Approximation of Ice Phenology of Maine Lakes using Aqua MODIS Surface Temperature Data

Abstract:

Large-scale climatic phenomena known to affect air temperature can also leave their signatures in the duration and timing of ice cover of lakes across the globe. Although studies of lake ice phenology have revealed significant long-term changes in the timing of ice out, empirical knowledge of the spatial and temporal coherence associated with lake ice phenology remains limited. Here we developed a model that uses daytime and nighttime surface temperature observations from the MODIS (Moderate Resolution Imaging Spectroradiometer) sensor onboard the Aqua satellite to approximate 2002 to 2018 ice in for 13 lakes and ice out for 58 lakes in Maine. Several approximation models were developed, calibrated, and evaluated using ground-based records. In the best approximation model, ice out was signaled by reaching a threshold of cumulative positive degrees following the first day above 0˚C. If in the same time period a threshold of negative degrees was also reached, the algorithm restarted the running sum at a later date to prevent early approximations when temperatures temporarily warmed. Ice in was signaled by the reversed version of these methods. The comparison of observed and remotely sensed ice out dates showed relative agreement with a correlation coefficient of 0.67 and an average error of ±8.8 days. Stretches of days with no MODIS data available, likely due to cloud cover, influenced model outputs. A 30-day moving average of the surface temperature data filled the gaps but led to late ice-out approximations. Years with warmer winter temperatures and earlier ice out had approximations that were less precise or too late. Lakes smaller in surface area and near the coast were more likely to exhibit greater error in model approximation of ice out. Ice in drew from a much smaller sample size, but approximations were similar in accuracy to ice out, with a correlation coefficient of 0.71 and an average error of ±9.8 days. The approximation model developed here will help build a more complete history of changes in ice phenology for northern temperate lakes, including those that lack in situ observations, and illuminate the influence of air temperature patterns on the process and timing of ice out and ice in.