| Summary: | The streaked horned lark (hereafter “lark”; Eremophila alpestris strigata) is a federally listed bird subspecies of the Pacific Northwest that occupies open landscapes with short vegetation and abundant bare ground. Across its breeding range, which has contracted dramatically, the lark relies primarily on human-modified habitats maintained in early successional states (e.g., agricultural fields, airfields, deposition islands). The focus of our study is the lower Columbia River where larks primarily occupy deposition islands created as a byproduct of shipping channel deepening (dredging) projects. The U.S. Army Corps of Engineers has proactively adopted a shifting-habitat-mosaic management approach when placing dredge (sand) materials to maintain lark habitat throughout the project area. To help achieve their goal, we created a geospatial tool capable of quantifying and tracking habitat suitability for larks throughout the project area with satellite imagery. We created spatial models of lark breeding habitat by incorporating lark survey data, Sentinel-2 satellite imagery, dredge-deposition maps, and surface-elevation models. We built candidate predictive models of lark breeding habitat after combining survey data (years 2016 and 2017), predictor variables, and logistic regression. In addition, we created a novel seral-stage predictor variable to identify young, mature, old, and unsuitable locations; critical information for lark conservation planning and dredge-disposal operational planning. We challenged probability models the year they were constructed (2016) and the following year (2017) with set-aside survey data. The best model contained terrain ruggedness, deposition age, vegetation greenness (e.g., lushness and density of vegetation) inside a 10-m cell, and heterogeneity in greenness within a 100-m radius. Verification accuracy ranged from 75 to 83% depending on the model tested and year. Importantly, the seral-stage variable allowed us to quantify and locate early, optimal, and late habitats. Model ...
|
|---|