Crop Mapping using Remote Sensing
Crop type mapping at the field level is necessary for a variety of applications in agricultural monitoring and water resources management. Remote sensing imagery is a powerful input from which crop type maps can be created. hydrosolutions ltd is currently designing various apps that facilitate users with no or minimal background in remote sensing to access and use such data for their purposes. Here we present a new browser application that we have designed to generate high-resolution annual maps of crop types and cropping patterns in different irrigation systems in the global drylands.
The methodology for crop type mapping that we have implemented in our tool is based on an unsupervised classification technique proposed and evaluated by Wang et al. 2019. It uses harmonic regression in Google Earth Engine to extract features from time series of optical remote sensing data. The features are then used to cluster the pixels, i.e. to group pixels with similar features into different classes. The final step is to label the classes, which means that we must attribute a crop type to every cluster of pixels. This last step is not automatized, and we rely on the expert knowledge of the local users to label the clusters with the main crop types that are growing in his region of interest. To facilitate this task, users can display the characteristic time series of vegetation indices, or they can identify the crop type by looking at the provided high-resolution satellite imagery. The advantage of our method is that it is not constrained by the common lack of field-level crop labels for training and can therefore theoretically be applied anywhere on the planet where optical satellite imagery is available. Clouds and haze are the methods’ main adversaries, and it works best in arid or semi-arid regions with distinct seasons and crop growth calendars.
More information and access to a sample application for crop type mapping in different areas of interest can be obtained on the dedicated application site.