Combining citizen science data and the hierarchical structuring of temporary streams to reconstruct the patterns of channel wetting and drying

Temporary streams (i.e., non-perennial streams) cover more than half of the global stream network. They are highly dynamic systems and important habitats. Still, they have so far not been thoroughly monitored because gauging stations are expensive, not well suited for measuring zero flows, and provide only data for a single location in the stream network. An alternative way to monitor temporary streams is by visual assessment. However, on-the-ground surveys of stream networks tend to be highly time-consuming. Hence, visual observations by citizen scientists provide a great opportunity to collect high spatial- and temporal resolution data, even though there are challenges regarding the accuracy and irregularity of the observations.To assess the potential of citizen science data to obtain temporal resolved information on the state of temporary streams, we used the observations submitted by citizen scientists using the CrowdWater app for 63 locations on a 5km2 forested hill in Zurich, Switzerland. The number of observations per location during the last three years varied from 1 to 257 (median: 40). There was at least one stream state observation for 402 days, with a maximum of 42 observations per day and a median of 10 observations per day. In addition, trained staff monitored 59 streams (30 overlap with the citizen science data set) at an almost bi-weekly resolution during six months (24 days of observations at all 59 points).The hierarchical structure of channel network dynamics postulates the existence of a fixed, unique order according to which stream segments are activated during network expansion (from the most to the least persistent). To understand the hierarchical structure of stream wetting and drying at the study site, we applied the graph-based method developed by Durighetto et al. (2023) on the available data. This data-driven method would allow us to fill the gaps of the irregular citizen science data (leading to 25,728 reconstructed observations compared to the 4,354 original observations). The hierarchical structures for the two datasets differed, even if only locations that were part of both data sets and the same period were used to determine the hierarchical structure. In the citizen science dataset, the order of activation of the observed stream locations is less clearly identifiable (i.e., more uncertain). This is likely due to the non-systematic and sporadic nature of the data, i.e., only a few stream observations on the same date, as well as errors in the data. Nonetheless, this information can be used to give guidance to the citizen scientists on which streams to observe more frequently because they provide the most crucial information about the wetting and drying patterns of the network.Reference: Durighetto, N., Noto, S., Tauro, F., Grimaldi, S., Botter, G. (2023). Integrating spatially-and temporally-heterogeneous data on river network dynamics using graph theory. Iscience, 26(8).