Tracking cropland change and its spatiotemporal characteristics can provide a scientific basis for assessments of ecological restoration in reclamation areas. In 1998, an ecological restoration project (Converting Farmland to Lake) was launched in Dongting Lake, China, in which original lake areas reclaimed for cropland were converted back to lake or to poplar cultivation areas. This study characterized the resulting long-term (1998–2018) change patterns using the LandTrendr algorithm with Landsat time-series data derived from the Google Earth Engine (GEE). Of the total cropland affected, ~447.48 km² was converted to lake and 499.9 km² was converted to poplar cultivation, with overall accuracies of 87.0% and 83.8%, respectively. The former covered a wider range, mainly distributed in the area surrounding Datong Lake, while the latter was more clustered in North and West Dongting Lake. Our methods based on GEE captured cropland change information efficiently, providing data (raster maps, yearly data, and change attributes) that can assist researchers and managers in gaining a better understanding of environmental influences related to the ongoing conversion efforts in this region.

Land cover changes, especially excessive economic forest plantations, have significantly threatened the ecological security of West Dongting Lake wetland in China. This work aimed to investigate the spatiotemporal dynamics of forests in the West Dongting Lake region from 2000 to 2018 using a reconstructed monthly Landsat NDVI time series. The multi-type forest changes, including conversion from forest to another land cover category, conversion from another land cover category to forest, and conversion from forest to forest (such as flooding and replantation post-deforestation), and land cover categories before and after change were effectively detected by integrating Breaks For Additive Seasonal and Trend (BFAST) and random forest algorithms with the monthly NDVI time series, with an overall accuracy of 87.8%. On the basis of focusing on all the forest regions extracted through creating a forest mask for each image in time series and merging these to produce an ‘anytime’ forest mask, the spatiotemporal dynamics of forest were analyzed on the basis of the acquired information of multi-type forest changes and classification. The forests are principally distributed in the core zone of West Donting Lake surrounding the water body and the southwestern mountains. The forest changes in the core zone and low elevation region are prevalent and frequent. The variation of forest areas in West Dongting Lake experienced three steps: rapid expansion of forest plantation from 2000 to 2005, relatively steady from 2006 to 2011, and continuous decline since 2011, mainly caused by anthropogenic factors, such as government policies and economic profits. This study demonstrated the applicability of the integrated BFAST method to detect multi-type forest changes by using dense Landsat time series in the subtropical wetland ecosystem with low data availability.

In recent years, deep-learning-based hyperspectral image (HSI) classification networks have become one of the most dominant implementations in HSI classification tasks. Among these networks, convolutional neural networks (CNNs) and attention-based networks have prevailed over other HSI classification networks. While convolutional neural networks with perceptual fields can effectively extract local features in the spatial dimension of HSI, they are poor at capturing the global and sequential features of spectral–spatial information; networks based on attention mechanisms, for example, Transformer, usually have better ability to capture global features, but are relatively weak in discriminating local features. This paper proposes a fusion network of convolution and Transformer for HSI classification, known as FusionNet, in which convolution and Transformer are fused in both serial and parallel mechanisms to achieve the full utilization of HSI features. Experimental results demonstrate that the proposed network has superior classification results compared to previous similar networks, and performs relatively well even on a small amount of training data.

Accurate and timely information about rice planting areas is essential for crop yield estimation, global climate change and agricultural resource management. In this study, we present a novel pixel-level classification approach that uses convolutional neural network (CNN) model to extract the features of enhanced vegetation index (EVI) time series curve for classification. The goal is to explore the practicability of deep learning techniques for rice recognition in complex landscape regions, where rice is easily confused with the surroundings, by using mid-resolution remote sensing images. A transfer learning strategy is utilized to fine tune a pre-trained CNN model and obtain the temporal features of the EVI curve. Support vector machine (SVM), a traditional machine learning approach, is also implemented in the experiment. Finally, we evaluate the accuracy of the two models. Results show that our model performs better than SVM, with the overall accuracies being 93.60% and 91.05%, respectively. Therefore, this technique is appropriate for estimating rice planting areas in southern China on the basis of a pre-trained CNN model by using time series data. And more opportunity and potential can be found for crop classification by remote sensing and deep learning technique in the future study.