The annual certification of organic agriculture products includes an in situ inspection of the fields declared organic. This inspection is more difficult, time-consuming, and costly for large farms or in production regions located in remote areas. The global objective of this research is to assess how spatial remote sensing may support the organic crop certification process by developing a method that would enable certification bodies to target for priority in situ control crop fields declared as organic but that would show on satellite imagery an appearance closer to conventional fields. For this purpose, the ability of multispectral satellite images to discriminate between organic and conventional maize fields was assessed through the use of a set of four satellite images of different spatial and spectral resolutions acquired at different crop growth stages over a large number of maize fields (32) that are part of an operational farm in Germany. In support of this main objective, a set of in situ measurements (leaf hyperspectral reflectance, chlorophyll, and nitrogen content and dry matter percentage, crop canopy cover, height, wet biomass and dry matter percentage, soil chemical composition) was conducted to characterize the nature of the biochemical and biophysical differences between organic and conventional maize fields. The results of this research showed that highly significant biochemical and biophysical differences between a large number of organic and conventional maize fields may exist at identified crop growth stages and that these differences may be sufficiently pronounced to enable the complete discrimination between crop management modes using satellite images issued from quite common multispectral satellite sensors through the use of spectral or spatial heterogeneity indices. These results are very encouraging and suggest, for the first time, that satellite images could effectively support the organic maize certification process.

This paper presents a novel system that produces multi-year high-resolution irrigation water demand maps for agricultural areas enabling a new level of detail for irrigation support for farmers and agricultural stakeholders. The system is based on a scalable distributed Deep Learning (DL) model trained on dense time series of Sentinel-2 images and a large training set for the first year of observation, and fine-tuned on new labeled data for the consecutive years. The trained models are used to generate multi-year crop type maps, which are assimilated together with the Sentinel-2 dense time series and the meteorological data into a physically based agro-hydrological model to derive the irrigation water demand for different crops. To process the required large volume of multi-year Copernicus Sentinel-2 data, the software architecture of the proposed system has been built on the integration of the Food Security TEP and the data-intensive arpngicial intelligence Hopsworks platform. While the Food Security TEP provides easy access to Sentinel- 2 data and the possibility of developing processing algorithms directly in the cloud, the Hopsworks platform has been used to train DL algorithms in a distributed manner. The experimental analysis was carried out in the upper part of the Danube Basin for the years 2018, 2019 and 2020 considering 37 Sentinel-2 tiles acquired in Austria, Moravia, Hungary, Slovakia and Germany.

The Danube Basin has been hit by several droughts in the last few years. As climate change makes weather extremes and temperature records in late winter and early spring more likely, water availability and irrigation possibilities become more important. In this paper, the crop water demand at field and national scale within the Danube Basin is presented using a dense time series of multispectral Sentinel-2 data, for crop type maps derived with deep learning techniques and physically-based models for crop parameter retrieval and crop growth modelling.

Keywords: M4Land, land management classification system, Sentinel, canopy reflectance model SLC, crop growth model PROMET To support food security, information products about the actual cropping area per crop type, the current status of agricultural production and estimated yields, as well as the sustainability of the agricultural management are necessary. Based on this information, well-targeted land management decisions can be made. Remote sensing is in a unique position to contribute to this task as it is globally available and provides a plethora of information about current crop status. M4Land is a comprehensive system in which a crop growth model (PROMET) and a reflectance model (SLC) are coupled in order to provide these information products by analyzing multi-temporal satellite images. SLC uses modelled surface state parameters from PROMET, such as leaf area index or phenology of different crops to simulate spatially distributed surface reflectance spectra. This is the basis for generating artificial satellite images considering sensor specific configurations (spectral bands, solar and observation geometries). Ensembles of model runs are used to represent different crop types, fertilization status, soil colour and soil moisture. By multi-temporal comparisons of simulated and real satellite images, the land cover/crop type can be classified in a dynamically, model-supervised way and without in-situ training data. The method is demonstrated in an agricultural test-site in Bavaria. Its transferability is studied by analysing PROMET model results for the rest of Germany. Especially the simulated phenological development can be verified on this scale in order to understand whether PROMET is able to adequately simulate spatial, as well as temporal (intra- and inter-season) crop growth conditions, a prerequisite for the model-supervised approach. This sophisticated new technology allows monitoring of management decisions on the field-level using high resolution optical data (presently RapidEye and Landsat). The M4Land analysis system is designed to integrate multi-mission data and is well suited for the use of Sentinel-2’s continuous and manifold data stream.