This study aims to develop a deep learning based classification framework for remotely sensed time series. The experiment was carried out in Yolo County, California, which has a very diverse irrigated agricultural system dominated by economic crops. For the challenging task of classifying summer crops using Landsat Enhanced Vegetation Index (EVI) time series, two types of deep learning models were designed: one is based on Long Short-Term Memory (LSTM), and the other is based on one-dimensional convolutional (Conv1D) layers. Three widely-used classifiers were also tested for comparison, including a gradient boosting machine called XGBoost, Random Forest, and Support Vector Machine. Although LSTM is widely used for sequential data representation, in this study its accuracy (82.41%) and F1 score (0.67) were the lowest among all the classifiers. Among non-deep-learning classifiers, XGBoost achieved the best result with 84.17% accuracy and an F1 score of 0.69. The highest accuracy (85.54%) and F1 score (0.73) were achieved by the Conv1D-based model, which mainly consists of a stack of Conv1D layers and an inception module. The behavior of the Conv1D-based model was inspected by visualizing the activation on different layers. The model employs EVI time series by examining shapes at various scales in a hierarchical manner. Lower Conv1D layers of the optimized model capture small scale temporal variations, while upper layers focus on overall seasonal patterns. Conv1D layers were used as an embedded multi-level feature extractor in the classification model which automatically extracts features from input time series during training. The automated feature extraction reduces the dependency on manual feature engineering and pre-defined equations of crop growing cycles. This study shows that the Conv1D-based deep learning framework provides an effective and efficient method of time series representation in multi-temporal classification tasks.

农作物遥感识别是地理学和生态学研究的前沿和热点，多源数据在农作遥感识别中日益发挥重要作用。笔者从多源数据融合的角度，归纳了2000年后多源数据在农作物遥感识别中应用的总体概况，系统梳理并提炼了当前多源数据融合的主要融合技术和融合模式。围绕与多源数据融合和农作物遥感识别相关的关键词，在Google学术、ISI Web of Knowledge和中国知网中对2000—2014年间国内外发表的论文进行检索，并统计不同传感器的使用频率及结合方式。研究表明，以提高空间分辨率为目标的多源数据融合和以提高时间分辨率为目标的多源数据融合技术是当前的两种主要方式，可以在一定程度上实现时空尺度的扩展。前者的融合技术包括图像融合、正态模糊分布神经网络模型、成分替换、半经验数据模型融合及多分辨率小波分解等，可以提升遥感数据的空间分解力和清晰度，较好弱化混合像元产生的影响，但农作物光谱信息有一定程度的丢失或扭曲，农作物空间分布局部细节信息与纹理特征依然会缺失；后者的融合技术形式灵活多样，可分为同源数据联合扩展时序的时空优化技术和异源数据联合扩展时序的时空优化技术，其可以有效排除短时间段内农作物生育期交叉，但易受不同遥感数据源间光谱反射率或植被指数转换模型及光谱波段设置差异的影响。在融合模式方面，根据数据类型分为光学数据的融合、光学数据与微波数据的融合以及遥感与非遥感数据的融合，以实现卫星资源优势互补为宗旨，充分挖掘不同类型农作物在遥感数据上呈现的光谱、时间和空间特征差异信息。同样，农作物遥感识别研究中的多源遥感数据融合也存在诸多挑战，在未来一段时间内，完善不同传感器之间的合作、更深层次挖掘融合信息以及多尺度长时间序列的中高分辨率农作物空间分布数据集的需求是多源数据融合的农作物遥感识别研究的重点发展方向和亟待解决的问题。研究结果有助于更好地理解多源遥感数据融合的技术和模式，为摸清多源数据融合在农作物识别中总体进展提供支撑，同时也为其他多源数据融合研究提供借鉴。

Crop mapping by using the remotely-sensed images provide basic information for further geographical and ecological researches. A systematic review on the recent literature regarding crop mapping was carried out in order to improve our understanding on the integration and application of multi-source remote sensing data. The literature search was performed in Google Scholar, the ISI Web of Knowledge and CNKI (e.g. Topic =”crop + mapping”; Topic =”classification + multi-source”; timespan = 2000-2014). According to the thorough analysis on the existing publications, it is suggested that (1) there are two main ways to identify crop types based on the integration of multi-source data in order to expand the spatial and temporal scales. The techniques of multi-source data fusion, which are aimed at improving the spatial resolution, include image fusion, normal fuzzy distributed neural networks, component substitution, semi-physical fusion approach, and multiresolution wavelet decomposition. With the integrated application of such approaches, the spatial resolution and clarity of remote sensing images are raised; the effect of mixed pixels is weaken to some extent. Nevertheless, crop spectral information is partly lost or distorted. The techniques of multi-source data fusion, which are aimed at improving the temporal resolution, can be categorized into two types: the integration of the same data source, and the integration of different data sources. By using such approaches, the crossover of growth period among different crops can be effectively eliminated. But such approaches are susceptible to transformation models of spectral reflectance or vegetation indices, and the differences in band coverage among different remote sensing data. (2) The modes of multi-source data fusion can be categorized into three types according to the data types applied: integration of optical data, integration of optical and microwave data, and integration of remote sensing and ancillary data sources. Taking complementary advantages of various satellite data resources, these techniques of data fusion fully mine the differences of spectral, temporal and spatial characteristics, among various crop species. However, there still remain challenges in previous researches about the crop identification based on the fusion of multi-source remotely sensed data.

