PS 28-146 - Linking satellite image texture metrics to field measures of tree basal area to monitor post-fire vegetation changes in Madrean woodlands and forests of Mexico and United States

Abstract: Quantifying the rate and trajectory of environmental change after wildfires with use of satellite imagery can provide useful insights into temporal and spatial dynamics of forest structure, disturbance characteristics, and ecological processes. Little is known about the impact of fire on wildlife habitats in Madrean woodlands and forests of the southwestern USA and northern Mexico. To address this, we developed remotely sensed indicators of forest structural changes that can be used to evaluate how fire characteristics (i.e., severity, times burned) and climate influence the rate and timing of environmental changes across this region. We first assembled a time series (1984-2022) of 30-m Landsat spectral indices for the pre-monsoon season (April-June) of each year. We then calculated Gray Level Co-occurrence Matrix (GLCM) features to characterize pixel variation (texture) at a range of spatial scales (3x3, 5x5 and 7x7). Using field samples of tree basal area (n = 919) from plots across our study region, we identified the best predictors of basal area from repeated runs (n =1,000) of generalized additive models, selecting random sets of eight texture metrics in each run and controlling for the influence of vegetation type: pine oak woodland (Pinus spp.; Quercus spp.), oak woodland, pine forest, mixed conifer forest (e.g., Pinus arizonica, Pseudotsuga menziesii) and pinyon juniper woodland (Pinus discolor, Juniperus spp.). Top predictors varied across scale and included metrics calculated with the Normalized Differenced Water Index (NDWI) and the Normalized Burn Ratio (NBR). Using top predictors, we then modeled trends across time at a sample of random points to describe changes in fire history (e.g., timing, number of times burned) and environmental attributes (e.g., elevation, aspect). We then calculated an index of vegetation change linked to magnitude of post-fire vegetation change relative to pre-fire time series. Our results showed a universal decrease in structure immediately after a fire, with varied rates of change in the 5-15 years following fire; the magnitude of post-fire changes were strongly linked to site fire severity, number of times previously burned, and pre-fire vegetation type. These results illustrate the potential benefits of using both spectral and textural information for identifying temporal changes in forest structure.