Characterizing baseline legacy chemical contamination in urban estuaries for disaster-research through systematic evidence mapping: A case study

Abstract

Natural disasters such as floods and hurricanes impact urbanized estuarine environments. Some impacts pose potential environmental and public health risks because of legacy or emerging chemical contamination. However, characterizing the baseline spatial and temporal distribution of environmental chemical contamination before disasters remains a challenge. To address this gap, we propose using systematic evidence mapping (SEM) in order to comprehensively integrate available data from diverse sources. We demonstrate this approach is useful for tracking and clarifying legacy chemical contamination reporting in an urban estuary system. We conducted a systematic search of peer-reviewed articles, government monitoring data, and grey literature. Inclusion/exclusion criteria are used as defined by a Condition, Context, Population (CoCoPop) statement for literature from 1990–2019. Most of the peer-reviewed articles reported dioxins/furans or mercury within the Houston Ship Channel (HSC); there was limited reporting of other organics and metals. In contrast, monitoring data from two agencies included 89–280 individual chemicals on a near-annual basis. Regionally, peer-reviewed articles tended to record metals in Lower Galveston Bay (GB) but organics in the HSC, while the agency databases spanned a wider spatial range in GB/HSC. This SEM has shown that chemical data from peer-reviewed and grey literature articles are sparse and inconsistent. Even with inclusion of government monitoring data, full spatial and temporal distributions of baseline levels of legacy chemicals are difficult to determine. There is thus a need to expand the chemical, spatial, and temporal coverage of sampling and environmental data reporting in GB/HSC.

Graphical Abstract

1. Introduction

With a history of hurricanes and flooding, Texas coasts and inland cities face numerous natural hazards that can affect the fate and transport of contaminants. However, within the field of disaster research, pre-existing conditions research and long-term fate and transport research are largely lacking (Knap and Rusyn 2016). In 2017 Hurricane Harvey struck the Texas Coast, resulting in extreme flooding over the span of four days. The resulting damages were estimated as being approximately $125 billion with more than 270,000 residences flooded (Blake and Zelinsky 2018; Harris County Flood Control District 2018; Texas Commission on Environmental Quality 2018). However, it has been difficult to characterize the degree to which Harvey changed or redistributed chemical contaminants in the Houston area. This region has a long history of legacy contamination, such as contamination with polycyclic aromatic hydrocarbons (PAHs), polycyclic biphenyls (PCBs), organochlorine pesticides, PCDD/Fs, and heavy metals, from intensive industrial activity over the past 100 years (Qian et al. 2001; Santschi et al. 2001; Yeager et al. 2007; Lakshmanan et al. 2010; Howell et al. 2011; Sappington et al. 2015; Hieke et al. 2016; HARC and Galveston Bay Foundation 2018; Louchouarn et al. 2018; Al Mukaimi, Kaiser, et al. 2018; Oziolor et al. 2018; HARC and Foundation 2019; HARC and Galveston Bay Foundation 2020). Therefore, these contaminant classes collectively are of interest because of their legacy presence and prevalence within GB/HSC environmental matrices (e.g., sediments, water) and biota (e.g., fish).

As an active port, the Houston Ship Channel (HSC) sees vessel activity comparable to the ports in Long Beach, California and New York/New Jersey (Campo 2020). In turn, regular maintenance is required to deepen and widen the channel through dredging. The earliest dredging efforts date back to the mid-1850s when the navigational channel from the Gulf of Mexico to Houston was not readily accessible (Mark Vincent et al. 2015). Because of these regular maintenance activities, the region’s sediment composition can vary since some marsh areas and private lands have been filled in with the dredge materials; Atkinson Island for example, served as a man-made dredge island (Mark Vincent et al. 2015).

Given the regular maintenance and historical contamination, both the dredged sediments and non-dredged sediments are of interest for public health as the public may be exposed to these sediments after a natural disaster or may be indirectly affected through environmental exposures. For example, sediment deposits were documented in local parks after Hurricane Harvey (Karaye et al. 2019). This particular example highlights the role floods and extreme events can play in the redistribution of sediments and thus implicate the potential of chemical redistribution.

