An automated algorithm for river ice detection using MODIS data: Examining spatial and temporal patterns in Arctic river ice breakup (2015)

Undergraduate: Sarah Cooley

Faculty Advisor: Tamlin Pavelsky
Department: Geology

---

The annual spring river ice breakup has vast consequences for northern ecosystems as well as significant economic implications for Arctic industry and transportation. The timing of breakup can be used to understand regional climate variability, and there is substantial interest in how patterns in ice breakup may change as a result of climate warming. River ice breakup research is limited by the sparse distribution of hydrologic stations in the Arctic region. The use of satellite imagery allows for breakup research over the entire reach of rivers, yet only a few such studies exist. In this talk I will describe an automated algorithm for detecting the timing of river ice breakup using MODIS imagery and present an analysis of spatial and temporal trends in breakup for the four largest pan-Arctic rivers. Through splitting each river into 10 km segments and classifying each river pixel as snow, ice, mixed ice/water or open water, the algorithm determines the date of breakup, here defined as the first day where 75% of the river segment is open water. Cloud-obscured imagery is by far the largest source of error, and the average window of breakup error is around +/- 1.5 days. There is noticeable variability in breakup timing both temporally and across the entire river length. Statistically significant trends towards earlier occurrence of breakup are found for the lower Mackenzie and upper Lena rivers.