Reciprocal Environmental Decision Support (REDS): better tailored advice in return for data

Abstract 1. Environmental Decision Support systems provide model-based predictions, tailored to user-inputted information about local ecosystems, which can support management decisions by citizens. Citizen Science systems accept user input to improve models. Thus, each system type emphasises automatic data transfer in one direction: model to citizen or citizen to model, respectively. 2. We introduce a new system type that combines automated data transfer in both directions. 3. Reciprocal Environmental Decision Support (REDS) systems process user-contributed information to (1) provide tailored predictions supporting management, and (2) improve the underlying predictive model. 4. For our proof-of-concept REDS system, Garden Advice, we began with a Bayesian species-habitat association model (for the House Sparrow Passer domesticus ) based on existing data, and obtained new data from UK residents about habitat structure and sparrow observations in domestic gardens. The model made predictions about sparrow presence, and effects of planned garden changes. Model parameters were updated by contributed data; parameter estimates generally tightened. One notable update likely reflects reality (a positive association with grass), while another (a reduced association with roof proximity) likely reflects observation bias. 5. The updated model predicted the new data better than the original model. Thus, untrained observers can provide data of sufficient quality to refine a model of trained observer data. Notwithstanding important questions about distinguishing observation bias from ecologically meaningful information, our system demonstrates that important synergies can be obtained from the REDS approach. Later users of the system obtained better advice thanks to automatically incorporated contributions from earlier users. Lay Summary A summary for non-specialists is available at https://bit.ly/reds_brochure .