Assessing the risk of dreissenid mussel invasion in Texas based on lake physical characteristics and potential for downstream dispersal

Abstract

ebra mussels ( Dreissena polymorpha ) and quagga mussels ( Dreissena bugensis ) were likely introduced from Ponto-Caspian Eurasia to the Laurentian Great Lakes inadvertently via ballast water release in the 1980s and have since spread across the US, including Texas. Their spread into the state, including reservoirs in both Brazos River and Colorado River basins, has resulted in a need to delimit suitable dreissenid habitat and dispersal potential in Texas. The objective of our research was to assess invasion risk in Texas by 1) predicting distribution of suitable habitat of zebra and quagga mussels using Maxent models; 2) refining lake-specific predictions for present zebra mussels via collection of physicochemical data; and 3) assessing the potential for downstream spread of zebra mussels by applying environmental DNA (eDNA) methods in the Leon and Lampasas Rivers downstream from the invaded Lakes Belton and Stillhouse Hollow, respectively. Maxent models did not predict the occurrence of suitable habitat for quagga mussels within Texas. However, our models accurately identified global zebra mussel habitat (AUC = 0.919), and Bioclim layers representing temperature and precipitation data both strongly influenced predictions. Predicted “hotspots” of suitable zebra mussel habitat in Texas occurred along the Red and Sabine Rivers of north and east Texas, as well as patches of suitable habitat in central Texas between the Colorado and Brazos Rivers and extending inland along the Gulf Coast. Most of the Texas panhandle, west Texas extending toward El Paso, and the Rio Grande valley were predicted to provide poor habitat suitability. Collection of physicochemical data (dissolved oxygen, pH, specific conductance, and temperature on-site as well as laboratory analysis for Ca, N, and P) from zebra mussel invaded lakes and a subset of identified high-risk lakes of North and Central Texas, did not aid predictions. Visual inspection of biplots of the first three components of a principle component analysis, which together accounted for ~80% of data variability, did not reveal separation between invaded and uninvaded lakes, and logistic regression analysis also failed to identify predictive relationships between measured variables and invasion status. Using eDNA analysis, we detected the presence of zebra mussel eDNA at 11 of 12 sites and up to at least 90.7 river km downstream from a pair of infested reservoirs. Rate of positive detection among water samples at each site ranged from 1/5 to 5/5, and within positive water samples, rate of detection among technical replicates ranged from 1/8 to 8/8, suggesting considerable heterogeneity in the zebra mussel eDNA signal in both rivers. Furthermore, no clear spatial pattern in detection rate occurred. Thus, a monitoring strategy that combines traditional sampling (e.g. settlement substrate samplers and microscopy) at sites immediately below a dam, and transitioning to more sensitive eDNA analysis at distances further from the dam may represent the most successful strategy for detection of dreissenid mussel downstream dispersal. Overall, we have demonstrated that while quagga mussels do not appear to represent an invasive threat in Texas, suitable habitat for continuing zebra mussel invasion exists within Texas, and stream and river connections may contribute to their spread. The threat of continued expansion of this poster-child for negative invasive species impacts warrants further prevention efforts, management, and research.