2.1.1 Wildfire occurrence detection in Alaskan tundra
We defined the extent of Arctic tundra in Alaska with the commonly used Circumpolar Arctic Vegetation Map (Walker et al., 2009). MODIS Thermal Anomalies/Fire locations product (MCD14ML; Giglio et al., 2003) was chosen to determine the locations and dates of fire occurrences. We first identified individual fire events with MCD14ML data based on its consistent information of active fire points. We designed a spatiotemporal clustering method designed based on the Density-Based Spatial Clustering of Applications with Noise (DBSCAN; Ester et al., 1996) algorithm (Text S1; Figure S1). The maximum distance between two neighboring fire points in a cluster was set to 2.5 km (Loboda and Csiszar, 2007). Since fire events that occurred during different time periods could be grouped into the same cluster, we further separated points of different fire events in a spatial cluster with a temporal gap of 4 days, as suggested by Loboda and Csiszar (2007). The locations and dates of the active fire points with the earliest acquisition time were then extracted to represent the tundra fire occurrences.