Abstract
In the early 2000s, severe drought and beetle outbreaks brought on by climate change led to extensive piñon pine mortality in northern New Mexico. Mapping piñon mortality is critical for understanding and monitoring impacts to ecosystem services. I used Landsat imagery, finer-resolution reference data, and statistical modeling to predict absolute within-pixel percent tree mortality (PTM). I found that a balanced random forest model had the highest accuracy. I applied the model to a time series and assessed spatial and temporal patterns of the estimated tree mortality. Pixels within the study area experienced a mean of 42.3% tree mortality, a mean duration of 3 years, and a mean rate of mortality of 22% per year. From 1987 to 2009, cumulative tree mortality reached 67,190 ha, or 19.8% of the woodland area, the equivalent of 48.9 million piñons. These results demonstrate the capability of mapping drought-induced PTM using a Landsat time series.