Monitoring and modeling land surface dynamics in bermejo river basin, argentina: time series analysis of MODIS and AMSR-E data

BARRAZA, VERONICA; GRINGS, F.; PERNA, P.; SALVIA, M.; CARBAJO, A. E.; FERRAZZOLI, P.; KARSZENBAUM, H.

Conferencia; IGARSS; 2012

IEEE, ESA

In this paper, we show how MODIS NDVI and AMSR-E microwave vegetation indexes (MVI) data can be used to monitor land surface phenology in the Bermejo River Basin. For this purpose, the statistical nature of the study area’s NDVI and MVI time series was analyzed and several modeling strategies were tested. For NDVI widely known time series models were tested and discarded. Reasons for the unsuccessful modeling attempts with NDVI series were provided in terms of noise sources usually present in the study area (mostly clouds and geometrical effects). Finally, a second order polynomial model has proven to be successful, explaining a significant portion of data variations, without over-fitting. NDVI temporal variation curves show functional forms that originate from the general annual performance of land surface phenology. Using these functional forms, a classification scheme is proposed. We also explore the MVIs behavior within large areas of homogeneous NDVI functional forms. We analyzed MVIs temporal trend using NDVI seasonal patterns and assumptions about canopy structure that influence vegetation emissivity and opacity. We fit a second order polynomial to model the MVI index as a way to compare the information provided by both optical and microwave indexes. By analyzing their temporal correlation, it was possible to evaluate the annual and inter-annual ecosystem dynamics in terms of land surface phenology.