With extreme events such as severe flooding events, hurricanes, tsunamis, and cyclones common in other parts of the world, understanding relevant exposures in global coastal and estuarine environments is timely for disaster-research. However, multiple factors are necessary to understand the impact of such events within the environment. Some examples of required data include contaminant concentrations over space and time as well as relevant exposure pathways (Oswer 2002; Hanrahan 2012). Sediment mixing after extreme events such as Hurricane Harvey are also necessary to note as estuaries eventually recover depending upon regional inputs (Du, Park, Timothy M. Dellapenna, et al. 2019; Du, Park, Timothy M Dellapenna, et al. 2019; Kiaghadi and Rifai 2019; Dellapenna et al. 2020).

As transparency needs and regulatory reform pertaining to data requirements increase in the scientific community, both scoping reviews and systematic evidence maps (SEMs) are being implemented in environmental sciences and environmental health. These tools’ methodologies and applications have grown since they first appeared in early 2000s literature (Munn et al. 2015; Miake-Lye et al. 2016; Haddaway et al. 2018; Haddaway and Macura 2018; Kohl et al. 2018; Munn, Stern, et al. 2018a; Munn, Peters, et al. 2018; Macura et al. 2019; Wolffe et al. 2020). An SEM serves as a guidance document so audiences can consider the state of science for a particular research topic. This means that an SEM will often include grey literature sources in addition to peer-reviewed literature. In turn, the outcomes of a SEM are reported through tabular or visual summaries so that knowledge and research gaps can be identified (James et al. 2016; Miake-Lye et al. 2016; Munn, Stern, et al. 2018b; Munn, Peters, et al. 2018; Saran and White 2018).

One challenge facing SEMs within environmental science is the lack of standardized methods (Miake-Lye et al. 2016; Saran and White 2018). Nonetheless, organizations like the Collaboration for Environmental Science (CEE) and Campbell Collaboration are working to develop guidance documents for environmental SEMs and systematic reviews (Miake-Lye et al. 2016; Saran and White 2018; Evidence 2018a). The benefits of SEMs are the a priori protocols that ensure comprehensive literature searches while maintaining well-defined inclusion and exclusion criteria.

In contrast, scoping reviews can serve as precursors to systematic reviews as they provide a comprehensive review of the literature typically identified by a SEM. This tool also utilizes a priori protocols like SEMs, which aid in reproducible searches and additional data extraction (Munn, Peters, et al. 2018). However, while SEMs and scoping reviews are both structured summaries, their final results differ. As discussed in the previous paragraph an SEM provides a tabular or visual summary, while a scoping review will provide a detailed analysis of the studies identified and included. Additionally, neither scoping reviews nor SEM necessarily include an evaluation of study quality or risk of bias, as such evaluations are reserved for full systematic reviews when the study question is narrowed and a protocol for evaluating study limitations has been established. Finally, none of these tools are not to be confused with traditional narrative reviews, which are often subjective and expert-based (Golash-Boza 2015; Munn, Peters, et al. 2018). Overall, SEMs and scoping reviews can serve as a query tool to identify and characterize evidence for broad or research questions, thereby supporting greater objectivity and transparency (Rooney et al. 2014; James et al. 2016; Haddaway 2018; Munn, Peters, et al. 2018; Saran and White 2018; Wolffe et al. 2020).

In this work, we apply a systematic evidence mapping approach to comprehensively identify and integrate prevalence and occurrence data of legacy chemicals in Galveston Bay (GB)/HSC. We consider the GB/HSC estuary as the region has documented legacy contamination. We also aim to determine whether a baseline chemical dataset exists, whether there are any discernable spatial or temporal trends within available chemical data, and share summative visuals displaying available chemical literature. To our knowledge, there have been no other systematic evidence maps of a similar scope or ones interested in estuarine environments. This project provides an overview of relevant regions of concern in GB/HSC, as well as identifies challenges faced with integrating environmental data from multiple sources.

