Data from: Some like it cold: A general habitat association model for smallmouth bass in stratified lakes

Associations between fish distribution and environmental or spatial gradients at the population level represent habitat use models with potential to be transferable if model parameters are stationary among populations and over years. Further, population-level habitat use models represent the scale of effect – habitat relevant to within population distribution and processes. Here we show that for a widely recognized warm water fish species (smallmouth bass; Micropterus dolomieu), habitat use based on occupancy models extends their distribution in lakes deeper into the metalimnion and hypolimnion  – some bass like it cold. Lake depth at net sites was the gradient used to model habitat use in a comparative lake set (n=11 lakes) and for a large lake with three surveys over a decade. Bayesian quantile regression, with lakes as random effects, revealed all sizes of bass were present in shallow, warm areas of lakes as expected based on thermal preferences. Large adult bass, and not smaller fish, were found at deeper, colder depths producing a significant pattern of large fish found at deeper sites, thus confirming earlier tracking studies. The depth covariate was similar across lakes (no significant difference among lakes with one exception) in the comparative lake set. Over years, the depth covariate was also similar (no significant difference among years) and comparable to the comparative lake set. This consistency points to a general stationary habitat model for smallmouth bass during the summer season. We provide guidance for habitat managers based on this stationarity.

Methods
All lakes were surveyed using a depth stratified randomized netting survey with each depth stratum area based on the aerial coverage of a stratum.

The netting survey for each lake used a depth stratified random design with each lake partitioned into depth categories as follows: 1–3 m; 3-6 m; 6–12 m; 12–20 m; 20–x m; more. Net sites were allocated based on aerial percent of each stratum in each lake and therefore represents a fair sample of bottom habitat in the lake as reflected in depth. A net consisted of 8 panels of varying mesh size (38, 51, 64, 76, 89, 102, 114, and 127mm; totaling to 25m in length X 1.8 m in height). See article for availability of depth maps.

Each net set was comprised of two nets tied together in tandem (“strapped”) and separated by approximately 3 m. Netting locations were separated by a minimum of 250 m. Nets were set on bottom for one hour with each net location separated by a minimum of 250 m to reduce or eliminate gear competition. Employing 1-hour sets approached operational limits of field crews and was as close to an instantaneous point sample site estimate as could be achieved for habitat use analysis. Netting effort on all lakes occurred between 07:00 and 16:00. With each net lift, fish species were recorded, measured for length (FL, mm), and returned to the lake if alive. Each net set included covariates consisting of depth of set (m) and temperature at net depth (°C) derived from a spline model of a depth: temperature profile recorded in mid-survey for each lake. Our definition of metalimnion is the depth range of a temperature profile with the highest rate of change in temperature from warm surface waters to deep cold water. The thermocline is the single depth of greatest change in temperature with the smallest change in depth within the metalimnion (greatest Δ temp over smallest Δ depth; based on the minima of the first derivative of the spline function using function “deriv”). Dissolved oxygen (DO) was above 7 ppm in each lake at all depths so DO was not included as a habitat covariate in this study.