Authors: Dar Roberts*, University Of California, Santa Barbara, Christopher Kibler, University of California, Santa Barbara, Conor McMahon, University of California, Santa Barbara
Topics: Remote Sensing, Land Use and Land Cover Change, Biogeography
Keywords: Drought, disturbance, time series, fire, spectral mixture analysis, remote sensing
Session Type: Paper
Presentation File: No File Uploaded
Widespread availability of multi-temporal remote sensing data makes it possible to identify surface changes due to disturbance over broad spatial and temporal scales. We present a generalized framework for using standardized spectral mixture analysis to identify departures from mean behavior over an extended time series. Reference endmembers for green vegetation (GV), non-photosynthetic vegetation (NPV – litter, bark) and Soil are identified from a reference library. A simplified form of Multiple Endmember Spectral Mixture Analysis (MESMA) is applied to a Landsat time series; the number of endmembers is allowed to vary on a per-pixel basis, but only from a limited set. The mean and variance for GV, NPV, Soil and Shade fractions are then calculated for each pixel. Residuals are calculated by subtracting the pixelwise mean from each fraction in the time series. This approach was applied to Landsat data acquired over the Santa Barbara, California area from 1984 to 2019 using anniversary dates acquired primarily in June. MESMA was applied to each date to generate mean and variance images for each fraction for each year. Inter-annual variation in precipitation and fire were the dominant sources of anomalies. High rainfall years generated positive GV residuals. Drought generated negative GV and positive NPV residuals. Major fires generated the largest changes, resulting in decreased GV and increased NPV fractions that persisted for five to seven years. Extensive positive NPV anomalies were observed throughout the 2012-2016 drought. Debris flows in Montecito, California were expressed as a positive Soil and NPV anomaly in impacted watersheds.
To access contact information login