Short-Term Variability in Alaska Ice-Marginal Lake Area: Implications for Long-Term Studies

Abstract

Lakes in direct contact with glaciers (ice-marginal lakes) are found across alpine and polar landscapes. Many studies characterize ice-marginal lake behavior over multi-decadal timescales using either episodic ~annual images or multi-year mosaics. However, ice-marginal lakes are dynamic features that experience short-term (i.e., day to year) variations in area and volume superimposed on longer-term trends. Through aliasing, this short-term variability could result in erroneous long-term estimates of lake change. We develop and implement an automated workflow in Google Earth Engine to quantify monthly behavior of ice-marginal lakes between 2013 and 2019 across south-central Alaska using Landsat 8 imagery. We employ a supervised Mahalanobis minimum-distance land cover classifier incorporating three datasets found to maximize classifier performance: shortwave infrared imagery, the normalized difference vegetation index (NDVI), and spatially filtered panchromatic reflectance. We observe physically-meaningful ice-marginal lake area variance on sub-annual timescales, with the median area fluctuation of an ice-marginal lake found to be 10.8% of its average area. The median signal (slow lake growth) to noise (physically-meaningful short-term area variability) ratio is 1.5:1, indicating that short-term variability is responsible for ~33% of observed area change in the median ice-marginal lake. The magnitude of short-term area variability is similar for ice-marginal and nonglacial lakes, suggesting that the cause of observed variations is not of glacial origin. These data provide a new context for interpreting behaviors observed in multi-decadal studies and encourage attention to sub-annual behavior of ice-marginal lakes even in long-term studies.

Keywords

• Landsat
• Optical Remote Sensing
• Google Earth Engine
• land cover classification
• glacier
• cryosphere