An assessment of grassland loss, woody encroachment, and pesticide use on North American grassland bird populations

Abstract

The North American Breeding Bird Survey monitoring effort has consistently documented grassland bird declines since 1966. Within Kansas, greater prairie-chicken (Tympanuchus cupido) surveys suggest declines over the past 30 years. Grassland loss has largely been cited as the main driver of declines; however, I predicted other factors, such as pesticide application and woody encroachment, may have induced additive and interactive effects. I evaluated grassland loss, woody encroachment, and pesticide use as potential threats to meadowlarks (Sturnella magna and S. neglecta) and greater prairie-chickens (Tympanuchus cupido) within the Great Plains and Kansas, respectively. I examined meadowlark abundance throughout the Great Plains of the United States to assess the relative influence of each potential threat between 1996 and 2019. The linear least square regressions showed strong relationships between observed meadowlarks and the three factors. With consideration of different starting populations between Bird Conservation Regions, the grassland model performed best based on the slope of the beta coefficient (β = 0.513, 95% CI [0.463, 0.562]; ΔRMSE = 3.46), while the woody encroachment model reported the highest predictive accuracy (β = -0.453, [ -0.487, -0.419]; ΔRMSE = 6.52). The pesticide model had a negative relationship with observed meadowlarks (β = -0.292, [ -0.328, -0.256]; ΔRMSE = 0.09), and indicated negative effects beyond grassland loss alone, the only correlated factor (r = -0.70). To evaluate these three threats on greater prairie-chickens in Kansas, I used a random forest modeling approach to compare grassland, woody cover, and pesticide use between occupied (n = 5,552) and “pseudoabsent” (n = 11,223) greater prairie-chicken lek site locations from 1993 to 2020. My best-supported variables were the amount of grassland at 5 km, woody cover percentage within 800 m, and pesticide application rates within 5 km of lek locations. I validated the model fit using independent holdout data (~25%) and ROC curve (AUC = 0.956); the model correctly predicted the use or absence of leks with 89.95% accuracy. My analysis suggests that though the amount of grassland has slightly declined, the amount of available habitat could have increased on average in Kansas. However, I estimated a 15% loss of greater prairie-chicken habitat in the Flint Hills coinciding with woody encroachment and increased pesticide use. My greater prairie-chicken species distribution model likely overestimates expected habitat without considering finer-scale factors (i.e. anthropogenic features) and vegetation structure but provides a foundation for local assessments. These results supported my hypotheses that grassland loss, woody encroachment, and pesticide use all present negative demographic consequences for meadowlarks, greater prairie-chickens, and other grassland birds within Kansas, and across the Great Plains. Current trends may be altered with grassland conservation action. The findings emphasize the importance of reducing the consequences of agricultural intensification and promoting bountiful grasslands for wildlife and producers alike.