2. Methods

After Hurricane Harvey in 2017, this project was undertaken to understand available chemical data in GB/HSC. By identifying relevant chemical data, this project would aid researchers working to identify areas of vulnerability. A similar approach is relevant to other environments both in the United States and elsewhere in the world. Because of limited resources and time, we developed our primary question to capture a broad and unique scope within the GB/HSC. Therefore, we developed the following primary question: What is the historical spatial/temporal distribution of legacy contaminants in GB/HSC? Any secondary questions that may have arisen would have addressed any trends within the specific chemical classes.

To ensure our scope was both relevant and applicable, several experts from library sciences, oceanography, risk assessment, toxicology, environmental chemistry, and occupational health were engaged during this project. As this was a SEM, we also considered multiple grey literature sources in addition to databases which are described in Section 2.2.

2.1. Components of the Primary Question

To identify literature relevant to our primary question, we used the Condition, Context, and Population (CoCoPop) framework (Munn et al. 2015; Munn, Stern, et al. 2018b). This framework, which was developed in 2013 (Munn et al. 2015), fit the scope of this project because the approach addresses questions pertaining to the prevalence and occurrence. For this project, the GB/HSC estuarine system, served as the context. The chemicals served as the condition, and sediments served as the population. Table 1 expands on the elements comprising the elements of this project’s CoCoPop statement.

Table 1:

CoCoPop Statement elements with specific descriptions of the eligibility criteria used for inclusion and exclusion of articles during the screening and data extraction processes.

Primary Question Elements	Inclusion	Exclusion
Populations (P): Environmental Matrix	Environmental Matrix: Sediments are primarily analyzed, sediments are analyzed with other environmental matrices (water, air, fish tissue)	Label to use for Environmental Matrix: no sediment Study does not analyze any sediments
Condition (Co) Chemicals of Interest	Chemicals of Interest:  PAHs (polycyclic aromatic hydrocarbons):   Naphthalene; Phenanthrene; Anthracene; Fluoranthene; Chrysene; Benzo(b)fluoranthene; Pyrene; Perylene; EPA Priority 16 PAHs  PCBs (polychlorinated biphenyls):   Specific Individual Conegers (e.g. PCB 1–209); Specific Arochlor mixtures  OC Pesticides (organochlorine pesticides):   DDT p,p; DDT o,p; DDE p,p; DDD p,p; Heptachlor; Aldrin; Chlordane; Lindane; Mirex  Heavy Metals:   Arsenic (As); Copper (Cu); Mercury (Hg); Lead (Pb); Chromium (Cr); Copper (Cu); Nickel (Ni); Cadmium (Cd); Antimony (Sb); Iron (Fe); Trace elements; Zinc (Zn); Manganese (Mn)  Dioxins:   TCDD; 2,3,7,8-tetrachlorodibenzo para-dioxin; 2,3,4,7,8-PeCDF	Label to use: no chemicals of interest  Any other chemicals analyzed or other factors analyzed
Context (Co): Type of Environment Region	Type of Environment:  Studies must report estuary or wetland as a descriptor of the study region. Inner coastal tidal zones and bay descriptors may be considered as Galveston Bay may not always be labeled as an estuary Geographically relevant region:  Texas; regions of Gulf of Mexico near Texas; Galveston Bay; Matador Bay; Houston Ship Channel; San Jacinto Estuary; Morgan’s Point; Trinity Bay; Upper/Lower Galveston Bay These descriptors may be used in the abstracts and are specific areas of interest - general names of Galveston Bay and the Houston Ship Channel may be more frequently used.	Label to use for Type of Environment: wrong environment  No estuary or wetland study environments describe the study region (e.g. coastline, ocean, rivers, streams, lakes)  For example, a study in the Gulf of Mexico investigates sediments, but they were from an oil spill Label to use: wrong region  Not geographically relevant region within Texas or if the study is within the Gulf of Mexico it is not near Texas coast(s)
Study Design Status	Primary Data point(s) reported or trends analyzed	Label to use: review  Review Articles

