Explainable machine learning improves interpretability in the predictive modeling of biological stream conditions in the Chesapeake Bay Watershed, USA Archives

Entry Thumbnail

Explainable machine learning improves interpretability in the predictive modeling of biological stream conditions in the Chesapeake Bay Watershed, USA

Anthropogenic alterations have resulted in widespread degradation of stream conditions. To aid in stream restoration and management, baseline estimates of conditions and improved explanation of factors driving their degradation are needed. We used random forests to model biological conditions using a benthic macroinvertebrate index of biotic integrity for small, non-tidal streams (upstream area ≤200 km2) in the Chesapeake Bay watershed (CBW) of the mid-Atlantic coast of North America. We utilized several global and local model interpretation tools to improve average and site-specific model inferences, respectively. The model was used to predict condition for 95,867 individual catchments for eight periods (2001, 2004, 2006, 2008, 2011, 2013, 2016, 2019). Predicted conditions were classified as Poor, FairGood, or Uncertain to align with management needs and individual reach lengths and catchment areas were summed by condition class for the CBW for each period. Global permutation and local Shapley importance values indicated percent of forest, development, and agriculture in upstream catchments had strong impacts on predictions. Development and agriculture negatively influenced stream condition for model average (partial dependence [PD] and accumulated local effect [ALE] plots) and local (individual condition expectation and Shapley value plots) levels. Friedman’s H-statistic indicated large overall interactions for these three land covers, and bivariate global plots (PD and ALE) supported interactions among agriculture and development. Total stream length and catchment area predicted in FairGood conditions decreased then increased over the 19-years (length/area: 66.6/65.4% in 2001, 66.3/65.2% in 2011, and 66.6/65.4% in 2019). Examination of individual catchment predictions between 2001 and 2019 showed those predicted to have the largest decreases in condition had large increases in development; whereas catchments predicted to exhibit the largest increases in condition showed moderate increases in forest cover. Use of global and local interpretative methods together with watershed-wide and individual catchment predictions support conservation practitioners that need to identify widespread and localized patterns, especially acknowledging that management actions typically take place at individual-reach scales.

Find more information on the ScienceDirect page.

Entry Thumbnail

Potomac River Water Quality at Great Falls: 1940 – 2019

The U.S. Army Corps of Engineers operates Washington Aqueduct and provides drinking water to the Washington, D.C. area. Washington Aqueduct routinely samples its source of water, the Potomac River. Each year, it reports the monthly averages for basic water parameters and several pollutants and metals. Reports since 2001 are available online. Reports from 1905 to 2000, however, had limited distribution and their legibility has faded over time.

Dr. Norbert A. Jaworski recognized the historical value of these reports. To prevent their loss, he digitized the monthly values for several parameters. The Interstate Commission on the Potomac River Basin (ICPRB) later updated his dataset through 2019 and checked the entered data for accuracy. This report focuses on changes in temperature, hardness, pH, total solids, chloride, nitrate, and sulfate over the 80 years since ICPRB was formed in 1940. Visual representations (“heatmaps”) and trend analysis show significant increasing trends in all these parameters except nitrate. The report is intended to introduce the historical Washington Aqueduct water quality data to a broader audience and highlight their potential value to Potomac studies.

The Supplemental Materials document contains additional graphical representations of the data.

See the video summary of the report:

Entry Thumbnail

Rapid Response Survey of Cyanobacteria Toxin Levels Downstream of North Fork Shenandoah River Algal Bloom After Tropical Storm Ida, 2021

The Virginia Department of Health issued a Harmful Algae Bloom (HAB) Advisory for a 53-mile stretch of the North Fork of the Shenandoah River on August 10, 2021 (Figure 1, left). Samples from multi-species algal mats on the river bottom contained harmful levels of toxins produced by cyanobacteria. Three weeks later, Tropical Storm Ida passed over the North Fork, dumping torrential rain on the watershed. Sharply rising streamflows were expected to scour the benthic algal mats, potentially lysing their cells and releasing toxins as they washed downstream. The ICPRB’s Emergency River Spill Model (ERSM) indicated the scoured material’s leading edge would reach the Potomac River mainstem by September 2nd – 4th and Great Falls near Washington, D. C. by September 3rd – 6th.

Virginia Department of Environmental Quality staff confirmed the algal mats were scoured off the river bottom. Water samples collected by ICPRB at the Shenandoah River mouth indicate the storm’s high flows diluted the algal cells and their associated toxins to below-detection levels before they reached the Potomac River. If flows had been less intense, we hypothesize the scoured material and toxins could potentially have reached the Potomac River mainstem. More advanced flow modeling and additional sampling during algal blooms could better characterize the potential transport of scoured or senescing algal blooms in the Shenandoah River under different river conditions.

