Modeling climate change effects on transport and fate of pathogens in an urban coastal area of Lake Michigan Open Access

CCR

UW Madison Center for Climatic Research

SEWRPC

Southeastern Wisconsin Regional Planning Commission

GLCFS

Great Lakes Coastal Forecasting System

ELPC

Environmental Law and Policy Center

GCM

General circulation models (GCM)

NOAA-OHHI

National Oceanic and Atmospheric Administration Oceans and Human Health Initiative

GLERL

Great Lakes Environmental Research Laboratory

POM

Princeton Ocean Model

NWS

National Weather Service

ASOS

Automated Surface Observing System

NBDC

National Buoy Data Center

CDF

Cumulative Distribution Function

KMKE

General Mitchell Airport

MG

Main gap

SG

South gap

NG

North gap

BB

Bradford Beach

SSB

South Shore Beach

LI

Linnwood

HA

Howard Avenue

The Great Lakes are the main source of drinking water for more than 34 million people, and the receiving body for tributary flows and wastewater from most coastal urban centers in the region. Tributary runoff and sewage overflows deliver pathogens and nutrients into Lake Michigan's urban coastal areas following storm events (McLellan et al. 2007). Pathogens and nutrients discharged to Great Lakes coastal waters are transported by lake circulation to coastal waters, beaches, water intakes, and other sensitive areas, where they may pose threats to human health and the environment (Murray et al. 2001; Corso et al. 2003).

Climate change can affect lake elevation, water temperature, dissolved oxygen, and ice and snow covers (Fang & Stefan 2009; Sharma et al. 2015; Magee et al. 2016; Woolway & Merchant 2018; Deoli et al. 2021, 2022), alter seasonal flow regimes and hydrologic extremes (O'Gorman & Schneider 2009; Byun et al. 2019), lake mixing regimes, and aquatic ecosystem productivity (O'Reilly et al. 2003), modify waterborne disease risk in the Great Lakes region of the U.S. (Patz et al. 2008). It is well documented that the changes in hydrometeorological extremes under climate change will affect human health and communities by alteration of temperature, precipitation, droughts, and flooding (Vörösmarty et al. 2000; Rosenberg et al. 2010; Tohver et al. 2014). In particular, climate change can affect the patterns of circulation and transport of pathogens in Great Lakes coastal waters. Planning for infrastructure renewal, allocation of resources for public health and recreation, and restoration efforts require the prediction of climate change effects on the occurrence of pathogens in Lake Michigan urban coastal areas. As the effects of climate change are known to be regionally impacted especially due to the current climate (baseline) regional scaling for hydrological studies is very important (Tohver et al. 2014; Verma et al. 2015).

There is a substantial body of literature on climate change modeling and on the effects of climate change on the Great Lakes. Beletsky et al. (1999) examined potential lake circulation changes using general knowledge of lake physics. They discussed how general circulation models (GCM) prediction of milder winters might lead to longer stratified periods and less ice cover, and the predicted decrease in wind speed and a decrease in water level because of changes in evaporation. They discussed the relation between the amplitude of currents, wind speed, density gradients, and depth, and the effect of decreased ice cover and wind speed on currents. Beletsky et al. (1999) summarized the potential climate change effects on Great Lakes hydrodynamics and water quality, and the conditions that should be met before the problem could be addressed properly. The conditions reported were knowledge of current climatology as a reference for future changes, knowledge of boundary conditions, and proof of the reliability of hydrodynamic models. Beletsky & Schwab (2008) addressed the climatology condition, and Beletsky & Schwab (2001, 2008) reported significant progress in the reliability of long-term hydrodynamic modeling. A report from leading Midwest University scientists (Environmental Law and Policy Center, ELPC 2019) warned of dangers to the Great Lakes and regional economy from climate change. The report documented impacts on water quality, health, infrastructure, agriculture, and tourism. The report examined regional climate change in the Great Lakes, changes in Great Lakes watershed hydrology, impacts on lake ecology, and public and economic impacts of changes to the Great Lakes.

Climate change modeling is an active area of research. Veloz et al. (2012) described significant progress in the knowledge of forcing functions because of improved reliability of GCM predictions and downscaling climate models to address the coarse spatial resolution of GCMs. Xiao et al. (2018) presented a dynamical downscaling projection of future climate change in the Great Lakes region using a coupled air-lake model. The Ontario Climate Consortium provided a summary of the state of climate modeling in the Great Lakes Basin and assessed model strengths and limitations, the knowledge gaps climate modelers face, the state of climate model users and translators, and recommendations for modelers, users, and translators moving forward, including the coupling of models that incorporate the atmosphere, land, and lakes in 3D models (Delaney & Milner 2019).

