"Everything should be made as simple as possible, but not simpler." - Albert Einstein

Seminar Announcement

PREDICTING SEASONAL ENHANCED VEGETATION INDEX (EVI) CURVES AND ASSOCIATED UNCERTAINTY USING A TWO-STAGE REGRESSION MODEL FITTING PROCEDURE

Ram Gurung , M.S. Candidate, Department of Statistics, Colorado State University

Tuesday, June10, 2008

1 p.m. Weber 223

ABSTRACT
The Enhanced Vegetation Index (EVI) is a numerical measure of the relative
greenness of vegetation. It is computed from remote sensing data collected
since 2000 by NASA’s satellites Terra and Aqua using Moderate Resolution
Imaging Spectroradiometer (MODIS).  EVI is smooth and seasonal, and varies
site by site and year by year. The purpose of this paper is to describe the
relationship between EVI and climate indexes (temperature and precipitation),
crop information and irrigation information using statistical tools. We use
time series data from 2000 to 2003 for farmland National Resources Inventory
(NRI) points of the mid-continent spanning ten states in the US. A two-stage
multiple regression fitting procedure within a semi-parametric mixed effect
(SPME) model framework was used to explain the seasonal EVI curves and to
predict past years, future years and completely new locations. First, a linear
mixed effect (LME) model was fitted to the EVI with the climate indexes, crop
information and irrigation information as predictor variables. Second,
Penalized B-splines were used to explain the behavior of the smooth residuals, which result from a smooth model fit to the smooth data. This allows us to describe the uncertainty of the EVI curve. Individual models were fitted to individual Major Land Resources Areas (MLRAs). Predicted seasonal EVI, derived from our regression equations, shows a strong agreement with the observed EVI and is able to capture the site by site and year by year variation in the EVI curve. Out-of-sample prediction is generally excellent, except on sites without clear seasonal patterns, which may be caused by disturbance due to cloud contamination and/or snow cover.