2.2. Literature Search Methodology

Both peer-reviewed literature and grey literature were searched in the English language, with the primary searches occurring between June 2018 and July 2019. Additionally, several national and regional monitoring programs (e.g., from the Texas Commission on Environmental Quality (TCEQ) and National Oceanic and Atmospheric Administration (NOAA)) were identified in a search conducted from November to December 2020. The organics (PAHs, PCBs, Dx/F, and pesticides) and metals data reported in sediments were extracted in December 2020 from the following databases: NOAA Data Integration Visualization Exploration and Reporting (DIVER) (National Oceanic & Atmospheric Administration 2020); NOAA National Centers for Coastal Ocean Science (NCCOS) (National Oceanic & Atmospheric Administration 2017); Texas Commission on Environmental Quality (TCEQ) Surface Water Quality Web (SWQW) Reporting Tool (Texas Commission on Environmental Quality 2020a); and the Texas Clean Rivers Program (CRP) (The Texas Clean Rivers Program 2020).

The TCEQ data included in this project came from 18 TCEQ segments within the GB/HSC (1001D, 2428, 2430, 2425, 2430A, 2423OW, 1006, 1007, 1005, 2439, 1001, 2427, 2429, 2426, 2425A, 2422, 2421, 2424OW). These segments can be found using the two web tools cited in the previous paragraph, or they can be readily viewed using the TCEQ Surface Water Quality Viewer (Texas Commission on Environmental Quality 2020b). Of the 18 TCEQ segments identified, only 15 were used, as 1001D, 2423OW, and 2424OW contained no data. The data extracted from these two TCEQ programs were for individual metals and organics (e.g. PAHs, PCBs, and pesticides) reported in sediments.

Furthermore, the Texas A&M University OAKTrust, which is the digital repository for Texas A&M, and the ProQuest Dissertation & Theses Global, which is a global repository, were separately searched in December 2020. The theses/dissertations title/abstracts, publication year, and whether metals, organics, or both metals and organics were recorded. Data were extracted from documents that were no longer embargoed.

The scope and eligibility criteria for the initial search (Table 1) and corresponding data coding form via Google forms were established between November 2017 and June 2018. During this iterative process, article identification was limited if detailed search terms such as individual chemical names or a specific bay were used. In contrast, irrelevant articles were identified when broad search terms only considered ‘estuary’ or ‘sediment.’ The online software tool Rayyan aided in finalizing the scope and eligibility criteria (Figure S1).

In addition to searches of the peer-reviewed literature searches, general searches of the grey literature were conducted both in Google and through Carrot2.org. The purpose of this search was to identify additional articles or documents not readily found through bibliographic or library database searches (Evidence 2018b). The grey literature was screened using the same eligibility criteria as database articles, with the first 100 searches set as the cutoff. Documents after the first 100 diminished in relevance.

To diminish bias, two reviewers independently screened title/abstracts identified for both the peer-reviewed articles and grey literature. Disagreements were settled through discussions between the reviewers. After title/abstract screening, data was extracted from the peer-reviewed articles and grey literature by utilizing a predesigned Google form. The purpose of this form was to code all extracted data. To verify relevant data would be extracted, the Google form was tested using the following articles: Almukaimi (2016), Louchouarn (2018), Santschi (2001), Suarez (2005), and Yeager (2007). All identified literature were fully screened using this Google form.

Upon completion of the screening, the coded information was then imported to Microsoft Excel for further data visualization and analysis. There was one central spreadsheet summarizing the findings of initial search, and a secondary spreadsheet compiling both the NOAA and TCEQ data. The data used to identify possible trends were publication and sampling years, metal concentrations, organic concentrations, sampling site descriptors, and geocoordinates (if available). Table 2 summaries the search strategies used in MEDLINE (OVID), while Supplementary Table 1 summarizes the search string terms used for our search in EBSCO databases. Details regarding search strategies used for grey literature which include the NOAA and TCEQ databases and OAKTrust and ProQuest Dissertation & Theses Global are outlined in Supplementary Table 2.

