Using Citizen Science Data to show Declines in Riverine Sentinel Invertebrates

Abstract:

Citizen science data is collected based on available funds. Thus, survey locations often lack consistent data collection and missing values are common. While citizen science water quality datasets can be large, on a regional and local scale they are often undervalued. However, citizen science datasets represent an opportunity for managers to define broad ecosystem trends if challenges such as missing data can be addressed. We used presence and absence data of sentinel invertebrates (stonefly, order Plecoptera) collected by citizen scientists over 17 years to approximate trends in stream health in urban Detroit, Michigan, USA. To overcome the missing data hurtle we used a combination of spatial (inverse distance weighting and spatial stream network), temporal (Bayesian state-space), and machine learning (ensemble random forest) models combining atmospheric, hydrologic, and biologic data. Using the estimated missing values, we determined negative population trends in stonefly driven by stream temperature via a dynamic occupancy model. Urban streams present a challenge to resource managers because data is collected at disparate locations, frequencies, and inconsistently recorded. However, a combination of methods pulling together publicly available and citizen science data from across disciplines can inform managers and support land-use decisions.