Conference Agenda

Feeding a growing global population while minimizing the subsequent environment impact are twin challenges faced by the international communities on food production and security. While agriculture is the largest fresh water consuming sector on the globe, climate change has created further uncertainties in water availability by changing climate patterns or reducing glacier size, putting to a greater extent, food security at stake. Although effort for timely monitoring food production has been made by international communities active on food security, for example using earth observation technologies (https://www.earthobservations.org/area.php?a=fssa), the environmental impact of water use needs to be further addressed. There are tremendous differences in the quantum of water used to produce a unit of grains, also called water productivity (WP), between farm fields in various part of the world, because of various cropping conditions and different water and farming management practices (1). It is therefore very opportune and important not only to measure this indicator, but also to dissect potential drivers of this parameter in context of food security, by identifying areas where the variability occurs, and to propose subsequently the strategies of improvement.

Agricultural water productivity (WP) is a measure of water use efficiency expressed as a ratio of crop production or crop yield to the water consumption for this production. The objective of the proposed project is to assess both the agricultural output (enumerator) and the water consumption for crop growth (denominator) using satellite information and compute subsequently the water productivity, this on two study areas in Europe and China. The outcome of the research could be used as a scientific evidence for water use policy making by considering the environmental impact while meeting food security imperatives.

Retrieving the crop growth information from multiple source satellite data in support of the agricultural management

Abstract: Remote sensing community has entered into a new era with the huge volume of satellite images at around 10 to 30 meter resolution fully and open available, including the sentinel series satellite in Europe and GF series satellite in China. These satellites brought more data options for the application in agricultural monitoring. The capability of agricultural monitoring in general is expected to be enhanced and improved with these satellite data in term of the monitoring spatial extent and the quality of the retrieved crop growth information. However, the agricultural cultivation is diverse in China and the rest of the world. There are existing large fields with mono crop and small fields with multiple strips of various crop types. This fact is impacting on the application of satellite data for agricultural monitoring. Therefore, the compromise of application has to be made between the optimized spatial resolution of satellite data and the field size of the study area. In general, the field size is quite small in many parts of farm land in China in comparison with that in Europe. The fine resolution satellite data are always expected to be used in the agricultural monitoring in China. In this study, 8 study sites are selected representing the major cropping systems, including winter wheat, maize, rice, sugarcane and vegetables. These sites also are representing the agricultural systems in the flat area or in hilly area, irrigated or rainfed, in the North and the South. The Sentinal1/2 and GF1/3/5/6, CBERS data are to be mainly data sources to support this study. The remote sensing parameters, like LAI/FPAR/FCOVER/NDVI are being retrieved with the adapted algorithm. The crop classification algorithm is applied to make crop type maps. Through this joint project and the heavy involvement of young scientists from Europe and China, the satellite data finely processing and information retrieval algorithm is being exchanged and it is expected to bring a step forwards to support agricultural monitoring at fine scale.

Agriculture is the largest consumer of water worldwide, accounting for about 70% of the global fresh water withdrawals. Irrigation efficiency and crop water use efficiency are key concerns for agricultural water management. The objective of the project is to assess irrigation water needs and crop water productivity based on the integrated use of satellite data with high resolution, ground hydro-meteorological data and numerical modelling, which is particularly significant for large un gauged agricultural areas. In such studies, satellite observation-based products or information with high accuracy and continuously spatial and temporal coverage are essential to support monitoring and modelling of agricultural water use and efficiency at farm and basin scales. The following progresses have been made in the first year of project implementation:

1) Soil moisture retrieval from SMOS data by a new multi-temporal and multi-angular approach. Improvement of SMOS (Soil Moisture and Ocean Salinity) soil moisture retrieval accuracy was made by a proposed multi-temporal and multi-angular approach. This approach can simultaneously retrieve soil moisture, vegetation optical depth and two soil parameters. Compared with the ground measurements, the results from this new approach in most sites showed advanced accuracy against the existing SMOS soil moisture products from SMOS.

