Vegetation Index Time Series Analysis for environment monitoring. Examples from the MODIS NDVI time series (2000-2022) on Sri Lanka (Keynote Speech)

Abstract

Remote sensing is a powerful tool for environment monitoring for because most of the remotely sensed data is intercomparable in space and time. In other terms a correct analysis of remotely sensed data enables to detect the same objects in different places (therefore, to map all the objects of a region) (Gosh et al., 2022) and same objects in different periods (therefore, to map changes) (Andrieu, 2017; Andrieu, 2018, Andrieu et al., 2019). Remote sensing is progressively changing from a little number of data sets with low to moderate resolution (spatial and temporal) (Andrieu, 2017; Andrieu et al 2019) to a large number of data sets of high to very high resolution (Andrieu, 2018; Gosh et al., 2022). Intercomparability has somehow been the “victim” of the target put on resolution. Mapping changes in land cover, or in vegetation, in a country like Sri Lanka with data as LANDSAT time series would suffer from 2 main weaknesses. First, a lot of different scenes are necessary, creating errors on the edges between two scenes. then most of the images contain important cloud cover. It’s even worse with Very High-resolution imagery either the expensive one or the free one accessible on Google Earth. For these reasons this conference will illustrate the potential of a very different kind of data, rarely used on Sri Lanka, and however having a high potential to monitor environment. The Very high intercomparability of NDVI and the interesting temporal resolution (16 days) of the MOD13Q1 dataset indeed offers the possibility of a very accurate bioclimatic mapping a robust capacity to detect changes in land cover and vegetation such applications will be presented. Some statistical methods will be presented with their applications to Sri Lanka.