IrrMapper is an annual classification of irrigation status in the 11 Western United States made at Landsat scale (i.e., 30 m) using the Random Forest algorithm, covering years 1986 - present. While the IrrMapper paper describes classification of four classes (i.e., irrigated, dryland, uncultivated, wetland), the dataset is converted to a binary classification of irrigated and non-irrigated. 'Irrigated' refers to the detection of any irrigation during the year. The IrrMapper random forest model was trained using an extensive geospatial database of land cover from each of four irrigated- and non-irrigated classes, including over 50,000 human-verified irrigated fields, 38,000 dryland fields, and over 500,000 square kilometers of uncultivated lands.
IrrMapper是使用随机森林算法对美国西部11个州的灌溉状况进行的年度分类（即30米），涵盖1986年至今。虽然IrrMapper论文描述了四个等级的分类（即灌溉、旱地、非耕地、湿地），但数据集被转换为灌溉和非灌溉的二元分类。灌溉的 "指的是在这一年中检测到任何灌溉的情况。IrrMapper随机森林模型是使用一个广泛的地理空间数据库来训练的，该数据库包括来自四个灌溉和非灌溉类别的土地覆盖，包括50,000多块经人类验证的灌溉田，38,000块旱地，以及500,000多平方公里的未开垦土地。
Dataset Availability
1986-01-01T00:00:00 - 2020-01-01T00:00:00
Dataset Provider
University of Montana / Montana Climate Office
Collection Snippet
ee.ImageCollection("UMT/Climate/IrrMapper_RF/v1_0")
Resolution
30 meters
Bands Table
Name | Description |
|---|---|
classification | Irrigated pixels have the value of 1, the other pixels are masked out. |
引用：
Ketchum, D.; Jencso, K.; Maneta, M.P.; Melton, F.; Jones, M.O.; Huntington, J. IrrMapper: A Machine Learning Approach for High Resolution Mapping of Irrigated Agriculture Across the Western U.S.. Remote Sens. 2020, 12, 2328. doi:10.3390/rs12142328
代码：
代码语言：javascript复制var dataset = ee.ImageCollection("UMT/Climate/IrrMapper_RF/v1_0");
var irr_2013 = dataset.filterDate('2013-01-01', '2013-12-31').select('classification');
var visualization = {
min: 0.0,
max: 1.0,
palette: ['blue']
};
Map.addLayer(irr_2013, visualization, 'IrrMapper 2013');
Map.setCenter(-112.516, 45.262, 10);