2) Crop mapping from Sentinel-2 MSI data by machine learning method. Timely and accurate crop classification is crucial information for agriculture management. However, such information is often not available during the agricultural practice. The European Space Agency (ESA) satellite Sentinel-2 has multi-spectral bands ranging in the visible-red edge-near infrared-shortwave infrared (VIS-RE-NIR-SWIR) spectrum. Understanding the impact of spectral-temporal information on crop classification is helpful for users to select optimized spectral bands combinations and temporal window in crop mapping when using Sentinel-2 data. We have developed a crop mapping algorithm by applying multi-temporal Sentinel-2 data acquired in the growing season to a machine learning algorithm, i.e., the Random Forest algorithm, to generate the crop classification map at 10 m spatial resolution. This algorithm was applied to the Shiyang River Basin, in the northwest of China with arid/semi-arid climate and scarce water resource, proper agricultural planting structure is of importance for efficient use of limited water resource. Four experiments with different combinations of feature sets were carried out to explore which Sentinel-2 information was more effective for crop classification with higher accuracy. The results showed that the augment of multi-spectral and multi-temporal information of Sentinel-2 improved the accuracy of crop classification remarkably, and the improvement was firmly related to strategies of feature selections. Compared with other bands, red-edge band 1 (RE-1) and shortwave-infrared band 1 (SWIR-1) of Sentinel-2 showed a higher competence in crop classification. The combined application of images in the early, middle and late crop growth stage is significant for achieving optimal performance. A relatively accurate classification (overall accuracy = 0.94) was obtained by utilizing the pivotal spectral bands and dates of image. In addition, a crop map with a satisfied accuracy (overall accuracy > 0.9) could be generated as early as late July.

3) Estimate of gross/net crop water requirements, actual crop water use and irrigation efficiency by high resolution satellite Sentinel-2 and ETMonitor model of center pivot irrigation system at farm scale. A case study was conducted for two major crops, i.e. wheat and potato, in Inner Mongolia autonomous region of China, where modern equipment and adequate irrigation methods are deployed for efficient use of water resource. The method estimated and mapped explicitly the net crop water requirements, the water losses (water droplet evaporation directly to the air during irrigation application before droplets fall on the canopy) and canopy interception loss, and the gross irrigation water requirements were mapped finally. Daily estimates of crop water requirement and actual water use were generated using data from Multi Spectral Instrument (MSI) of Sentinel-2 with fine resolution combined with meteorological forcing data and soil moisture retrievals. A good agreement between the estimated values and ground observations for crop actual water use and for water losses were obtained. It also showed that the losses of total irrigated volume were 25.4% for wheat and 23.7% for potato, respectively, and found that the water allocation was insufficient in fulfilling the water requirement in this irrigated area. This suggested that the amount of gross irrigation water was inadequate to meet the crop water requirement and the inherent water losses occurred during water application by center pivot irrigation systems.

4) FEST-EWB energy-water balance model is coupled with derived vegetation and land surface temperature (LST) data over two of the project case studies in Italy: the Chiese and Capitanata irrigation consortia.

Remotely- sensed data at different temporal and spatial resolution of vegetation parameters (leaf area index (LAI), fractional coverage of vegetation, albedo) which are used as inputs to hydrological model are obtained at high spatial and temporal resolution merging Sentinel 2 data with Landsat 7 and 8 for the Capitanata area and also with MODIS data for the Chiese basin.

Satellite LST is further retrieved from Landsat 7 and 8 at 30 m spatial resolution as well as from MODIS and Sentinel 3 data at 1km resolution, to be used for the hydrological model calibration.

Indeed, the energy–water balance FEST-EWB model (flash flood event-based spatially distributed rainfall–runoff transformation—energy–water balance model) computes continuously in time and is distributed in space soil moisture (SM) and evapotranspiration (ET) fluxes solving for a land surface temperature that closes the energy–water balance equations. The comparison between modelled and observed LST was used to calibrate the model soil parameters with a newly developed pixel to pixel calibration procedure. The effects of the calibration procedure were analysed against ground measures of soil moisture and evapotranspiration.

The calibrated and validated hydrological models coupled with satellite data will provide consistent outputs of the different hydrological processes overcoming the limitation of remote sensing data caused by cloud cover, retrieval algorithms, temporal and spatial resolutions, etc.

