Discrete choice modeling of shovelnose sturgeon habitat selection in the Lower Missouri River

Abstract

Substantive changes to physical habitat in the Lower Missouri River, resulting from intensive management, have been implicated in the decline of pallid (Scaphirhynchus albus) and shovelnose (S. platorynchus) sturgeon. To aid in habitat rehabilitation efforts, we evaluated habitat selection of gravid, female shovelnose sturgeon during the spawning season in two sections (lower and upper) of the Lower Missouri River in 2005 and in the upper section in 2007. We fit discrete choice models within an information theoretic framework to identify selection of means and variability in three components of physical habitat. Characterizing habitat within divisions around fish better explained selection than habitat values at the fish locations. In general, female shovelnose sturgeon were negatively associated with mean velocity between them and the bank and positively associated with variability in surrounding depths. For example, in the upper section in 2005, a 0.5 m s )1 decrease in velocity within 10 m in the bank direction increased the relative probability of selection 70%. In the upper section fish also selected sites with surrounding structure in depth (e.g., change in relief). Differences in models between sections and years, which are reinforced by validation rates, suggest that changes in habitat due to geomorphology, hydrology, and their interactions over time need to be addressed when evaluating habitat selection. Because of the importance of variability in surrounding depths, these results support an emphasis on restoring channel complexity as an objective of habitat restoration for shovelnose sturgeon in the Lower Missouri River.

Subcutaneous injection of visible implant elastomer in sterlet (Acipenser ruthenusLinnaeus, 1758): a study on compatibility and retention

“…Mark‐recapture studies allow assessments of links between subpopulations, and estimations of population size (Coggins et al., ). Also, monitoring the movement of individual fish is crucial in identifying behavioural traits including the use of distinct habitat types (Bonnot et al., ; Cunjak et al., ; Young & Scarnecchia, ; Zeller, ).…”

Foraging plasticity by a keystone excavator, the white-headed woodpecker, in managed forests: Are there consequences for productivity?

“…We completed five iterations of this method, each time removing another 20% of cases. Based on random chance alone we expected 50% (1 in 2) of used sites to be correctly identified and values >50% suggested that our model predicted use better than random (Bonnot et al, 2011).…”

“…We validated our top model using a modified k-fold approach following the methods of Bonnot et al (2011). We randomly removed 20 percent of cases as ''test data", fit the model with the remaining cases (''training data"), and tested the ability of the model to identify used points within choice sets of the test data.…”

Pallid sturgeon seasonal habitat selection in a large free‐flowing river, the lower Mississippi River

“…To evaluate selection for depth and current velocity, availability must be measured at a time concomitant with fish location. The availability of such hydrologic and substrate data measured within 1 day of sturgeon location have been used in short (2.4 km mean length) reaches of the Missouri River to test habitat selection of shovelnose sturgeon (Bonnot et al, ); but in the LMR, where channel widths can exceed 1 km, such reach‐scale mapping would take several days and habitat conditions can change within that timeframe dependent on discharge. If future researchers desire to compare the availability of continuous environmental conditions, such as depth and current velocity, over the categorical method applied, scale and timeframe of sampling must be relevant to the conditions at the time of fish location.…”

“…We used a discrete choice model to estimate the probability of use and selection of the seven habitat types where pallid sturgeon were encountered. Discrete choice models have been used to measure probability of using categorical habitats (Dailey, Gitelman, Ramsey, & Starcevich, ; Kneib, Knauer, & Küchenhoff, ) and selection based on presence‐only data (Bonnot et al, ; Davis & Wagner, ; Kneib et al, ). Habitat selection was defined as the ratio of the probability of use and proportion available () with values <1 indicating avoidance and >1 indicating positive selection (Kneib et al, ).…”