农作物种植结构信息对农业生产管理、农业可持续发展及国家粮食安全等具有重要意义。本文中概括了农作物种植结构遥感提取的理论基础，归类了近10年间不同农作物种植结构遥感提取技术方法，重点评述了不同技术方法的特点及应用情况，讨论和展望了未来农作物种植结构遥感提取研究的发展方向。当前，光谱特征、时相特征和空间特征是农作物种植结构遥感提取的三大理论基础。基于单一影像源的种植结构提取方法操作简单，但往往难以获取种植结构“最佳识别期”的遥感影像；基于多时序影像源的种植结构提取方法可以充分利用农作物季相节律特征，成为当前农作物种植结构遥感提取的主流方法。在基于多时序影像源的种植结构提取方法中，多特征参量法较单一特征参量法更适用于农作物种植结构复杂区域，基于多特征参量的统计模型法一定程度上解决了混合像元问题，但模型的鲁棒性有待提高。此外，遥感与统计数据融合的农作物种植结构提取法在国家及全球大尺度的农作物种植结构提取中具有优势，但较低的制图分辨率使得数据产品的区域适宜性较差。未来农作物种植结构遥感提取将以区域“作物一张图”为目标，充分发挥多源数据组合利用的优势，围绕多类型作物同步提取和大范围作物种植结构提取开展深入研究，重点加强遥感数据预处理、特征参量提取和分类器高效选择等关键技术研究，从而提升农作物种植结构遥感提取的时空尺度，满足多方位的农业应用需求。

Mapping crop patterns with remote sensing is of great implications for agricultural production, food security and agricultural sustainability. In this paper, the theoretical basis behind the mapping was summarized, mapping methods were classified into several categories, characteristics and applicabilities of different mapping methods in the latest decade were discussed intensively, and some important directions and priorities for future studies were proposed. Currently, spectral, temporal and spatial features are the major theoretical bases for crop pattern mapping. The mapping method based on single imagery is characterized by its simple implementation, but with difficulty of capturing imagery at the best time for distinguishing different crops. Instead, the mapping method based on time-series of imagery can make full use of temporal features and is thus widely used for crop mapping, among which the methods using multiple features are more suitable than the ones using a single feature for regions with complicated planting structure. To some extent, feature-oriented statistical modeling method can resolve the mixed-pixel problem but its robustness needs to be improved. Furthermore, large-scale crop pattern mapping can be done by combining the remote sensing and agriculture statistics. However, due to coarse resolution, the derived maps show poor region suitability. Future crop pattern mapping should target at developing “a map of crops”, the emphasis must be put on covering more crop types, enlarging the mapping areas, utilizing the superiority of blending multi-source data, strengthening the data preprocessing, optimizing the feature extraction and classifier selection, and improving the temporal and spatial scales of crop pattern mapping so as to better meet the needs of multi-faceted agricultural applications.

Area and spatial distribution information of paddy rice are important for understanding of food security, water use, greenhouse gas emission, and disease transmission. Due to climatic warming and increasing food demand, paddy rice has been expanding rapidly in high latitude areas in the last decade, particularly in northeastern (NE) Asia. Current knowledge about paddy rice fields in these cold regions is limited. The phenology- and pixel-based paddy rice mapping (PPPM) algorithm, which identifies the flooding signals in the rice transplanting phase, has been effectively applied in tropical areas, but has not been tested at large scale of cold regions yet. Despite the effects from more snow/ice, paddy rice mapping in high latitude areas is assumed to be more encouraging due to less clouds, lower cropping intensity, and more observations from Landsat sidelaps. Moreover, the enhanced temporal and geographic coverage from Landsat 8 provides an opportunity to acquire phenology information and map paddy rice. This study evaluated the potential of Landsat 8 images on annual paddy rice mapping in NE Asia which was dominated by single cropping system, including Japan, North Korea, South Korea, and NE China. The cloud computing approach was used to process all the available Landsat 8 imagery in 2014 (143 path/rows, similar to 3290 scenes) with the Google Earth Engine (GEE) platform. The results indicated that the Landsat 8, GEE, and improved PPPM algorithm can effectively support the yearly mapping of paddy rice in NE Asia. The resultant paddy rice map has a high accuracy with the producer (user) accuracy of 73% (92%), based on the validation using very high resolution images and intensive field photos. Geographic characteristics of paddy rice distribution were analyzed from aspects of country, elevation, latitude, and climate. The resultant 30-m paddy rice map is expected to provide unprecedented details about the area, spatial distribution, and landscape pattern of paddy rice fields in NE Asia, which will contribute to food security assessment, water resource management, estimation of greenhouse gas emissions, and disease control. (C) 2016 Elsevier Inc.

