Sentinel 2 image RGB:843 Sentinel 2 image RDVI

Biophysical variables and land cover products

GEOMATRIX UAB has developed automated work-flow for computing standard vegetation indexes and bio-physical parameters (BPP), which we use in various agricultural, hydrological and environmental applications. We use BPPs for fast extraction of the main  land cover classes and monitoring of landscape dynamics within a time-series of high-resolution satellite images provided by the European Space Agency (ESA) for implementation of Copernicus Land Monitoring Services and Downstream Services. Automated production work-flow was designed for creation of products containing aggregated maximum values detected in a time-series of satellite images over a selected area of interest. Analysis of several on-line references on remote sensing indexes and IDB database in particular pointed out a group of BPPs which seem to be most useful in terms practical applications, however automated processing work-flow allows computing of any BPP product as long as satellite data provides the required spectral bands.

For agricultural monitoring applications we use Sentinel 2 MS L2A products (top left example shows S2 image taken in Lithuania in August 2015) for computing Renormalized Difference Vegetation Index (RDVI, top-right image) which has a high correlation with a very accurate but mostly empirical Leaf Area Index (LAI) parameter, along with Vegetation Moisture Index (mid right image) and Vegetation Stress Index (mid left image) based on SWIR spectral band. Those biophysical indexes reflect status of vegetation in terms of a shifting balance of NIR/SWIR ratio from high-NIR indicating fast and "healthy" growth towards high-SWIR indicating low water content related to vegetation "aging" or stress.

Biophysical indexes provide continuous coverages of standard numerical expressions related to certain vegetation, soil and wetness conditions, therefore they can be used (alone or in combination with each other) for automated creation of certain "pre-classified" land cover products, as shown in the bottom line example image from northern Italy. Biophysical land cover thematic layers have no thematic attributes, but they contain narrow ranges of BPP values systematically indicating presence of certain land cover classes. In the example pre-classified image we can easily notice water bodies, grassland, crops, different forest types, etc. Although BPP classification alone can not detect all potentially existing land cover classes, those products nevertheless can be used rather efficiently for the majority of natural habitats and - most importantly - they enable fast and systematic detection and mapping of natural and rural landscape changes within a time series of MS images. Land cover classes can be aggregated directly from BPP values in a fast and efficient way (i.e. without complicated and demanding object-based classification), while change detection is done by the use of time-series statistical analysis ("permanent" vs "temporary" pixels of a certain class) or direct raster algebra operations applied incrementally.

Sentinel 2 image VSI Sentinel 2 image VMI
Pre-classified BPP land cover