All studies show a rise in temperature in future for the Great Lakes region, but still the downscaling GCM for regional data is a complicated problem and requires heavy computational modeling. These cause uncertainty in the meteorological data that are used for local-scale studies and may affect the hydrodynamic modeling of the lakes (Li et al. 2016; Hattermann et al. 2017).

Bravo et al. (2017) presented a study on the fecal coliform footprint in the Milwaukee Harbor coastal area using field sampling and modeling, sponsored by the National Oceanic and Atmospheric Administration Oceans and Human Health Initiative (NOAA-OHHI) Program. That whole-lake model/nested model provided sufficient spatial and temporal resolution and simulated the appropriate transport mechanisms. Measurements and model results showed that the footprint can extend up to 8 km into the lake during sewer overflow events, depending on the meteorology that drives the interaction between tributary loads, the whole lake, and local circulation.

McLellan et al. (2013) studied climate change risks and impacts on urban coastal waters in the Great Lakes. They developed a climate change impact component into watershed and nearshore models to improve predictive capabilities to urban coastal regions. In that study, the UW Madison Center for Climatic Research (CCR) created downscaled climate change data for meteorological stations around Milwaukee and Lake Michigan. The Southeastern Wisconsin Regional Planning Commission (SEWRPC) and Tetra Tech implemented a hydrologic model that predicts flows and bacteria loads for the tributary watersheds. This research presents the part of that study that addressed the effect of climate change on the transport and fate of bacteria in Milwaukee's urban coastal area, using the model developed by Bravo et al. (2017). Although the model, developed in this research, is tested for the Milwaukee coastal region, it will be used for other regions in the Great Lakes. The finding of this research could extend the prediction of water quality in other coastal regions, too. The following objectives are the focus of this study:

• (1)

  To determine the important meteorological drivers in the production of watershed flows and bacteria loads, and in the transport and mixing in lake coastal waters.

• (2)

  To introduce a practical selection of climate change scenarios appropriate to investigate and link watershed and lake transport processes.

• (3)

  To study the transport and fate of bacteria in the Great Lakes urban coastal areas under changing climate.

It is important to have prediction ability to study the hydrodynamics of the coastal areas of the Great Lakes, to evaluate the water quality and ecosystem of the lakes under future climate. This is important for any future planning, resource allocation, and restoration efforts.

This section summarizes the methods used in developing lake meteorological forcing for hydrodynamic modeling, methods for the creation of scenarios for watershed hydrologic and bacteria load modeling and for hydrodynamic modeling, the hydrodynamic and bacteria transport modeling methods, and the field data used in model validation.

A model of lake circulation, thermal regime, and bacteria transport in Lake Michigan's urban coastal area was developed and validated by Bravo et al. (2017). In this study, that model was used to evaluate the effects of different climate change scenarios on the hydrodynamic and biogeochemical behavior of the urban coastal areas in the Great Lakes. This model builds upon the Great Lakes Coastal Forecasting System (GLCFS) Lake Michigan model developed by NOAA GLERL. The nested model that has high resolution to represent the Milwaukee Harbor was expanded to include a fecal contamination transport module and tributary flows and fecal indicator bacteria loads. Both the GLCFS model and the nested model are based on a Princeton Ocean Model (POM, Blumberg & Mellor 1987) version adapted to the Great Lakes by NOAA GLERL (Schwab & Bedford 1994). The nested model gets its boundary condition from the whole-lake model. The whole-lake model has a coarser grid of 2 km by 2 km. This may reduce the accuracy of the hydrodynamic model. Based on the validation of this nested model, presented by Bravo et al. (2017), the accuracy of the model outputs both in hydrodynamic and transport of bacteria is in the acceptable range compared to other studies (Thupaki et al. 2010). It is possible to develop other models such as Finite Volume Community Ocean Model (FVCOM) to be able to use finer grids in the area of interest in the whole-lake model in future to possibly improve the accuracy. In this case, there is no need to have two models; one for the whole lake and one for the area of interest, nested in the whole-lake model.

The bacteria transport module simulates the processes of advection, dispersion or mixing, bacteria fall through the water column, light-dependent inactivation rate, and base mortality, as described by Thupaki et al. (2010). The bacteria transport module and the hydrodynamics module are solved simultaneously and use the same numerical methods.

