The caddis aren’t alright: modeling Trichoptera richness in streams of the northcentral United States reveals substantial species losses

Abstract

Despite the importance of caddisflies in aquatic ecosystems, few studies have assessed the most important variables affecting their species richness throughout a large region or attempted to model such richness throughout the region. We sampled nearly 500,000 adult caddisfly specimens representing 18,288 species occurrences and 282 total species using ultraviolet light traps from nearly 800 streams of varying size and condition throughout the northcentral US, an area of about 1.3 million km2. We tested 18 candidate variables reflecting natural habitat conditions, anthropogenic disturbance, and weather differences for their ability to account for variation in the caddisfly species richness of our samples. Multiple linear regression, including significance testing and ranking by Akaike Information Criterion importance values, determined a best model including eight variables. Caddisfly species richness increased with percentage of intact natural upstream habitat, elevation, dew point, longitude, depth of soil organic matter, and distance of soil to bedrock, and decreased with the percentage of non-native plants and total runoff value. The percentage of intact upstream habitat alone accounted for >40% of the variation in caddisfly species richness. After correcting for dew point to equalize weather conditions, our 7-variable model predicted that, on average, a 50% loss of intact upstream habitat would cause a 30% decrease in caddisfly species richness relative to undisturbed conditions, a 75% loss would cause a 55% decrease, and a near total loss would cause a 75% decrease. Applying our model to the 760,047 stream segments of our study area estimated that, in the absence of disturbance, nearly 90% of stream segments were predicted to contain 31–40 caddisfly species. Based on calculated disturbance levels, however, only 15% of stream segments had maintained at least 90% of that richness, whereas 58% had lost at least half. Species extirpations were generally lowest in the northern forested portion of our study area and, except for some relatively small protected areas, higher in the southern agricultural portion. Overall, our data indicated a tremendous number of site-level extirpations over a large area owing principally to watershed-level habitat disturbance. Such losses probably occur in other aquatic insect taxa throughout much of the United States and elsewhere.