3.4 Pond inundation regimes
Based on HMM estimates, ponds varied in both initial timing and duration of inundation (Figure 4), with initial inundation dates ranging from 10 July 2018 to 7 August 2018, over a small geographic area (Figure 1). Eleven of the fourteen ponds in our study had at least one predicted wet state during our monitoring period. Ten of these ponds had wet states predicted from both datasets. Based on the state predictions by both models, ponds in the central range of our study area filled first, with ponds T8, T9, T12, and T13 all inundated between 10 - 17 July 2018. Ponds in the northern and southern portions of the study area had more variation in their initial inundation dates, which were predicted to occur between late July and mid-August (Table S6, Figure 1, Figure 4).
Inundation dates inferred from pond-only and pond-control models largely aligned (Figure 4). Only pond T20 was predicted to have a wet state by one model (the pond-only model) and not the other. The presence of vegetation at the perimeter of the pond and mud inside the housing of the T20 pond logger during our visit in April 2019 suggest that the pond may have been inundated with water at some point during logger deployment. Visual inspection of the T20 temperature standard deviation readings revealed a slight difference between August and October 2018. Ponds T11 and T15U, which had no predicted inundation dates, also showed slightly lower readings from the pond loggers relative to the control loggers at certain points in the monsoon season and had mean state values close to but slightly above our wet state thresholds. It is possible that some water accumulated in these ponds and that our wet state threshold for the HMMs lacked the sensitivity to capture these low signals. The tSD threshold may need to be adjusted to increase precision in cases where small amounts of water accumulate for durations shorter than 5 days.