Historical land cover and paddy rice mapping for Northwest Bangladesh 1989 to 2016

High resolution (30 m) land cover and cropping maps in GeoTIFF format for two main rice types in northern Bangladesh, dry season Boro rice (January to May) and wet season Aman rice (October to January) for the cropping seasons of 1989–1990 to 2015–2016. Other land cover types include other vegetated type, water, water non-permanent, and bare. The values in the Boro season are as follows: 10 represents Boro, 11 and 13 represent other vegetated areas, 14 represents water, 15 represents water non-permanent and 16 represents bare. The values in the Aman season are as follows: 20 represents Aman, 23 represents other vegetated areas, 24 represents water, 25 represents water non-permanent and 26 represents bare. Value 0 is a null value in both rice season maps. Associated GeoTIFF maps show the number of months missing in each pixel per mapping season per cropping year (using the unfilled monthly composite images) as a guide for quality. \nLineage: The data used to produce the maps encompassed nearly three decades of Landsat TM/ETM+/OLI TOA reflectance data from several satellite platforms, sourced and pre-processed through the freely available petabyte archive and geostatistical processing power of Google Earth Engine. Geospatial techniques were used to reduce gaps in the data. A combination of unsupervised K-means clustering and supervised Random Forest Machine Learning algorithms were implemented to produce a predictive model that includes vegetation indices and other covariates, which explain the phenology of different land cover types.

本数据集包含孟加拉国北部两种主要水稻类型的高分辨率（30米）土地覆盖与种植分布图，格式为GeoTIFF，覆盖1989–1990至2015–2016年间的两个种植季：旱季布里（Boro）水稻（1月至5月）与雨季阿曼（Aman）水稻（10月至次年1月）。其他土地覆盖类型包括其他植被类、水体、季节性水体与裸地。布里季的像素取值规则如下：10代表布里水稻，11与13代表其他植被区域，14代表水体，15代表季节性水体，16代表裸地。阿曼季的像素取值规则如下：20代表阿曼水稻，23代表其他植被区域，24代表水体，25代表季节性水体，26代表裸地。两个水稻种植季的地图中，值0均代表空值。配套的GeoTIFF图层展示了每个种植年、每个制图季中，各像素的缺失月份数量（基于未填充的月度合成影像），可作为数据质量评估参考。

数据谱系：本数据集所依托的原始数据涵盖近三十年来自多颗卫星平台的Landsat TM/ETM+/OLI大气顶层（TOA）反射率数据，通过谷歌地球引擎（Google Earth Engine）免费开放的拍字节级存档与地理统计处理能力完成数据获取与预处理。研究采用地理空间技术填补数据间隙，并结合无监督K-means聚类与有监督随机森林机器学习算法，构建包含植被指数与其他协变量的预测模型，以此阐释不同土地覆盖类型的物候特征，最终生成上述土地覆盖与种植分布图。