Using Remote Sensing Data Fusion Modeling to Track Seasonal Snow Cover in a Mountain Watershed

Abstract

Seasonal snowfall is the largest component of the water budget in many mountain headwater regions around the world. In addition to sustaining biological water needs in drier, lower elevation areas throughout the year, mountain snowpack also provides essential water inputs to the Critical Zone (CZ) - the outer layer of the Earth’s surface, which hosts a variety of biogeochemical processes responsible for transforming inorganic matter into forms usable for life. Water is a known driver of CZ activity, but uncertainty exists in its spatial and temporal interactions with CZ processes, particularly in the complex terrain of heterogeneous mountain areas. Increasing pressure on the CZ due to climate change and human land use needs creates an urgency to better understand the CZ system and how it may change in the future. An important variable for water driven CZ behaviors in mountain areas is the spatial extent of snow, also known as snow-covered area (SCA). SCA in mountain areas can change quickly over small scales of time and space with large impacts on the rest of the system. It has been difficult historically, however, to measure snowpack extent for large areas on very fine spatial and temporal scales due to a lack of remote sensing datasets with both of these fine scale characteristics. In this study we use the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) to fill this historic knowledge gap for the East River watershed in Colorado, USA. By fusing low spatial and high temporal resolution data from MODIS (500-m, daily) with high spatial and low temporal resolution data from Landsat (30-m, 16 days), a fine resolution, 30-m daily dataset can be created. This study is one of the first to use this model with the primary intent of monitoring SCA in a mountain watershed. The first component of the study in this thesis presents a comprehensive validation of STARFM for use in monitoring snow cover in mountain areas. Normalized Difference Snow Index (NDSI) values from MODIS and Landsat are used as input to the STARFM model, and synthetic NDSI values at 30-m resolutions are obtained for days without Landsat data acquisitions. After converting NDSI to binary snow cover, we then examine the temporal performance of STARFM for an entire calendar year. The model’s performance is also analyzed for different landscape features known to influence snow cover. Accuracy, precision, recall, and F-score values indicate that the model is able to successfully predict the location of SCA in the landscape when validated with Landsat data. The second component of the study describes the process of creating the daily, 30-m NDSI dataset with STARFM for 20 water years of analysis and provides examples of how these data can be used to monitor SCA in a mountain watershed. We examine patterns of percent annual snow cover for three of the water years from the dataset, a dry, average, and wet water year. Here we find that predictable patterns of SCA occur over those years, with the highest percent annual snow cover occurring during the wet year and the lowest occurring during the dry year. Despite these differences, however, elevation is clearly the dominating factor in determining the spatial variability of snow cover in the landscape for all three water years. We also connect our snow cover analysis back to CZ processes by examining the timing of snow cover disappearance with the peak of annual stream discharge at the watershed outlet. The results of this work provide a multi-decadal dataset of snow cover information for the East River that can be used for future research into snowpack and streamflow forecasting, modeling of the movement of water through the CZ, and the effects that climate change may have on these processes. This study also provides examples of methods that can be used for further snow monitoring work in the East River watershed and other snow-dominated mountain catchments similar to it.

Untitled

“…STARFM-a weighted function-based methods that weight the spatial and temporal aspects, thus capable of reconstructing multi-time data that have gaps and then producing images with good spatial and temporal resolution, developed by Gao et al (2006) has been widely used and has shown success in producing synthetic data like Landsat which has good spatial and temporal resolution for identifying phenological events. Similar studies utilizing STARFM have demonstrated promising results in generating images with adequate spatial and temporal resolution, particularly for detecting phenology (Gallagher, 2018;Onojeghuo et al, 2018;Son et al, 2016;Vincent, 2021). In addition, downscaling MODIS and Landsat-8 data can produce a high-quality time-series data since MODIS and Landsat have similar orbital characteristics with only 30-minutes time difference when crossing equator (Hwang et al, 2011).…”

Automated snow cover detection on mountain glaciers using space-borne imagery

“…Several recent studies have worked to overcome gaps in spatial and temporal coverage of SCA estimates 70 associated with image repeat intervals, cloud cover, and spatial resolution. Techniques such as data fusion and spatial downscaling (Berman et al, 2018;Rittger et al, 2021;Vincent, 2021;Walters et al, 2014), leveraging multiple satellite image products (e.g., Gascoin et al, 2019), and the use of the 3-5 m-resolution ~daily PlanetScope imagery (Cannistra et al, 2021;John et al, 2022) have helped to overcome these gaps. Yet, applying the NDSI thresholding method to glacier surfaces is likely to yield inaccurate results due to the overlapping NDSI ranges for snow versus ice 75 and firn.…”