| Tule elk (Cervus elaphus nannodes) in Colusa and Lake Counties utilize habitats that are difficult to accurately survey through site-based methods and their abundance is unknown. We designed a survey method to collect fecal pellets for DNA-based monitoring, a method found to be effective for other ungulates. We used 1,207 elk-presence locations to develop a MaxEnt model, which included vegetation type, vegetation cover, and mean diurnal temperature range. We tested the model with independent data from GPS-collared elk (AUC = 0.885), then with occupancy surveys on the Cache Creek herd. We employed 6-km triangle transects and 4.5-km linear transects in 33 random 2-km2 grid cells, 24 in predicted-presence habitat, and 9 in predicted-absence habitat. We documented elk presence using fecal pellets and other elk sign at 92% (22 of 24) of the predicted presence cells and 44% (4 of 9) of the predicted absence cells. Non-detection in predicted-presence cells coincided with high-intensity agriculture (n = 2). Detections in predicted-absence cells were typically <0.5-km from a predicted presence location. Our use of the habitat model to stratify the landscape increased our survey efficiency, resulting in a 76% reduction of the landscape surveyed relative to a random survey. |