Table 2:

The search strings used for the three components of the CoCoPop (Condition, Context, Population) statement are summarized along with the corresponding total hits retrieved from both Medline OVID and EBSCO. The hits associated with the specific databases within EBSCO are also included in the parentheses.

Database Search		# Hits
Search string: Population - Environmental Matrix	1. exp Geologic Sediments/ 2. sediment*.ti.ab 3. 1 or 2	Medline OVID: 183 retrieved EBSCO: 70 retrieved; 39 unique Agricola (27) Environmental Complete (15) Wildlife & Ecology Studies Worldwide (15) GreenFile (10) Academic Search Ultimate (8)
Search string: Condition – Chemicals of Interest	4. exp Polycyclic Aromatic Hydrocarbons/ 5. exp DIOXINS/ 6. exp Metals, Heavy/ 7. aldrin/ or ddt/ or dieldrin/ or endrin/ or exp heptachlor/ or lindane/ or mirex/ 8. Polycyclic Aromatic Hydrocarbon*.ti,ab. 9. (dioxin* or aldrin or ddt or dieldrin or endrin or heptachlor or lindane or mirex).ti,ab. 10. (heavy metal* or lead or cadmium or mercury).ti,ab. 11. (Naphthalene or Phenanthrene or Anthracene or Fluoranthene or Chrysene or Benzofluoranthene or Pyrene or Perylene or DDD or DDE or TCDD or tetrachlorodibenzo or PeCDF or (Organochlorine adj1 pesticide*) or (Chlorinated adj1 hydrocarbon*) or Chlordane or (Trace adj1 (element* or metal*)) or USEPA priority polycyclic aromatic hydrocarbons or Copper or Nickle or Lead or Cadmium or Aresnic or Antimony or Mercury or Iron or Zinc or Manganese).ti,ab. 12. or/4–11 14. 3 and 12
Search string: Context – Type of Environment/Region	15. exp Texas/ 16. exp “Gulf of Mexico”/ 17. (texas or galveston or houston or gulf of mexico or gulf coast).ti,ab. (southeast* adj1 (“united states” or america or states)).ti,ab. 18. or/14–17 19. 13 and 18

2.3. Eligibility Criteria

The CoCoPop framework guided how the eligibility criteria were defined in this project. Because the primary question focused on historically contaminated sediments, the three elements of the CoCoPOP statement were divided into five additional categories: (1) Environmental Matrix, (2) Chemicals of Interest, (3) Type of Environment, (4) Region, and (5) Study Design Status. By utilizing each of these categories, we aimed to identify relevant information while also ensuring that multiple studies could be considered during the screening processes. If the articles did not meet the inclusion criteria as defined in Table 1, they received an exclusion label (Table 1). Of the five categories, the ‘chemicals of interest’ possessed an additional five categories of chemical classes as listed in Table 1. Each of these chemical classes have had historical relevance in the GB/HSC.

Since authors often did not describe an estuary or explicitly describe the geographical region, both of which were listed under Context, alternative descriptors such as “inner coastal tidal zone”, “bay”, “regions of the Gulf of Mexico near Texas” or “Upper/Lower Galveston Bay” were used. When the title and abstract did not identify the geographic region, this information was obtained from the full text. As reviews can be biased and may not adequately consider historical references, this project excluded all reviews.

2.4. Article Screening

All peer-reviewed articles and grey literature were screened for relevance using the eligibility criteria defined in the previous paragraph. To facilitate the screening process, the free and accessible software tool Rayyan is utilized (Collaboration for Environmental Evidence 2019). Rayyan includes a “blind-mode,” which ensures that each reviewer independently screens articles, thereby avoiding biases. Through machine learning, Rayyan also utilizes a user’s inclusion and exclusion decisions to make the screening process faster (Kohl et al. 2018). Based on the eligibility criteria established for this systematic map, Rayyan also enabled both reviewers to use and reference the same set of labels (Table 1) to identify which articles they planned to include, planned to exclude, or were uncertain to whether to include. If the reviewers were uncertain whether to classify an article as included or excluded, they shared their reasoning in order to reach a consensus. Figure S1 of the supplementary materials summarizes the screening process.

