Estimating Juvenile Salmonid Carrying Capacity with Quantile Random Forests

Improving juvenile salmonid carrying capacity through tributary restoration actions is a key component of the 2008 Biological Opinion for endangered Chinook and steelhead populations in the Pacific Northwest of the United States. Estimating the current carrying capacity for rearing parr and identifying the important habitat components that influence that capacity is a necessary challenge to effectively direct restoration actions as well as provide inputs for a variety of life-cycle models. Utilizing data from fish and habitat samples across a range of subbasins within the interior Columbia basin through the Integrated Status and Effectiveness Monitoring Program and the Columbia River Habitat Monitoring Program as well as several collaborating agencies we used a quantile random forest approach to empirically derive estimates of parr carrying capacity. Quantile random forests estimate the (potentially) non-linear relationships between fish and habitat, while incorporating interactions between habitat variables. We also demonstrate how to link reach-level estimates of capacity to greater spatial scales. We compared estimates of capacity to independent estimates where possible. We found QRF predictions corroborate other modeling predictions providing robust evidence that we are identifying habitat factors that limit salmonid densities.