. Large-scale, high-resolution maps of rapeseed (Brassica napus L.), a\nmajor oilseed crop, are critical for predicting annual production and\nensuring global energy security, but such maps are still not freely\navailable for many areas. In this study, we developed a new pixel- and\nphenology-based algorithm and produced a new data product for rapeseed\nplanting areas (2017–2019) in 33 countries at 10 m spatial resolution based\non multiple data. Our product is strongly consistent at the national level\nwith official statistics of the Food and Agricultural Organization of the\nUnited Nations. Our rapeseed maps achieved F1 spatial consistency scores of\nat least 0.81 when compared with the Cropland Data Layer in the United\nStates, the Annual Crop Inventory in Canada, the Crop Map of England, and\nthe Land Cover Map of France. Moreover, F1 scores based on independent\nvalidation samples ranged from 0.84 to 0.91, implying a good consistency\nwith ground truth. In almost all countries covered in this study, the\nrapeseed crop rotation interval was at least 2 years. Our derived maps\nsuggest, with reasonable accuracy, the robustness of the algorithm in\nidentifying rapeseed over large regions with various climates and\nlandscapes. Scientists and local growers can use the freely downloadable\nderived rapeseed planting areas to help predict rapeseed production and\noptimize planting structures. The product is publicly available at\nhttps://doi.org/10.17632/ydf3m7pd4j.3 (Han et al., 2021).\n

. Sugarcane is the most important source of sugar, and its\ncultivation area has undergone rapid expansion, replacing other crops,\npastures, and forests. Brazil is the world's largest sugarcane producer and\ncontributed to approximately 38.6 % of the world's total production in\n2019. Sugarcane in Brazil can be harvested from April to December in\nthe south-central area and from September to April in the northeast area. The\nflexible phenology and harvest conditions of sugarcane in Brazil make it\ndifficult to identify the harvest area at state to country scales. In this\nstudy, we developed a phenology-based method to identify the harvest area of\nsugarcane in Brazil by incorporating the multiple phenology conditions into\na time-weighted dynamic time warping method (TWDTW). Then, we produced\nannual 30 m spatial resolution sugarcane harvest maps (2016–2019) for 14\nstates in Brazil (over 98 % of the harvest area) based on the proposed\nmethod using Landsat-7, Landsat-8, and Sentinel-2 optical data. The proposed method\nperformed well in identifying sugarcane harvest area with limited training\nsample data. Validations for the 2018 harvest year displayed high accuracy,\nwith the user's, producer's, and overall accuracies of 94.35 %, 87.04 %, and\n91.47 % in Brazil, respectively. In addition, the identified harvest area\nof sugarcane exhibited good correlations with the agricultural statistical\ndata provided by the Brazilian Institute of Geography and Statistics (IBGE)\nat the municipality, microregion, and mesoregion levels. The 30 m Brazil\nsugarcane harvest maps can be obtained at\nhttps://doi.org/10.6084/m9.figshare.14213909 (Zheng et al., 2021).\n

【目的】当前对涉及到耕地内部不同作物空间分布及其变化的研究较少。本文旨在探讨大尺度作物种植面积和分布格局遥感提取方法及景观生态学中景观格局指数在作物格局动态变化分析中的应用。【方法】基于2005年和2010年作物生育期内遥感影像全覆盖的MODIS-NDVI数据，利用RS、GIS技术，通过分析东北地区主要作物（水稻、玉米、大豆）的种植结构、物候历及NDVI曲线特征，建立不同作物面积遥感提取模型，提取大尺度农作物空间分布格局信息。同时，利用景观格局指数方法分析农作物格局动态变化特征和变化规律。【结果】与多年平均统计数据比较，基于MODIS遥感数据提取的作物面积信息，2005年和2010年平均精度达到了90%以上；5年间，东北地区主要作物种植结构发生了较大变化。其中大豆平均斑块面积减少，面积年动态度为-4.47%，水稻和玉米平均斑块面积均增加，且5年的变化幅度均超过20%。【结论】成本和收益是作物面积增加或减少的主要原因；用中等分辨率的遥感数据进行大尺度作物面积提取的方法是可行的；景观生态学中格局指数可以用来分析耕地内部作物格局的动态变化规律。