Development of visible/infrared/microwave agricultural classification and biomass algorithms

Abstract

Due to inadequate crop acreage and biomass estimates using satellite and aircraft visible and infrared data, a study was conducted to (1) develop and test agricultural crop classification models using two or more spectral regions (visible through microwave), and (2) estimate biomass by including microwave with visible and infrared data. The study was conducted at two locations; Guymon, Oklahoma in 1978, and Dalhart, Texas in 1980. Aircraft multispectral data collected during the study included visible and infrared data (multiband data from 0.5 [mu]m - 12 [mu]m, passive microwave data [C band (6 cm) vertical and horizontal polarizations, and L band (20 cm) horizontal polarization] and active microwave data [K band (2 cm), C band (6 cm), L band (20 cm), and P band (75 cm) like and cross polarizations]. Ground truth data from each field consisted of soil moisture at both sites and biomass at Dalhart. The study was divided into four problems: (1) are differences in individual band responses related to crop type differences? (2) what is the most accurate multifrequency crop classifying dendrogram (tree classifier) at both locations? (3) what is the utility of microwave data alone or in combination with other spectral bands for classifying crops and estimating total biomass? and (4) is the multifrequency tree-classification model variability dependent on phenological or biomass differences? Results indicated that inclusion of C, L, and P band active microwave data from look angles greater than 35° from nadir with visible and infrared data improved crop discrimination and biomass estimates compared to results using only visible and infrared data. The active microwave frequencies were sensitive to different biomass levels. K and C band were sensitive to differences at low biomass levels, while P band was sensitive to differences at high diomass levels. In addition, two indices, one using only active microwave data and the other using data from the middle and near infrared bands, were well correlated to total biomass. Results from the study implied that inclusion of active microwave sensors with visible and infrared sensors on future satellites could aid in crop discrimination and biomass estimation.