CONTRIBUTIONS OF OPTICAL REMOTE SENSING TO PERMAFROST MAPPING IN DONNELLY TRAINING AREA, ALASKA
AN ABSTRACT OF THE THESIS OFKiran Thapa, for the Master of Science degree in Geography and Environmental Resources, presented on April 8, 2020, at Southern Illinois University Carbondale.TITLE: CONTRIBUTIONS OF OPTICAL REMOTE SENSING TO PERMAFROST MAPPING IN DONNELLY TRAINING AREA, ALASKA MAJOR PROF...
Main Author: | |
---|---|
Format: | Text |
Language: | unknown |
Published: |
OpenSIUC
2020
|
Subjects: | |
Online Access: | https://opensiuc.lib.siu.edu/theses/2764 https://opensiuc.lib.siu.edu/cgi/viewcontent.cgi?article=3778&context=theses |
Summary: | AN ABSTRACT OF THE THESIS OFKiran Thapa, for the Master of Science degree in Geography and Environmental Resources, presented on April 8, 2020, at Southern Illinois University Carbondale.TITLE: CONTRIBUTIONS OF OPTICAL REMOTE SENSING TO PERMAFROST MAPPING IN DONNELLY TRAINING AREA, ALASKA MAJOR PROFESSOR: Dr. Guangxing Wang Permafrost occupies about a quarter of the northern hemisphere land with 25.5 million ha. Global warming and anthropogenic activities affect the dynamics of permafrost. Snow and permafrost, in turn, serve as an indicator of climate change and human activity disturbance. The dynamics of permafrost are often estimated using interferometric Synthetic Aperture Radar (InSAR) methods. However, acquiring and processing InSAR images is costly and computation intensive. Due to various spectral variables and indices available from optical images, Landsat satellite images that are free-downloadable provide the potential for studying and monitoring changes of permafrost. The overall objective of this study was to explore the use of optical images as a cost-effective method to map permafrost in Donnelly Training Area (DTA) - an installation located in Alaska. First, Landsat 8 OLI/TIRS images from January 2014 to December 2018 were used to calculate various remote sensing variables. The variables included Land Surface Temperature (LST), albedo, Soil Moisture index (SMI), Normalized Difference Vegetation Index (NDVI), Normalized Difference Snow Index (NDSI), Normalized Difference Built-up Index (NDBI), Normalized Difference Water index (NDWI), Simple Ratio (SR), Soil Adjusted Vegetation Index (SAVI), Normalized Burn Ratio (NBR), Triangular Vegetation Index(TVI), Visible Atmospherically Resistant Index (VARI), and Active Layer Thickness (ALT). Moreover, elevation, slope, and aspect were obtained from a digital elevation model (DEM). The variables were used to estimate the probabilities of permafrost presence (POP) for DTA. The logistic and linear models were respectively selected and optimized based on logistic and linear stepwise regression for the estimation of and ALT. A total of 414 field observations that were collected from 1994 to 2012 were utilized for validation of models.The results showed that the POP in DTA was significantly affected by all the factors except aspect and EVI. The factor that was most correlated with ln((1-POP)/POP) was elevation, then NDVI, albedo, ALT, LST, NDWI, NDSI, slope, TVI, RSR, SMI, NDBI, SR, SAVI, NBR and VARI. A total of six prediction models were obtained. The elevation, NDVI, LST, TVI, ALT, SLOPE, RSR, SMI, NBR, and NDSI were finally chosen in the best model 5.6 with the smallest relative root mean square error (RMSE) and Akaike information criterion (AIC). The albedo used in previous studies was excluded in the final model, implying that the albedo was not critical to the prediction of POP. In addition to the previously used elevation, NDVI and SMI, other predictors including LST, TVI, ALT, SLOPE, RSR, NBR, and NDSI could not be ignored in the prediction of POP. The model generated reasonable spatial distribution of POP in which POP had greater values in the east, northeast, north, and northwest parts and smaller in the south and southwest parts. Except for NDVI, NDWI, NDSI, aspect, and RSR, moreover, all other predictors showed significant contributions to the prediction of ALT. The SMI, ELEVATION, SAVI, NDBI, SLOPE, LST, SR, EVI, VARI, and TVI were finally selected in the best model 5.14 with the smallest relative RMSE and AIC. The ALT highly varied over the study area with the spatial patterns inversely consistent with those of POP.The results are essential for the governments, policymakers, and other concerned stakeholders to estimate the degradation of permafrost in DTA and minimize the risk of policy decision-making for land use management and planning. This study will help to understand the global climate change, changing ecosystems, increasing concentration in the atmosphere, and human activity-induced disturbance. |
---|