The fraction of mulched film is of great significance for evaluating the agricultural water-saving effect and controlling environmental plastic pollution. Unfortunately, there is no work has been done to obtain this parameter due to the mixed pixel issue of satellite imagery with medium and low resolutions, till now. In this study, we proposed a novel two-step framework for mapping fraction of mulched film based on very high resolution satellite observation and deep learning. The first step is extracting the extent of the mulched film based on a new few-shot learning model named PT-CNN (Parameter Transition Convolutional Neural Network), which aims to increase the extraction accuracy and overcome the lack of labeled training data. The second step is retrieving the fraction of the mulched film at pixel scale based on a spectrum analysis method. The result shows that the proposed PT-CNN outperforms several state-of-the-art methods in mulched film extraction, with F1-scores at 97.09% and 98.65% for white and black mulched film, respectively. Meanwhile, the retrieved pixel scale fraction of mulched film shows high consistency to in-situ measurement, with a MAE of 0.0321. The proposed method can be useful in agricultural water resource management and environmental governance.

Plastic mulch has been widely used in agricultural cultivation for its significantly increasement toward crop yield since last century, and has recently draw lots of attention from the government side, due to the environmental concerns caused by the agricultural plastic mulch. In this study, a few-shot learning based deep learning model is designed for film mulching mapping using very high resolution satellite imagery. Firstly, the image restoration model is pre-trained by massive unlabeled very resolution satellite imagery samples. Then, the film mulching mapping model is set by the pre-training model weights and trained by few labeled film mulching imagery samples. Results show the proposed method could achieve well film mulching mapping performance, and the F1-score reaches 0.9, which is useful in the agricultural irrigation management and yield prediction.

Soil moisture (SM) plays an important role in the water–heat–energy exchange and water cycle of the land ecosystem. Long-term SM products are vital in the time series study of ecology and hydrology. Therefore, it is vital to extend the time span with limited SM monitoring sensors, since there is no single long-term SM product currently. In this study, an SM product prolonging method based on an artificial neural network (ANN) and moderate-resolution imaging spectroradiometer (MODIS) optical products was proposed. The prolonging results of Soil Moisture Active Passive (SMAP) and Fenyun-3B (FY3B) products were validated in Tibetan Plateau to present the feasibility of this method. The result shows this method is feasible in areas under medium vegetation cover (0.2$R $= 0.84, RMSE3${cm}^{-3}$) within situmeasurements for both SMAP and FY-3B products. The generated long-term SM will benefit the global water cycle study.

Highly quality soil moisture is significant for hydrological, meteorological and agricultural applications. At present, active and passive remote sensing are the only ways to monitor soil moisture directly at regional scale. However, the quality of single satellite-based soil moisture product is insufficient to meet the requirements of these applications. Hence, fusing these two soil moisture products to improve their quality of change capture ability and accuracy is a necessary and challenging work. This study proposes an improved double instrumental variable method to fuse active and passive soil moisture products. First, the method is improved in finding the best instrumental variables in time series based on correlation coefficient. Second, fused weights of input soil moisture products are estimated using the improved method. Finally, fused soil moisture products are obtained with higher change capture ability and higher accuracy. The Tibetan Plateau was selected as the study area to test the algorithm using both of the Climate Change Initiative (CCI) active and passive soil moisture products from the European Space Agency (ESA). The ground validation results show that, compared with the original soil moisture products, the change capture ability, expressed by the correlation coefficient (R), and the accuracy, expressed by the unbiased root mean square deviation (ubRMSD), have been both improved by about 10% on average. This study indicates that the proposed fusion method can effectively improve the quality of soil moisture products to further understand the global changing water cycle.

Irrigation is critical to agricultural production in arid and semiarid regions, and it is imperative to map high-resolution irrigated area to improve water productivity. This study proposes a field-scale (30-m resolution) irrigated area mapping method based on soil moisture change detection using remote sensing data only. First, normalized soil moisture is obtained using the optical trapezoid model (OPTRAM) and then converted to soil water content. Next, individual irrigation events are identified in the time series of soil water content using threshold detection. Finally, irrigation events are accumulated over the time series, and then, the irrigated area map can be obtained. This method was tested using Google Earth Engine (GEE) to analyze remote sensing images and map irrigated areas in a typical arid and semiarid region called Hexi Corridor in northwestern China in the past 30 years.In situvalidation shows that this method has an accuracy close to 100%. The shortcoming of low recall is also overcome by long-term observations. An application of the proposed method shows that the irrigated cropland of Hexi Corridor has increased by 4840 km2(42.2%) over a 31-year time period (1990–2020). This field-scale irrigated area mapping method can improve the management of water resources.

The rapid development in network technology has resulted in the proliferation of Internet of Things (IoT). This trend has led to a widespread utilization of decentralized data and distributed computing power. While machine learning can benefit from the massive amount of IoT data, privacy concerns and communication costs have caused data silos. Although the adoption of blockchain and federated learning technologies addresses the security issues related to collusion attacks and privacy leakage in data sharing, the “free-rider attacks” and “model poisoning attacks” in the federated learning process require auditing of the training models one by one. However, that increases the communication cost of the entire training process. Hence, to address the problem of increased communication cost due to node security verification in the blockchain-based federated learning process, we propose a communication cost optimization method based on security evaluation. By studying the verification mechanism for useless or malicious nodes, we also introduce a double-layer aggregation model into the federated learning process by combining the competing voting verification methods and aggregation algorithms. The experimental comparisons verify that the proposed model effectively reduces the communication cost of the node security verification in the blockchain-based federated learning process.