2.5. Coding Strategy

The data for this systematic map was coded via a Google form, the results were saved in Microsoft Excel spreadsheets, and the spreadsheets were organized to be searchable and user friendly (James et al. 2016). Several categories were used to extract the data and they included article details (e.g., title, authors, publication year), chemical specific data (e.g., chemical name, concentration, concentration units), regional descriptors (e.g., latitude, longitude, sampling season), analytical techniques (e.g., instrumentation, extraction and clean-up, QA/QC), and outcomes (e.g., environmental health, chemical origins, data applications). This Google form was verified by testing the previously listed test articles. Additional commentary or observations made by the reviewer were annotated in paragraph format at the end of the Google form. Each reviewer verified the other’s coding and once the coded data were agreed upon, it was then extracted into Microsoft Excel to help assess for general trends. A list of the coding descriptors used can be found in Supplementary Table 3.

2.6. Statistical Analyses

All coded data extracted were inputted into Microsoft Excel where the concentration units were verified and then imported to GraphPad Prism 9.0.0. for further data analysis. Tableau 2020.3.3 was also used to visualize and map the prevalence of organics versus metals reported in GB/HSC. Analysis of variance (ANOVA) was used to determine variability between temporal and regional groups once all the raw concentration data were normalized through log-transformation. The statistical software, R (version 3.6.1), was used to create a graphic that compared publication year and sampling year for reported chemicals. The summary values from both the NOAA and TCEQ databases are contrasted with NOAA’s Screening Quick Reference Tables (SQuiRT) effects range low (ERL) and effects range median (ERM) for sediment quality guideline (SQG) references. The purpose of considering these two SQGs is to provide additional context relating to potentially bioavailable concentrations to the benthic community (Burgess et al. 2013).

3. Results

3.1. Number and type of articles, contaminants, and spatial data

After duplicate removal, a total of 487 articles and documents were searched; 423 were excluded using the eligibility criteria defined in Table 1. There were 55 included articles included for full text screening, where an additional 15 articles were added after separate secondary searches (Figure 1). After critical appraisal, 36 articles were included for data visualizations and concentration analyses; detailed summaries of appear in Table 3. Figure 1 further details all stages of the screening process and indicates why given studies were included and excluded with their reasons (Haddaway et al., 2018).

Figure 1:

ROSES flow diagram (Haddaway NR, Macura B, Whaley P, and Pullin AS. 2017). ROSES flow diagram for systematic reviews. Version 1.0. DOI: 10.6084/m9.figshare.5897389) shows the process in which all articles considered for the narrative analysis were selected for inclusion/exclusion during the screening and data extraction processes.

Table 3:

Summary of all studies included for the systematic evidence map. Each row describes the relevant categories for each individual reference used. References (Ref); Reference Type (Ref Type); Peer-Reviewed Article (Peer-Rev.); Dissertation/Theses (D/T); Grey Literature (Grey Lit.); Undergraduate Research Scholars Thesis (Undergrad); Matrix (Mx); Surface Sediment (Surf Sed); Sediment Core (Sed Core); Sediments (Sed.); Sampling Year (Samp Year); Concentrations Reported (Concen.); Averages (Av.); Background (Bkg); Descriptive Statistics (Descrip. Stat); Individual Chemicals Reported (ICR); Maximum (Max); Organics (Orgo); Sediment Quality (SQ); Not Reported (NR); Supplemental Table (Suppl Table).

