Using Google Earth Engine, to classify irrigated and non-irrigated lands in Idaho's Snake River Plain using Landsat. Classification is based on a binary classification of seasonal NDMI (normalized difference water or moisture index) (also called NDWI). Classifications are perfomed for each year using Landsat data (1984-2016). This could easily be adapted to other regions, if you wanted to. All you need is a training dataset for your region.
The basic workflow is as follows:
-
determine thresholds on seasonal Max NDMI for set control points:
A. Export values at control points. Code is here: /Irrigation_Extent/NDMI_Max_Points_3x3.js https://github.com/ericwchance/Irrigation_Extent/blob/master/NDMI_Max_Points_3x3.js
B. organize them into 2 columns of irrigated and non irrigated values.
C. Put that data into matlab to get a threshold for each year. Code is here: /Irrigation_Extent/Simplified_threshold_selection.m https://github.com/ericwchance/Irrigation_Extent/blob/master/Simplified_threshold_selection.m
-
Apply those yearly thresholds to each year's data. /Irrigation_Extent_County_Export.js
-
Export desired information, such as the extent of irrigation by county or water right.
The main file you want to look at here is the /Irrigation_Extent_County_Export.js https://github.com/ericwchance/Irrigation_Extent_Time_Series/blob/master/Irrigation_Extent_County_Export.js. The other files in here are just variations on that one, set to export different things.