| Utility of satellite-based remote sensing (SRS) data, ETM7+ and AVHRR, for mapping East Texas forest resources was the primary dissertation theme. First, the utility of ETM7+ for providing stand-level data for a commercially managed forest tract was evaluated. A strong linear relationship was found between the infrared band of ETM7+ and age (r = −0.876), and density of stems (r = 0.854). The agreement between the classified ETM7+ imagery and stand inventory data was only 54% (raw) and 48% (transformed). Area estimates for clear-cuts and SMZ were lower than the estimate. Based on the results obtained, it was concluded that ETM7+ could be used with the existing inventory database. Second, the regional forest cover estimates were obtained using ETM7+ and AVHRR data. The overall accuracy of the classified ETM7+ and AVHRR data was 85% and 67%, respectively. Forest area estimates obtained from ETM7+ and AVHRR data were compared to traditional methods. SRS estimates were within the 95% confidence interval of the traditional estimate. The variability associated with ETM7+ estimates was relatively lower than the variability associated with AVHRR estimates. Third, the pixel-by-pixel agreement between the classified ETM7+ and AVHRR images ranged between 60% and 78%. Counties with contiguous forest cover had higher agreement values. Fourth, the influence of the proportion of forest area within an AVHRR pixel was analyzed. Most of the AVHRR pixels with high forest cover (>80%) were classified as forest. When the forest cover was low (<20%), most of the AVHRR pixels were classified as non-forest. However, several AVHRR pixels were misclassified when the proportion of forest area ranged between 20% and 60%. Finally, the spatial pattern of ETM7+ forest pixels within an AVHRR pixel was analyzed using three landscape metrics. SRS data could play an important role in collecting forest resource information for tactical and strategic management needs. At the current level of technology, SRS data may not be able to completely replace the current photo-based and field survey methods for collecting information about forest resources, but can form part of a multi-stage sampling scheme. |
