the increasing computation power available at relatively low cost, is requiring and at the same time allowing the development of new algorithms to automatically extract information from the data. In this work, we propose a new method for automatic detection, from series of satellite images, of possible anomalies relative to crop parcels declared by farmers in the framework of EU's Common Agricultural Policy (CAP). Differently from other recently explored methods, our technique is not based on a crop classification approach. On the contrary, we approached the problem as an anomaly detection problem, and our method bases only on the quite realistic and general assumption that declarations are mostly correct, with a moderate number of outliers. Therefore, our technique is robust to variations in weather conditions, terrain morphology and agriculture practices. In order to detect the anomalies, the method computes the “distances” between the different parcels with a given declared crop. In particular, the time series of the features extracted from the satellite data on the different parcels are compared. Their distance is defined according to the Dynamic Time Warping (DTW) method, robust to temporal variations. The tests performed were very good, and the technique has been already operationally used with satisfactory results. In particular, the automatic anomaly detection approach has made it possible to verify all the farmers' declarations in a large area (a significant portion of Italy), and will make it possible to process even larger areas, such as the whole Italy or the whole Europe.

Remote sensing present some new challenges for deep learning, because (also to compensate the scarce detail level) multimodal, multisource and multitemporal data should be jointly exploited. For example, time series of optical multispectral/hyperspectral or synthetic aperture radar (SAR) data probe different properties of the observed scene, based on their different wavelength, acquisition geometry, etc., and with possible data gaps. To address this task, we propose a new deep learning architecture that exploits a sequence of deep convolutional neural networks (CNN) and a recurrent neural network (RNN). In the proposed architecture, all the data (with their spectral, spatial and temporal information) are used jointly and optimally in the sense that no imputation is enforced, but the internal weights providing the best classification results are estimated from the data themselves (hence the proposed name ODIN - Optimal Data Imputation Network). We have tested the proposed architecture, using Sentinel SAR and multispectral image series, on land cover and crop classification, an important remote sensing application. The obtained results are very promising, with an error rate below 1%, and show good spatial consistency without loss of spatial resolution.