In this study, the meteorological forcing from climate change scenarios explained in the following sections is used in the hydrodynamic and transport model to find out the effect of different climate change scenarios on pathogen concentration in future.

This section presents general aspects of the meteorological forcing for both watershed modeling and lake transport. Climate change scenarios for lake modeling were developed using the arguments that bacteria load to Lake Michigan is most sensitive to the spring season, and transport in coastal waters is most sensitive to changes in wind speed and direction. Therefore, the analysis focused on the spring time frame, thus matching the time frame of interest in the watershed modeling. Meteorological forcing for watershed and lake transport modeling are described further in the following sections.

The main goal of this study was to simulate the effect of climate change on the transport of bacteria in the nearshore of Lake Michigan. Atmospheric forcing for the transport of bacteria in the lake includes the wind field, air and dew point temperature, and cloud cover. Previous studies have shown that the transport in the lake is most sensitive to the wind field over the lake (Bravo et al. 2013, 2017). A 12-year (2000–2011) baseline record of the atmospheric forcing variables was gathered and projected to mid-century using 13 global circulation models. The 156 projections of wind speed at the Milwaukee Airport meteorological station were analyzed, and the realizations corresponding to the third and first quartiles were selected to represent the wind-speed worst-case and best-case scenarios, respectively. Atmospheric forcing corresponding to those worst-case and best-case scenarios was used to drive the lake transport model, providing results deemed to represent the effects of climate change.

Tributary inflows and bacteria loads were analyzed using a similar approach. The main drivers of inflows and loads are watershed precipitation and air temperature. A 10-year meteorological period from 1988 through 1997 baseline record of those drivers was gathered, a watershed model was run to predict and verify baseline tributary inflows and bacteria loads, and the year with the highest tributary load within the baseline period was identified. Watershed precipitation and air temperature were projected to mid-century under the A1B emissions scenario. The watershed model, driven by the precipitation and air temperature projected using 14 GCMs, was used to predict tributary inflows and bacteria loads for the projected year with the highest bacteria loads. Tributary inflow and bacteria load that represent the top 10% of the projected realizations were used as input to the lake transport model. The watershed and lake transport model was thus linked to simulate the March–May spring period that produces the highest tributary inflows and bacteria loads.

CCR developed predictions of wind vectors, air temperature, dew point temperature, and cloud cover at the ASOS stations for the climate change scenarios using remapped predictions of the baseline observation period (2000–2011). CCR remapped the baseline observations to two future periods, namely 2046–2065 (mid-century) and 2081–2100 (end of century), using the 13 global circulation models shown in Table 3 (IPCC 2007) for remapping. The 2046–2065 remapped predictions were used in lake hydrodynamic and transport modeling, in consistency with the watershed model.

The data were created using the cumulative distribution function (CDF) remapping technique (Veloz et al. 2012). This method uses the time-mean CDF in the climate of the 20th-century experiment (20C3M) scenario and the time-mean CDF in a future scenario to map the nth percentile in the 20th century to the nth percentile in the future. This preserves all the covariances between the data. All these data are for the A1B scenario (IPCC 2007, middle-of-the-road scenario).

The model cccma_cgcm3_1 projection for 2005 yielded approximately the 75th percentile projected March–May average wind speed increment, and the model mri_cgcm2_3_2a projection for 2011 yielded approximately the 25th percentile increment. The meteorological predictions of those two models, for all 11 ASOS stations around the lake, were used to construct the worst-case and best-case meteorological forcing scenarios, respectively.

Table 4 summarizes baseline scenarios and climate change scenarios that the Lake Michigan model and the Milwaukee nested model were run for, using the 1990 baseline tributary flows and loads in all cases. Model-predicted bacteria concentration at stations main gap (MG), south gap (SG), north gap (NG), Bradford Beach (BB), South Shore Beach (SSB), and the Linnwood (LI) and Howard Avenue (HA) intakes were used to assess effects of climate change. The intakes of two water treatment plants are at stations LI and HA (not shown), located north and south of the Milwaukee Harbor.

Using a period that has a complete observation data set is a prerequisite for the development of climate change predictions. The watershed model was developed for a previous study using as a baseline the existing 1988 through 1997 complete observations data set at the General Mitchell weather station. A complete observation data set was not available for the 1988–1997 period at the 11 ASOS stations around Lake Michigan. The Center for Climatic Research used the existing 2000–2011 period of complete observations at the 11 ASOS stations around Lake Michigan to develop climate change projections for hydrodynamic modeling. Therefore, the baseline and mid-century climate change projections for watershed and hydrodynamic models are not completely synchronized.

