Document Type


Date Created



Grassland birds across both North America and globally are in rapid decline, and these population losses are primarily due to agricultural practices driving land-conversion and habitat degradation. While the effects of landscape characteristics like land cover, fragmentation, anthropogenic features, and land use on bird abundance and diversity are well-studied, the mechanisms behind those effects are poorly understood for many species. In our studies, we considered how individual condition and behavior may be underpinning some population losses in Western Meadowlarks (Sturnella neglecta). First, we investigated what natural, anthropogenic, and climatic landscape characteristics best predicted Western Meadowlark song structure. We used 421 song samples collected across the United States and measured a suite of 12 temporal and tonal song components for each sample. We then used random forest modeling to determine which of 30 measured landscape characteristics best predicted the first three temporal and tonal principal components of song structure. We found that final models all had negative R2 values, indicating that no landscape variables considered in this study predicted song structure. We proposed three hypotheses to explain this finding: 1) Sexual selection and cultural drift could obscure small environmental effects; 2) A positive selection bias among citizen scientists for species-typical habitats and high-quality recordings could skew our samples towards higher quality habitats; 3) Western Meadowlarks may not have behavioral plasticity to modify components of songs in their repertoire or songs selected from their repertoire, and therefore may be singing sub-optimal songs. In our second study, we sought to better understand how Western Meadowlarks respond to landscape characteristics by determining the effects of landscape on individual condition, then determining if there is a relationship between individual condition and song structure. We captured, measured (n = 29), and recorded songs from (n = 26) Western Meadowlarks in Weld County, CO. Then, we calculated the proximity of roads, energy development, and water resources to each capture location as well as the land cover surrounding the site in a 1-km radius circle. We measured the minimum and maximum song frequencies, the number of flat notes, and the inflection rate of several song types from each individual. We used linear regression modeling to determine that waterbodies moderately, negatively influenced mass to tarsus ratio. Roads and non-grassland habitat moderately, negatively influenced culmen length. Finally, proximity to rivers/streams and distance from oil/natural gas infrastructure moderately, positively influenced wing chord length. We then examined song structure as predicted by those three morphological measures using linear mixed effects regression models. We detected a weak, negative relationship between mass to tarsus ratio and inflection rate as well as a weak, negative relationship between wing chord and minimum song frequency. This second study investigated the trends detected across the United States Great Plains and led us to conclude that landscape is exerting selective pressure on Western Meadowlarks through condition. However, these differences in individual condition are not reflected as we might expect in the song characteristics measured in this study. These studies are novel in their use of landscape and behavior data to investigate possible mechanisms behind large-scale avian population trends, as well as the ground-truthing of the relationships identified through a field study. Finally, we call for further research on the effects of landscape on birds beyond the final population outcomes we currently are detecting.


Western Meadowlark; Sturnella neglecta; Landscape; Birdsong; Song; Body Condition; Random Forest Model; GIS; land cover; anthropogenic