Ecological risk response models for antecedent flow volatility effects on freshwater fish and macroinvertebrate communities across the conterminous United States

This data release contains the outputs, including model objects, of a causal dose-response analysis quantifying how Antecedent Flow Volatility (AFV) affects the probability of poor ecological condition in freshwater fish and macroinvertebrate communities across the conterminous United States (CONUS).

Antecedent Flow Volatility (AFV) is defined as the log10-transformed standard deviation of monthly observed-to-expected (O/E) flow ratios (Natural Flow Anomaly, or NFA) in the 12 months preceding a biological sample collection. Higher AFV values indicate greater hydrologic instability over the prior year. This workflow uses a causal inference framework based on Average Dose Response Functions (ADRFs) to estimate the marginal effect of increasing AFV on the probability of biological communities being classified in poor condition.

To isolate the causal effect of AFV from confounding landscape and land-use variables (drainage area, agricultural and developed land cover, dam influence, and wastewater discharge density), propensity-score-style balancing weights were estimated using energy balancing methods (WeightIt R package). Weighted generalized linear models (quasibinomial family) with ecoregion interaction terms were then used to estimate region-specific ADRFs, with significance assessed via Omnibus Curve Tests using the Imhof approximation (adrftools R package).

Files included in this data release are:

Cleaned biological condition datasets with associated AFV metrics, landscape confounders, and flow metrics calculated for both fish (fish_clean_AFV_landscape.csv) and macroinvertebrates (bugs_clean_AFV_landscape.csv).

Crosswalk between EPA Ecoregion Level III and the 9-category aggregated Ecoregions used by the EPA National Rivers and Streams Assessment, which are then later re-aggregated into 7-categories for this study (Ecoregion3_aggregation_crosswalk_NRSA_Herlihy2008.csv).

Fitted weighted outcome model objects for fish and macroinvertebrate response to AFV (fish_fit_glm_weightit_logsdprior12mon.rds and bugs_fit_glm_weightit_logsdprior12mon.rds)

Summary table of the omnibus curve significance tests for each ADRF for each taxa and regional combination for whether the ecological response is significantly different than flat (ADRF_regional_significance.csv).

A data dictionary of entity and attribute information for ease of use.