Prospects for Using Pseudocolor Image Processing in the Analysis of Multiyear Time Series of Satellite Data in the Task of Assessing the State of Vegetation Cover

Satellite images and the Normalized Difference Vegetation Index (NDVI) are often used to monitor the state of vegetation cover and a significant amount of information has now been accumulated. When processing long-term time series of satellite data, mathematical difficulties arise, for example, in data clustering. The solution to the problem can be a parameterization of the satellite monitoring information using characteristic moments. For each pixel position, hundreds of NDVI values from satellite data can be reduced to several characteristic functional parameters, in particular to the extreme NDVI and the average multiyear NDVI value, as well as other summing characteristics. This opens the way for constructing pseudocolor composite images and their subsequent clustering using any standard algorithms. This research examined Southern Kazakhstan, with a total area of more than 700 thousand km². Using Google Earth Engine, time series of NDVI from Sentinel-2 scenes (resolution 10 m) for the period April–October 2018–2022 (about 160 covers) served as the basis for describing the state of vegetation cover. An additional parameter was the multiyear maximum of the Vegetation Soil Salinity Index (VSSI). The RGB channels of the pseudocolor image were based on Sentinel-2 monitoring data from April–October 2017–2022 and included Red, a multiyear maximum of VSSI; Green, a multiyear maximum of NDVI; and Blue, a multiyear average of NDVI. The resulting pseudocolor image displayed in detail the state of vegetation, with a clear separation of agricultural vegetation from natural, with a ranking of irrigated arable land according to the features of growth and development of agricultural crops. This information can serve as a basis for segmentation of preprocessed satellite data for analyzing the vegetation state in South Kazakhstan in various applied tasks. As an example, using unsupervised ISODATA classification, salinity of irrigated arable land of the Kyzylkum rural district of the Zhetysay district of Turkestan region was estimated. The results demonstrated the prospects of such an analysis method and clarified the known results obtained earlier using MODIS satellite data.