The scenarios for the contributing watersheds and for the transport and mixing in the lake were linked by considering the spring season in all cases. The approach taken in this study is believed to provide an adequate sensitivity analysis of the effects of climate change.

The bacteria transport model needs, as a driving boundary condition, a continuous time series of bacteria loading at the mouth of the tributaries, but continuous monitoring is not yet available. However, continuous data on tributaries streamflow are available. Bravo et al. (2017) demonstrated the estimation of a reasonably accurate time series of bacteria loading based on the existing significant relation between tributary streamflow and bacteria loading measured between June 2009 and October 2011 by McLellan's laboratory and Steve Corsi (USGS) at the mouth of the Milwaukee River. The procedure was used to estimate fecal coliform loads from the Kinnickinnic, Menomonee, and Milwaukee River watersheds and the Oak Creek watershed, for the baseline and climate change hydrodynamic modeling scenarios.

A comparison of model runs showed negligible differences between the results illustrated in Figure 6(a) and 6(b). This is consistent with other research in the Great Lakes basin that shows the variation of the nutrients and sediment concentration under mid-century climate is less than 10% (Verma et al. 2015). The transport of baseline and projected fecal coliform at relevant sites showed negligible effect of using those two different loads for the same lake hydrodynamics. Changes in spring air temperature, and consequently evapotranspiration, counteracted with changes in precipitation, resulting in insignificant changes in watershed loads. Baseline watershed loads for the year with the largest spring watershed flows and bacterial loads, and their corresponding projection, transported under the same lake meteorological forcing, produced insignificant changes in concentration at selected critical coastal locations.

• (1)

  Watershed loads and the transport of bacteria in Great Lakes coastal waters are most sensitive to different meteorological drivers. Study of climate change effects on the occurrence of pathogens in Lake Michigan urban coastal areas requires therefore the analysis of different scenarios.

• (2)

  From the point of view of the contributing watersheds, it is most important to study the sensitivity of bacterial loads with respect to changes in spring-season precipitation and air temperature.

• (3)

  Baseline watershed loads for the year with the largest spring watershed flows and bacterial loads, and their corresponding projection, transported under the same lake meteorological forcing, produced insignificant changes in concentration at selected critical coastal locations.

• (4)

  From the viewpoint of transport and mixing in lake coastal waters, the most important driver is the wind field.

• (5)

  The use of spring-season watershed loading in the study of lake transport links the watershed and lake transport components of the study. Scenarios for the transport and mixing by lake hydrodynamics were developed by selecting climate projections over the lake that yielded the 75th and 25th percentile projected spring-season wind speed at the local meteorological station to define the worst-case and best-case climate change scenarios, respectively.

• (6)

  Downscaled meteorological data for stations around the lake were used to develop meteorological forcing for the whole lake under the worst-case and best-case climate change scenarios.

• (7)

  The patterns of bacteria transport showed significant changes under climate change conditions, and the changes in fecal coliform concentration at critical coastal locations were explained by changes in current vector fields.

• (8)

  The results of this study will provide predictive tools and guidelines for policymakers to adopt for future climate, especially for water intakes and other sensitive areas, where they may pose threats to human health and the environment. It should be noted that this model is tested for Milwaukee coastal area, but is applicable for other urban coastal regions.

• (9)

  The model here is nested on the whole-lake model. The bacterial load used for validation of the model is based on discrete measurement in tributaries and coastal areas. Adding more continuous measurements may improve the accuracy of the results in future.

• (10)

  The findings of this study will be helpful for preparation for future climate and adoptive strategies for the mitigation of bacteria concentration in the urban coastal areas and Combined Sewer Overflow (CSO) events in the watershed. It is recommended to expand this study in future by adding different watershed management scenarios.

The National Oceanic and Atmospheric Administration Sectoral Applications Research Program (SARP) (extramural grant number NA10OAR4310184) funded this research project. The Center for Climatic Research at UW Madison (David Lorenz) created downscaled climate change data for meteorological stations around Milwaukee and Lake Michigan. The hydrologic model that predicts flows and bacteria loads for the tributary watersheds was developed by Tetra Tech (Kevin Kratt and J. Butcher) under contract with SEWRPC (Michael G. Hahn).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.