Scientist sends testing equipment attached to a rope over the side of a bridge. Shenandoah river is below the bridge.

Rt. 340 bridge over Shenandoah River near Harpers Ferry, WV

Entry Thumbnail

Linking Altered Flow Regimes to Biological Condition: an Example Using Benthic Macroinvertebrates in Small Streams of the Chesapeake Bay Watershed

Regionally scaled assessments of hydrologic alteration for small streams and its effects on freshwater taxa are often inhibited by a low number of stream gages. To overcome this limitation, we paired modeled estimates of hydrologic alteration to a benthic macroinvertebrate index of biotic integrity data for 4522 stream reaches across the Chesapeake Bay watershed. Using separate random-forest models, we predicted flow status (inflated, diminished, or indeterminant) for 12 published hydrologic metrics (HMs) that characterize the main components of flow regimes. We used these models to predict each HM status for each stream reach in the watershed, and linked predictions to macroinvertebrate condition samples collected from streams with drainage areas less than 200 km2. Flow alteration was calculated as the number of HMs with inflated or diminished status and ranged from 0 (no HM inflated or diminished) to 12 (all 12 HMs inflated or diminished). When focused solely on the stream condition and flow-alteration relationship, degraded macroinvertebrate condition was, depending on the number of HMs used, 3.8–4.7 times more likely in a flow-altered site; this likelihood was over twofold higher in the urban-focused dataset (8.7–10.8), and was never significant in the agriculture-focused dataset. Logistic regression analysis using the entire dataset showed for every unit increase in flow-alteration intensity, the odds of a degraded condition increased 3.7%. Our results provide an indication of whether altered streamflow is a possible driver of degraded biological conditions, information that could help managers prioritize management actions and lead to more effective restoration efforts.

The report has been published in Environmental Management.

Entry Thumbnail

Nutrient limitation of phytoplankton in Chesapeake Bay: Development of an empirical approach for water-quality management

Understanding the temporal and spatial roles of nutrient limitation on phytoplankton growth is necessary for developing successful management strategies. Chesapeake Bay has well-documented seasonal and spatial variations in nutrient limitation, but it remains unknown whether these patterns of nutrient limitation have changed in response to nutrient management efforts. We analyzed historical data from nutrient bioassay experiments (1992–2002) and data from long-term, fixed-site water-quality monitoring program (1990–2017) to develop empirical approaches for predicting nutrient limitation in the surface waters of the mainstem Bay. Results from classification and regression trees (CART) matched the seasonal and spatial patterns of bioassay-based nutrient limitation in the 1992–2002 period much better than two simpler, non-statistical approaches. An ensemble approach of three selected CART models satisfactorily reproduced the bioassay-based results (classification rate = 99%). This empirical approach can be used to characterize nutrient limitation from long-term water-quality monitoring data on much broader geographic and temporal scales than would be feasible using bioassays, providing a new tool for informing water-quality management. Results from our application of the approach to 21 tidal monitoring stations for the period of 2007–2017 showed modest changes in nutrient limitation patterns, with expanded areas of nitrogen-limitation and contracted areas of nutrient saturation (i.e., not limited by nitrogen or phosphorus). These changes imply that long-term reductions in nitrogen load have led to expanded areas with nutrient-limited phytoplankton growth in the Bay, reflecting long-term water-quality improvements in the context of nutrient enrichment. However, nutrient limitation patterns remain unchanged in the majority of the mainstem, suggesting that nutrient loads should be further reduced to achieve a less nutrient-saturated ecosystem.

Published in the Journal of Water Research, Volume 188, January 2021:

Entry Thumbnail

Planning Assistance to States: Jennings Randolph Lake Scoping Study Phase II Report

The watershed of the North Branch Potomac River experienced severe environmental degradation and flooding in the 20th century. A dam across the river mainstem was completed in 1982, creating Jennings Randolph Lake. The lake and dam are operated by the U. S. Army Corps of Engineers for four authorized purposes: control floods, dilute downstream pollution, supply drinking to Washington DC during droughts, and provide recreation. Water quality in the North Branch watershed has improved considerably since the dam was built due to many factors, including regulatory enforcement, mine runoff mitigation, wastewater treatment, infrastructure improvements, forest regrowth and the abatement of acid rain (see ICPRB report 19-4). This pilot study was done to determine if an update of the 1997 Reservoir Regulation manual is appropriate at this time. The report reviews and evaluates each of the authorized purposes in terms of their original management goals and objectives, current relevance, and future application.