Ref.	Ref. Type	Mx	Samp Year	General Sampling Location	Metals	Organics	Concen	Reported Outcome
Aguilar et al. 2014	Peer-Rev	Surf Sed. (<5 cm)	2010	San Jacinto Waste Pits/Channelview	-	PCBs (PCB77; PCB-81; PCB-105; PCB-114; PCB-118; PCB-123; PCB-126; PCB-156; PCB-157; PCB-167; PCB-169; PCB-189)	Av. in pg/g dry wt	Improving PCDD/F extraction method using the San Jacinto Waste Pit samples
Al Mukaimi et al. 2018	Peer-Rev	Sed. Core	2012; 2014	West Bay; East Bay; Texas City; Trinity Bay; Upper Bay; Clear Lake; Taylor Lake; Houston Ship Channel	Hg; Pb; Al; Ni; Zn	-	Av.; Bkg; Max. in ng/g	SQ Check; Environmental Policy; Focus on Hg
AlMukaimi 2018	D/T (PhD)	Sed. Core	2012; 2014	West Bay; East Bay; Texas City; Trinity Bay; Upper Bay; Clear Lake; Taylor Lake; Houston Ship Channel	Hg; select cores with Pb, Al, Ni, and Zn	-	Av.; Bkg; Max. in ng/g	Subsidence; Sedimentation; Hg Detection
Apeti et al. 2012	Peer-Rev	Surf Sed. (<3 cm)	2006–2007	Confederate Reef; Offatts Bayou; Ship Channel; Todd’s Dump; Yacht Club	Total Hg	-	Av in ug/g dry wt	Potential of Bioaccumulation; Monitoring Efforts
Carr et al. 1996	Peer-Rev	Surf Sed.	1992	Morgan Point; Jack’s Pocket; Eagle Point; East Bay; West Bay; Burnett Bay; Cedar Bayou; Trinity Bay; Kemah Flats; Texas City; Jones Bay; Chocolate Bay; Alexander Island; Black Duck Bay; Atkinson island; Swan Lake; Dollar Bay	Al, Ba, Be, Cr, Cu, Fe, Mg, Mn, Ni, Sr, V, Zn	Pesticides (aldrin, dieldrin, endrin, mirex, chlordanes, BHCs, DDTs) PCBs PAHs	Analysis NR	Sediment Quality; Risk Assessment; Exposure Assessment
Davis 2018	D/T (MS)	Surf Sed.	1992–2017	TCEQ Segments: 1005 (HSC); 1006 (HSC); 1007 (HSC); San Jacinto Bay (2427); Burnett Bay (2430); Upper GB (2421); Bayport Channel (2438)	-	EPA 16 PAHs: Nap; A; AE; AY; F; FL; P; PY; C; BaA; BaF; BkF; BaP; ghi; IP; DA	Descrip. Stat for 1992–1997; 1999–2002; 2004–2008; 2009–2013; 2014–2017 all in ug/kg	Temporal & Spatial Distribution of PAHs; Sourcing of PAHs; SQGs
Dean et al. 2009	Peer-Rev	Surf Sed. (<5 cm)	2002–2004	Houston Ship Channel (HSC)	-	Individual 18 PCDD/F	Median & Ranges in pg/g	Understand the Bioaccumulation of PCDD/F
Dobberstine 2007	D/T (MS)	Sed. Core	2004; 2005	lower Garpenter Bayou; lower Cedar Bayou; East Fork of Double Bayou; Robinson Bayou; Little Cedar Bayou	As; Cd; Cu; Pb; Ni; Sn; Hg; Zn	Organo-chlorine/ phosphorus pesticides; PAHs	ICR in mg/kg (metals) & ug/kg (orgo)	Identify a reference site within upper GB via evaluation of Sediment Quality Triad
Galveston Bay Estuary Program; TCEQ, USEPA, HARC 2019	Grey Lit.	NR	1973–2009	Houston Ship Channel, Trinity Bay, Upper & Lower Galveston Bay, East Bay, West Bay, Christmas Bay Complex	As; Ca; Cr; Cu; Pb; Hg; Ni; Ag; Zn	-	General Trends	Trends
