NOTE from WRI: WRI decided to stop updating FORMA alerts. The goal was to simplify the Global Forest Watch user experience and reduce redundancy. We found that Terra-i and GLAD were more frequently used. Moreover, using GLAD as a standard, found that Terra-i outperformed FORMA globally.
FORMA alerts are detected using a combination of two MODIS products: NDVI (Normalized Difference Vegetation Index) and FIRMS (Fires Information for Resource Management System). NDVI updates are processed every 16 days, while fire updates are processed daily. Models are developed individually for each ecogroup to relate the two inputs to the area of clearing, using the Hansen annual tree cover loss data to train the model. The minimum threshold to qualify as an alert is 25% of the pixel cleared, though thresholds vary by ecogroup to minimize false positives. Here is an example script for a quick introduction to the FORMA datasets.
The images in this ImageCollection contain the “reversed rectified t-statistics” used in calculating NTT, the vegetation color index derived from MODIS NDVI that FORMA uses to measure browning. Using a sum reducer on over various date ranges in this ImageCollection produces an “NTT” image.
The images are broken by "ecogroup". Ecogroup geometries can be found here.
WRI的注意：WRI决定停止更新FORMA警报。其目的是简化全球森林观测的用户体验，减少冗余。我们发现，Terra-i和GLAD的使用频率更高。此外，以GLAD为标准，发现Terra-i在全球的表现优于FORMA。
FORMA警报是使用两个MODIS产品的组合来检测的。NDVI（归一化植被指数）和FIRMS（资源管理系统的火灾信息）。NDVI更新每16天处理一次，而火灾更新则每天处理一次。为每个生态组单独开发模型，将这两个输入与清理区域联系起来，使用汉森年度树木覆盖损失数据来训练模型。有资格成为警报的最低阈值是25%的像素被清除，尽管阈值因生态组而异，以尽量减少假阳性。下面是一个快速介绍FORMA数据集的脚本示例。
该图像集中的图像包含用于计算NTT的 "反转校正t统计"，NTT是由MODIS NDVI得出的植被颜色指数，FORMA用来测量褐化。在该图像集的各种日期范围内使用总和还原器，产生一个 "NTT "图像。
这些图像按 "生态组 "划分。生态组的几何形状可以在这里找到。
Dataset Availability
2012-01-01T00:00:00 - 2019-04-23T00:00:00
Dataset Provider
World Resources Institute / Global Forest Watch
Collection Snippet
ee.ImageCollection("WRI/GFW/FORMA/vegetation_tstats")
代码：
代码语言：javascript复制var dataset = ee.ImageCollection('WRI/GFW/FORMA/vegetation_tstats')
.filter(ee.Filter.date('2018-07-01', '2018-07-15'));
var tstat = dataset.select('tstat_r');
var visParams = {
min: 0.0,
max: 1.0,
};
Map.setCenter(25.73, -7.61, 2);
Map.addLayer(tstat, visParams, 'Reversed rectified t-statistics');
