An analysis of the effects of ground parameters and multitemporal compositing techniques in the passive microwave vegetation index.

摘要

The response of the Passive Microwave Vegetation Index (MVI) in the State of California is determined. Five different methods are used to investigate MVI. First, the effects of different multi-temporal compositing techniques on MVI are determined. Second, the correlation between MVI and four parameters; water bodies, precipitation, ground cover, and the Normalized Difference Vegetation Index (NDVI), are determined. Third, MVI values are predicted using the four parameters. Fourth, NDVI values are predicted using MVI as the independent variable. Fifth, MVI, precipitation, and NDVI multi-temporal trajectories are compared at three study sites in California.; Two different multi-temporal compositing techniques were used. The two methods were the Marshall Space Flight Center (MSFC) technique using the average value for a pixel, and the Choudhury method using the second lowest value for a pixel. Three datasets were produced. One dataset using the MSFC technique was created. One dataset using solely ascending (morning) satellite overpass data was created using the Choudhury technique. One dataset using solely descending (evening) satellite overpass data was created using the Choudhury technique. Correlations between the three datasets were on the order of 0.85 to 0.95. ANOVA tests showed the three datasets to be statistically different.; The correlation between MVI and water bodies was approximately 0.2 to 0.5. The correlation between MVI and precipitation was approximately {dollar}-{dollar}0.38 to {dollar}-{dollar}0.41. The correlation between MVI and NDVI was approximately {dollar}-{dollar}0.69. The relationship between MVI and ground cover was determined by splitting ground cover into five categories based on gross vegetation morphology. The five categories were forest, grassland, scrub, shrub, and woodland. Representative MVI values for the five categories were forest 2.5, grassland 7.6, scrub 15.7, shrub 6.2, and woodland 4.9.; Eight independent variables could have been used to predict MVI. The eight variables were forest, grassland, scrub, shrub, woodland, water bodies, precipitation and NDVI. The optimal subset of predictor variables was forest, woodland, and NDVI. Using these three variables the correlation between predicted and actual MVI values was approximately 0.7.; NDVI values were predicted using MVI values. A quadratic equation was found to produce the optimal result. The correlation between predicted and actual NDVI values was approximately 0.37.; MVI, precipitation, and NDVI multi-temporal trajectories were compared at three study sites in California. The three sites were representative of desert, chaparral, and agricultural vegetation.