Tree mortality detection and assessment using AVIRIS imaging spectroscopy data in Sierra Nevada
Abstract
Warmer and drier climate and decades of active land management have changed forest compositions and functions significantly, affecting forest resilience to extreme climatic events and their long term sustainability. Recent prolonged 2012-2016 drought has resulted in more than 100 million dead trees since 2013 in the Sierra Nevada mountains. Regular monitoring of forest water stress and tree mortality are critical for understanding the drivers for tree mortality and for local communities and land managers and. We aim to investigate the capability of the imaging hyperspectral data in detecting dead trees and understanding the temporal evolution and spatial patterns of tree mortality in California.
We here used the AVIRIS airborne hyperspectral data acquired by JPL during the 2013-2016 HyspIRI summer campaigns along Tahoe, Sierra and Sequoia National Forests. Random Forest (RF) models were developed to detect dead trees with spectral metrics derived from AVIRIS, climate and topographic data. The high spatial resolution tree mortality estimated from multispectral UAV imagery by Koontz et al (in prep) over 32 forest plots, dominated by pine/mixed-conifer forests, across an elevation gradient of 1000m, were used to train and test the models. Our preliminary results showed that the RF regression model was able to identify the percentage of tree mortality with R2 higher than 0.6. The top five most important predictors include the absorption spectral feature between 898 and 1071 nm, absorption spectral feature between 1100 and 1252 nm, elevation, climate water deficit and cellulose absorption index. The spatial patterns of tree mortality also agreed with those from the USFS Aerial Detection Surveys (ADS) maps. Similarities of models among the study sites were further evaluated to develop a more generalized model that can be applied to the whole Sierra Nevada. With the availability of upcoming hyperspectral data from various satellite instruments, the method developed here will provide complementary tools and products to ADS surveys and Landsat-based monitoring, to track forest condition and mortality, and inform rapid response to forest stress, develop adaptive forest management practices, and ultimately improve the resilience and sustainability of California's Natural Ecosystems.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.B53N2585J
- Keywords:
-
- 0426 Biosphere/atmosphere interactions;
- BIOGEOSCIENCES;
- 1807 Climate impacts;
- HYDROLOGY;
- 1813 Eco-hydrology;
- HYDROLOGY;
- 1851 Plant ecology;
- HYDROLOGY