3.2 | Occupancy Models
We retained the most supported detection probability model structure for each taxa when assessing various larval fish occupancy model combinations. Similar to detection models, the combination of shad and percids, with carp and freshwater drum occupancy estimated separately, received the most support (ΔAICc = 0.00,wi = 0.44; Table 2). Other likely combinations include all taxa estimated separately (ΔAICc = 1.63, wi = 0.20) and freshwater drum estimated separately from bigheaded carp, shad, and percids (ΔAICc = 1.78, wi = 0.18; Table 2). Through the addition of habitat to this taxa model structure, the most supported model included backwater estimated separately and thalweg combined with channel border habitats for shad and percids (ΔAICc = 0.00, wi = 0.21; Table 2). Two of the six top and competitive model structures incorporated this habitat grouping among different taxa (Table 2). Finally, we added environmental covariates to determine the combination that best improved the taxa × habitat model. The most supported taxa × habitat × environmental covariate model (ΔAICc = 0.00, wi = 0.69) incorporated Julian date and CV temperature for all taxa with the added effect of temperature on bigheaded carp and freshwater drum (Table 2). Bigheaded carp were the only taxonomic group to retain both discharge and CV discharge in the top model (Table 2).
Freshwater drum occupancy among all habitats (Ψ = 0.84, 95% CI = 0.73 to 0.91) was higher than bigheaded carp (Ψ = 0.29, 95% CI = 0.15 to 0.47) and shad and percids in backwaters (Ψ = 0.53, 95% CI = 0.41 to 0.66) or thalweg/channel border (Ψ = 0.42, 95% CI = 0.34 to 0.51). Julian day had a quadratic effect on occupancy probability for all taxa: shad and percids occupancy was highest on day 144 (May 24th) followed by bigheaded carp on day 170 (June 19th) and freshwater drum on day 184 (July 3rd; Figure 5). Bigheaded carp (β = -0.63, 95% CI = -1.03 to -0.23) and freshwater drum (β = -0.89, 95% CI = -1.24 to -0.55) occupancy declined with increased variation in weekly water temperature while shad and percids occupancy increased with increasing water temperature CV (β = 0.33, 95% CI = 0.16 to 0.51; Figure 5). Weekly water temperatures also had a quadratic effect on occupancy of bigheaded carp and freshwater drum: bigheaded carp occupancy was highest at 20.7°C and drum was highest at 23.9°C (Figure 5). Bigheaded carp occupancy was also positively related to mean weekly discharge (β = 1.29, 95% CI = 0.19 to 2.39) whereas the slope estimate of CV discharge incorporated zero (β = -0.01, 95% CI = -0.35 to 0.36; Figure 5).