Conference Agenda

In the execution period of Dragon 5, our team have made a quantitative use of remote sensing information in vegetation monitoring and developed products targeted at optimizing the agricultural production and allowing a more sustainable ecology. The project has been concentrated on the following themes: retrieval of biophysical variables of vegetation, estimation of bare soil properties, prediction of crop yield and monitoring vegetation pests and diseases. Multi-source remote sensing data including Sentinel-2, Gaofen-1/2 and PRISMA were used. Three experimental sites were settled, Maccarese farm in Italy, some sites in Central East-Africa, and some typical sites of crop, forest and grassland in Hebei, Hubei and Inner Mongolia.

For what concerns the retrieval of crop biophysical variables, we have mainly exploited the hyperspectral mission currently available (i.e. PRISMA and ENMAP). In order to harmonize the dataset, we implemented the Ross-Li BRDF correction procedure by determining coefficients in accordance with the increasing abundance of vegetation (based on NDVI classes). Results showed that due to the small swath of the hyperspectral mission (i.e. 30km), even if the geometry of acquisition impacts the retrieval quality, its influence its small and it does not require a BRDF correction. As regards the optimization of the biophysical retrieval we have implemented the hybrid approaches on PRISMA images acquires on the test sites of Jolanda and Maccarese in Italy to retrieve structural parameter and pigments. In particular, we have implemented for the ongoing and furthers HYP missions (i.e. PRISMA, ENMAP and CHIME) the GPR MLR algorithm on the first 15 PC bands on which an active learning procedure has been performed. Furthermore, to overcome the differences between the large data set of simulated RTM spectra with the real HYP data (i.e. PRISMA, ENMAP and GF-5) the Domain Adaptation techniques have been adopted to consider the nonlinear effects in the data.

Similarly, as crop status depends also on soil fertility, we investigated the suitability of PRISMA and Sentinel-2 satellite imagery for the retrieval of topsoil properties, such as soil Organic Matter (OM), Nitrogen (N), Phosphorus (P), Potassium (K) and pH in croplands, based on different Machine Learning (ML) algorithms in the Quzhou County area, North of China. The best performing algorithms were the Support Vector Regression (SVR) for the estimation of Total N (RMSE=0.13, R2=0.58, RPD=1.56), the Random Forest (RF) for the available P (RMSE=4.70, R2=0.60, RPD=1.58) and the available K (RMSE=24.59, R2=0.59, RPD=1.37). The results revealed the potential of the new HYP sensors to predict soil nutrients in a real test case.

In China, the demand for accurate and up to date information on wheat Grain Protein Content (GPC) has gained increased urgency, driven by the rising demands in the food consumption market and intensifying international market competition. To address this challenge and the data gap, the first 500-meter spatial resolution of GPC, long-term winter wheat dataset covering major planting regions in China (CNWheatGPC-500) from 2008-2019, was created by integrating multi-source data from ERA5/MODIS and Hierarchical Linear Model (HLM), established by deciphering the spatial hierarchical relationships among temporal phenological, meteorological, and remote sensing data. CNWheatGPC-500 provides valuable insights for improving wheat production, enhancing quality control, and supporting decision-making in the agricultural sector. The CNWheatGPC-500 product is freely accessible at https://doi.org/10.5281/zenodo.10066544.

For what concerns crop threats, the core of the system aiming at detecting yellow rust outbreaks in maize and wheat crops, will be built on PRISMA satellite. Several VIs (NDVI, SIPI, PRI, PSRI, MSR) computed by using hyperspectral images will be used to implement a Diseases Infection Index. The impact of spatial resolution on the capability to detect yellow rust in crops will be one of the results of the activity. For desert locust presence risk forecasting, we have proposed a dynamic prediction method of desert locust presence risk at Somalia-Ethiopia-Kenya. Monthly prediction experiments until now were conducted, extracting high, medium and low risk areas of desert locust occurrence in the study area. Results demonstrated that the overall accuracy was 77.46%, and the model enables daily dynamic forecasting of desert locust risk up to 16 days in advance, providing early warning and decision support for preventive ground control measures for the desert locust.

In addition, we have established the remote sensing-based monitoring and risk assessment methods for biological disaster of forest and grassland. For pine wilt disease monitoring, we proposed a spatiotemporal change detection method to reduce false detections in tree-scale PWD monitoring under a complex landscape. The results showed that the proposed method effectively distinguished wilted pine trees from other easily confused objects. For grasshopper, we took the two steppe types as the research object and established a remote sensing monitoring model for grasshopper potential habitat. The results demonstrated that the most suitable and moderately suitable areas were distributed mainly in the southern part of the meadow steppe and the eastern and southern parts of the typical steppe. We also found that the soil temperature in the egg stage, the vegetation type, the soil type, and the precipitation amount in the nymph stage were significant factors both in the meadow and typical steppes.

In summary, as of the last year of the Dragon 5, the project's execution progress is consistent with the schedule, and most of the activities have achieved good results. Additionally, some scholars in the project team are conducting scientific research using the data obtained through the cooperation between the two sides.