Preliminary results of the amount of precision irrigation water supply and the Evapotranspiration deficit at pixel scale will also be shown.

The overall objective of project 59197 is to carry out agro-ecosystem health diagnosis and to investigate agricultural processes based on various in situ and EO data, allowing to conserve, protect and improve the efficiency in the use of natural resources to facilitate sustainable agriculture development. Two study areas are identified: the Rur basin observatory in Germany and the Huaihai Economic Zone in China. This selection enables us to investigate the transferability of results between European and Asian agricultural systems and to ensure global applicability.

Five work packages (WP) are proposed to fulfill the project’s research goal, including: 1) Crop classification based on multi-source remote sensing data; 2) Retrieval of soil parameters and plant growth stress factors; 3) Monitoring of crop biophysical variables; 4) EO-based evaluation of cropland carbon budgets; 5) Data assimilation of remote sensing products for synoptic systems analysis.

This paper presents the progress of our research activities since the kick-off of the Dragon 5 programme.

For crop classification, to take full advantage of high spatial resolution of panchromatic images and polarimetric synthetic aperture radar (PolSAR) data with rich scattering information, a novel dual-domain data fusion method is explored by combining spherically invariant random vector (SIRV) model with a novel generalized adaptive linear combination approximation (GALCA) technology. Gaofen (GF)-2, 3 and Radarsat-2 data are used. Experimental results show that this method is able to significantly improve the spatial resolution of PolSAR data without degrading polarimetric information.

Studies in agricultural hydrology are highly stressed, including the estimation of soil moisture, evapotranspiration, and groundwater table depths. Surface soil moisture is estimated by a Copernicus Sentinel-1 time series approach at a sub-field scale. The C-band SAR data is processed and analyzed on the cloud computing platform Google Earth Engine. This allows for high-performance investigations at larger scales and high resolution as well as straightforward transfer to alternative regions. The results are validated against in situ soil moisture networks consisting of state-of-the-art time domain reflectometry and innovative cosmic-ray neutron sensors. Besides, we propose an optimal estimation approach combined using SAR and optical remote sensing imagery, in order to retrieve vegetation water content, roughness and soil moisture simultaneously. Three optical remote sensing indices are investigated. The proposed method is performed by using Sentinel-1and Landsat 8 data. Retrieved results are validated against ground measurements and show a good agreement between remote sensing estimates and ground measurements. Additionally, it is found that the result of estimated vegetation water content and the parameterization scheme of vegetation parameters have pronounced influence on the accuracy of soil moisture estimates. Evapotranspiration is estimated by Spinning Enhanced Visible and Infrared Imager (SEVIRI) as well as Landsat observations, the implementation of Sentinel-2 and Sentinel-3 is foreseen. The Evaporative Drought Index as a relationship between actual and potential evapotranspiration provides levels of water stress and informs about the irrigation demand in agriculture. Extensive validation is performed against in situ data of the Integrated Carbon Observation System (ICOS). To investigate the small-scale heterogeneity of evapotranspiration for the area of Eddy Covariance footprints, unmanned aerial vehicles with multispectral and thermal sensors are employed. Here we discuss the impact of soil texture on plant growth and evapotranspiration. To ensure sustainable groundwater abstraction for irrigation, we predict groundwater table depth anomalies by machine learning approaches. Long-Short-Term Memory networks are trained on integrated hydrologic simulations from groundwater to the upper atmosphere. This enables the utilization of precipitation and soil moisture information to predict groundwater table depth anomalies with high agreement to reference wells. Besides further vegetation and weather indicators, the hydrological conditions are also drivers for fire risks.

We propose a new method of surface soil salinity estimation in coastal areas based on ground-based digital photographs to obtain soil salinity information quickly and conveniently under complicated weather conditions. Color parameters obtained from digital images provides a new approach for soil salinity estimation effectively.

Crop parameters in farmland areas of the Huaihai economic region, such as leaf area index (LAI) and canopy chlorophyll content (CCC) are accurately retrieved by new spectral indices such as OSAVI_{[864, 866]} and SR_{[790, 631]} and a hybrid inversion model, which provides data support for crop growth monitoring and yield estimation. Then, the corresponding net ecosystem productivity (NEP) is estimated based on the improved Carnegie Ames Stanford approach (CASA) model and geostatistical model, which lay a foundation for the assessment of farmland ecosystem carbon budget in this region.