A copy of the report is available here.

Entry Thumbnail

A water quality binning method to infer phytoplankton community structure and function

Aspects of phytoplankton community structure (e.g., taxonomic composition, biomass) and function (e.g., light adaptation, net oxygen production, exudation) can be inferred with a binning method that uses water transparency (Secchi depth), dissolved inorganic nitrogen, and ortho-phosphate to classify phytoplankton habitat conditions in the surface mixed layer. The method creates six habitat categories, forming a disturbance scale from turbid, nutrient-enriched waters (“degraded”) to clear waters with bloom-limiting nutrient concentrations (“reference”). Across this disturbance scale, estuarine phytoplankton exhibit strong differences in chlorophyll a, count-based biomass, trophic mode, average cell size, photopigment cell content, taxonomic dominance, and the frequency of algal blooms. Differences in ambient dissolved oxygen and dissolved organic carbon are also observed. Two alternate states are apparent, separated primarily by water transparency, or clarity.Water transparency determines cellular light-adaptation and the potential for photosynthesis and growth; nutrient concentrations determine how much of that potential can be realized if and when light becomes available. In Chesapeake Bay, Secchi depth thresholds separating the two states are 0.7–0.9 m in shallow, well-mixed, low salinity waters and 1.2–2.1 m in deeper, stratified, higher salinity waters. The water quality binning method offers a conceptual framework that can be used to infer the overall state of a phytoplankton population more accurately than chlorophyll a alone.

The article was published in Estuaries and Coasts (2020). DOI link: https://doi.org/10.1007/s12237-020-00714-3. Please contact us for a full copy of the report.


Entry Thumbnail

Disentangling the potential effects of land‐use and climate change on stream conditions

Land‐use and climate change are significantly affecting stream ecosystems, yet understanding of their long‐term impacts is hindered by the few studies that have simultaneously investigated their interaction and high variability among future projections. We modeled possible effects of a suite of 2030, 2060, and 2090 land‐use and climate scenarios on the condition of 70,772 small streams in the Chesapeake Bay watershed, United States. The Chesapeake Basin‐wide Index of Biotic Integrity, a benthic macroinvertebrate multimetric index, was used to represent stream condition. Land‐use scenarios included four Special Report on Emissions Scenarios (A1B, A2, B1, and B2) representing a range of potential landscape futures. Future climate scenarios included quartiles of future climate changes from downscaled Coupled Model Intercomparison Project ‐ Phase 5 (CMIP5) and a watershed‐wide uniform scenario (Lynch2016). We employed random forests analysis to model individual and combined effects of land‐use and climate change on stream conditions. Individual scenarios suggest that by 2090, watershed‐wide conditions may exhibit anywhere from large degradations (e.g., scenarios A1B, A2, and the CMIP5 25th percentile) to small degradations (e.g., scenarios B1, B2, and Lynch2016). Combined land‐use and climate change scenarios highlighted their interaction and predicted, by 2090, watershed‐wide degradation in 16.2% (A2 CMIP5 25th percentile) to 1.0% (B2 Lynch2016) of stream kilometers. A goal for the Chesapeake Bay watershed is to restore 10% of stream kilometers over a 2008 baseline; our results suggest meeting and sustaining this goal until 2090 may require improvement in 11.0%–26.2% of stream kilometers, dependent on land‐use and climate scenario. These results highlight inherent variability among scenarios and the resultant uncertainty of predicted conditions, which reinforces the need to incorporate multiple scenarios of both land‐use (e.g., development, agriculture, etc.) and climate change in future studies to encapsulate the range of potential future conditions. DOI link: https://doi.org/10.1111/gcb.14961

Entry Thumbnail

The Influence of Jennings Randolph Lake and Dam Operations on River Flow and Water Quality in the North Branch Potomac River

A multi-year study began in 2018 to determine if an update of the Army Corps’ 1997 Water Control Plan for Jennings Randolph Lake is needed. Watershed and river conditions have improved significantly since the turn of the century, an outcome of regulatory enforcement, mine runoff mitigation, wastewater treatment,  infrastructure improvements, forest regrowth and the abatement of acid rain. The Commission, in partnership with the Corps, has produced a draft Scoping Study report that reviews the dam’s long-running operational objectives and procedures, and assesses the current importance of these procedures in achieving the four mandates. It also reviews various modeling approaches that incorporate modern science and technology for better future management. Learn more…