Evaluation of sampling and analytical approaches to assess fish population responses to floodplain restoration

Abstract Introduction The effectiveness of floodplain restoration projects in terms of creating habitat for sensitive fish species like salmonids is well‐studied, but inferences about large‐scale responses are sometimes intractable due to smaller‐scale data collection. Objectives This study considered how sampling design and analytical methods affect the ability to make inferences about fish responses to floodplain restoration across spatial scales. It also compared the descriptive and predictive performance of different modeling approaches to describe habitat use at each scale. Finally, it tested whether fish density‐weighted habitat diversity was associated with reach‐scale fish abundance in restored and reference floodplains. Methods We evaluated physical habitat attributes that strongly correlate with juvenile Chinook Salmon ( Oncorhynchus tshawytscha ) density. Furthermore, density estimates from randomized and non‐randomized sampling were compared using linear and nonlinear models. Modeled density was then used to generate a weighted habitat diversity index. Results Nonlinear models provided a better fit to the data. Data simulation methods showed that non‐randomized sampling gave similar estimates of reach‐scale density as randomized methods. The density‐weighted habitat metric was a correlate of population size across subbasins. Restored floodplains tended to have higher habitat diversity associated with larger reach‐scale populations. Conclusions Non‐randomized, small‐scale sampling methods, combined with strong model fit and data simulation can assess restoration efficacy at larger scales.