To study the method of cropland carbon budgets evaluation, a projection-based model driven by satellite remote sensing data (GIMMS NDVI) that represent temporal dynamics of climate, vegetation, and land cover is developed to investigate the spatiotemporal changes of soil organic carbon (SOC) during two time periods: 1980s and 2010s. We find that the spatiotemporal patterns in SOC along the gradients of temporal covariates are similar across space and time. Model projections with temporal covariates result in more accurate estimates.

The project #57475 deals with the set up and testing of pre-operative algorithms and processing chain convolving ESA/TPM EO data and including the exploitation of the hyperspectral images provided by the ASI PRISMA mission (that can be considered as the European precursor of the Copernicus candidate CHIME) and Chinese multi-bands EO data. The project aims at developing operational thematic products tuned to match the farmer’s requirements. User requirements that, for EU, regard the EC policies related to “Agriculture & Food Security” application domain. The project intends to identify specific “use case demonstration” related to:

i) the agriculture and topsoil monitoring;

ii) forecast yields, also in terms of proteins content and crop pest and disease.

iii) identification of sustainable agricultural practices;

The project cross cutting methodological approach foresees the exploitation of the DIAS systems (e.g. ONDA) to support a multi-sensor/spatial/temporal approach for the “use case demonstration”.

As regards the agricultural crop and topsoil monitoring the project, within this first year of activity, has started a comparison of the different retrieval algorithms for both domains: vegetation and topsoil. For the crop-vegetation domain the different approaches for the improvement of the accuracy in the estimation of crop biophysical variables such as pigments (including carotenoids and anthocyanins) and variables related to nitrogen and water stress have been evaluated. To this aim, parametric methods like (vegetation indexes) and non-parametric both linear and non-linear regression methods (e,g. Linear Regression (LR), Partial Least Square Regression (PLSR), Random Forest regression (RFR)) and PROSAIL RTM based approaches as hybrid methods will be compared to evaluate their performance in estimating the crop biophysical variables of interest. Optimization of the retrieval process will be tested with a synergistic use of both S2, GF6 data set and PRISMA hyperspectral data when applied to local scale retrieval applications. As regards topsoil properties (i.e. texture and SOC) retrieval algorithms such as chemometrics techniques and multivariate calibrations, like multiple linear regression (MLR), principal components regression (PCR), partial least-squares regression (PLSR) neural networks (ANN), including support vector machines (SVM) have been explored. Moreover, in this first year of activity intensive, according to the limitation due to theCOVID-2019 pandemic, field campaigns in different sites on cultivar and on different bare soil fields have been conduction in order to define a cal/val data set to validate the retrieval algorithms performances and the products accuracies also considering errors and uncertainties in the remote sensing observations. Whenever possible campaigns have been performed contemporary to EO data acquisitions. First year results for vegetation and top soil domains, regard the comparison of the different retrieval procedures and the start of collection of a cav/val data set to be applied in the following years of activity

As for the retrieval for agronomical variables of interest like yields, grain quality and pest and disease the focus is on the development of data assimilation algorithms that specifically address issues of the multi-scale and multivariate nature of multitemporal optical (S-2, S-5 and GF-6) and eventual SAR datasets. At present we are evaluating two variables for the assimilation i.e. LAI and soil moisture. Within this year of activity, the two different assimilation algorithms (deterministic and stochastic) based on the Ensemble Kalman Filter (EnKF) and Particle Swarm Optimization (PSO) methods have been evaluated. These methods will update the state variables and/or parameters of the Aquacrop model, allowing to estimate variables of agronomic interests, such as crop yield and grain protein quality. First year results for the retrieval of agronomical variables of interest (e.g. yield and grain quality), regard the consolidation of the different assimilation procedures, while the collection of the cav/val data set is actually ongoing on the different test cases in Italy and China.

The presentation will provide an overview of all the ongoing activities for project ID#57475 in terms of: EO data collection; data processing: cal/val acquisition strategies’ set up for the different sites.