SCALABLE TIME SERIES CHANGE DETECTION FOR BIOMASS MONITORING USING GAUSSIAN PROCESS
Varun Chandola, Ranga Raju Vatsavai · Scopus · 2010
Biomass monitoring, specifically, detecting changes in the biomass or vegetation of a geographical region, is vital for studying the carbon cycle of the system and has significant implications in the context of understanding climate change and its impacts. Recently, several time series change detection methods have been proposed to identify land cover changes in temporal profiles (time series) of vegetation collected using remote sensing instruments. In this paper, we adapt Gaussian process regression to detect changes in such time series in an online fashion. While Gaussian process (GP) has been widely used as a kernel based learning method for regression and classification, their applicability to massive spatio-temporal data sets, such as remote sensing data, has been limited owing to the high computational costs involved. In our previous work we proposed an efficient Toeplitz matrix based solution for scalable GP parameter estimation. In this paper we apply these solutions to a GP based change detection algorithm. The proposed change detection algorithm requires a memory footprint which is linear in the length of the input time series and runs in time which is quadratic to the length of the input time series. Experimental results show that both serial and parallel implementations of our proposed method achieve significant speedups over the serial implementation. Finally, we demonstrate the effectiveness of the proposed change detection method in identifying changes in Normalized Difference Vegetation Index (NDVI) data. Increasing availability of high resolution remote sensing data has encouraged researchers to extract knowledge from these massive spatio-temporal data sets in order to solve different problems pertain- ing to our ecosystem. Land use land cover (LULC) monitoring, specifically identifying changes in land cover, is one such problem that has significant applications in detecting deforestation, crop ro- tation, urbanization, forest fires, and other such phenomenon. The knowledge about the land cover changes can then be used by policy makers to take important decisions regarding urban planning, natural resource management, water source management, etc. In this paper we focus on the problem of identifying changes in the biomass or vegetation in a geographical region. Biomass is defined as the mass of living biological organisms in a unit area. In the context of this study, we restrict our monitoring to plant (specifically crop) biomass over large geographic regions. In recent years biomass monitoring is increasingly becoming important, as biomass is a great source of renewable energy. Moreover, biomass monitoring is also important from the changing climate perspective, as changes in climate are reflected in the change in biomass, and vice versa. The knowledge about biomass changes over time across a geographical region can be used estimate quantitative biophysical parameters which can be incorporated into global climate models. The launch of NASA's Terra satellite in December of 1999, with the Moderate Resolution Imag- ing Spectroradiometer (MODIS) instrument aboard, introduced a new opportunity for terrestrial remote sensing. MODIS data sets represent a new and improved capability for terrestrial satel- lite remote sensing aimed at meeting the needs of global change research. With thirty-six spectral bands, seven designed for use in terrestrial application, MODIS provides daily coverage, of moderate spatial resolution, of most areas on the earth. Land cover products are available in 250m, 500m, or 1000m resolutions (17). MODIS land products are generally available within weeks or even days