We developed habitat suitability models for occurrence of three invasive riparian woody plant taxa of concern to Department of Interior land management agencies, as well as for three dominant native riparian woody taxa. Study taxa were non-native tamarisk (saltcedar; Tamarix ramosissima, Tamarix chinensis), Russian olive (Elaeagnus angustifolia) and Siberian elm (Ulmus pumila) and native plains/Fremont cottonwood (Populus deltoides ssp. monilifera and ssp. wislizenii, Populus fremontii), narrowleaf cottonwood (Populus angustifolia), and black cottonwood (Populus balsamifera ssp. trichocarpa and ssp. balsamifera). We generally followed the modeling workflow developed in Young et al. 2020. We developed models using five algorithms with VisTrails: Software for Assisted Habitat Modeling [SAHM 2.1.2]. We accounted for uncertainty related to sampling bias by using two alternative sources of background samples: random (10,000 spatially-filtered (50-kilometer [km]) random background samples) and Salix (10,000 randomly-selected occurrence records of Salix spp.). We constructed model ensembles with the 5 models for each taxon (five algorithms) with each background method, as well as with all 10 models for each taxon (five algorithms by two background methods), for three different occurrence likelihood thresholds (1st percentile, 10th percentile, and MSS (maximum sensitivity and specificity)). We also used the model ensembles to identify major watersheds where each taxon was under-represented in occurrence records relative to predicted habitat suitability, to evaluate risk of undetected or future invasion. For each 6-digit hydrological unit (HUC6, USGS Watershed Boundary Dataset) within the study area, we calculated the difference between actual occurrence record density and the density of occurrence records that would be expected if occurrence records were distributed among watersheds in proportion to habitat suitability in MaxSS 10-model ensembles. This data bundle contains the merged data sets used to create the models, occurrence locations that were used for independent assessments of model accuracy (not used in model training), the raster files associated with each taxon, and tabular summaries of actual and expected occurrence record densities by HUC6. The spatial data are organized in a separate folder for each taxon, each containing 9 rasters. Each of the rasters represent the following: 1) X1st_random - ensemble of 5 models with random background data and 1st percentile threshold 2) X10th_random - ensemble of 5 models with random background data and 10th percentile threshold 3) MaxSS_random - ensemble of 5 models with random background data and MaxSS threshold 4) X1st_Salix_1st - ensemble of 5 models with random background data and 1st percentile threshold 5) X10th_Salix - ensemble of 5 models with random background data and 10th percentile threshold 6) MaxSS_Salix - ensemble of 5 models with random background data and MaxSS threshold 7) X1st_combined - ensemble of 10 models with random and Salix background data and 1st percentile threshold 8) X10th_combined - ensemble of 10 models with random and Salix background data and 10th percentile threshold 9) MaxSS_combined - ensemble of 10 models with random and Salix background data and MaxSS threshold The bundle documentation files are: 1) 'RiparianSDMs_main.xml' (this file), which contains the project-level metadata 2) 'ModelTrainingData.csv' contains the merged data set used to create the models, including location and environmental data. 3) 'IndependentAssessmentData.csv' contains the data set used to assess accuracy of model predictions (occurrence locations not used for model training) 4) XX.tif where XX is the raster type explained above in taxa subfolders. 5) 'HUC6Summaries.csv' contains tabular summaries of actual and expected occurrence record densities by HUC6. 6) 'bison_citations.txt' contains the different data sources with occurrences from the BISON database. This file specifically describes the IndependentAssessmentData.csv that includes the data used to assess the accuracy of the model predictions.