Gardinali 1996	D/T (PhD)	Surf Sed.	1993	West Bay; East Bay; Lower GB; Upper GB; Trinity Bay; along the HSC	-	17 PCDD/F; Total PCBs (a focus on PCBs: 77; 81; 126; 169; 105; 114; 118; 123; 156; 157; 167; 189; 128; 138; 158; 166; 170);	ICR (pg/g); Total PCDD/Fs in pg/g; Total PCBs in ng/g	Bioaccumulation & distribution of halogenated aromatic hydrocarbons
HARC and Galveston Bay Foundation 2017	Grey Lit.	Sed.	2002–2016	Galveston Bay, Houston Ship Channel	Hg; Zn; Ni; Pb; As; Ag; Cu; Cr; Ca	Pesticides (DDT; Lindane; Dieldrin, Chlordane) PAHs (PY; AY; FL; AE; P; A; F; DA; BaP; C; Nap) PCBs (in general)	NR	Comparisons to Prior TECQ data
Hieke et al. 2016	Peer-Rev	Sed. Core (1–5 cm)	2006	inlet off Burnett Bay; Lower San Jacinto Bay; Negrohead Lake, San Jacinto Waste Pits; Beak Lake; Anahuac Channel; inlet by Kirby Inland Marine Oper Center; inlet south of Greens/ Baffalo Bayou split; inlet off entry to Buffalo Bayou; near Bay Shore Park; inlet off Tabbs Bay; Atkinson Island; transcect in GB	-	Sum of 18 PCDD/F	Av. in ng/g dry wt	SQ; Remedial Actions; Microbial Management & Trends
Howell et al. 2011	Peer-Rev	Surf Sed.	2002–2003	mouth of Patrick Bayou; near Patrick Bayou; main channel; near tributary; near Patrick Bayou/main channel	-	Sum of 209 PCBs	Av. in ug/g OC or ng/g OC	Sediment/Water Quality; Exposure Assessment; Bioaccumulation Factor
Kennicutt MC 2017	Peer-Rev	Sed.	1980s, 1991–1995, 1990–1997, 2000, 2001–2002, 2003–2006	-	Cr; Cu; Ni; As	Pesticides PAHs PCBs (general)	NR	SQ
Lakshmanan et al. 2010	Peer-Rev	Surf Sed. (<5 cm)	2002–2003; 2008	General HSC	-	209 PCBs 43 PCBs 18 PCBs	Av. in ng/g dry wt	Bioaccumulation; SQ; Risk Assessment; Environmental Monitoring
Leonard 2018	D/T (Undergrad)	Soils	2017	SJWP, Lynchburg Ferry landing, Burrnet Bay, Highland Acid Pit, French Limited, Sikes Superfund Site	T-Hg	Nap; A; AE; AY; F; FL; P; PY; C; BaA; BaF; BkF; BaP; ghi; IP; DA	ICR & Totals in ug/kg	Release & Remobilization after a natural disaster
Louchouarn et al. 2018	Peer-Rev	Sed. Core	2006	inlet off Burnett Bay; Lower San Jacinto Bay; Negrohead Lake; S. of I10 Bridge; San Jacinto Waste Pits; Beak Lake; Anahuac Channel	-	Individual 18 PCDD/F	Av. in pg/g	Fate & Transport (PCDD/F)
NOAA 2020/2017	Data-base	Sed.	1993, 1994, 1996, 2000–2006, 2010	Upper Galveston Bay; Lower Galveston Bay; HSC	See Supplemental Table 2	See Suppl.Table 2	mg/kg; ng/g; ug/g	Monitoring Data
Oziolor et al. 2014	Peer-Rev	Sed.	-	Houston Ship Channel (Vince Bayou, Patrick Bayou)	-	PCDD/F PCBs PAHs	NR	Microevolutionary Outcomes from Exposure
Oziolor 2017 (Embargo)	D/T (PhD)	-	-	-	-	-	-	-
Qian et al 2001	Peer-Rev	Sed.	1990, 1994	Ship Channel, Hanna Reef, Yacht Club, Todd’s Dump, Offatts Bayou, Confederate Reef	-	PAHs	Ranges and Mean in ng/g	NOAA Status & Trends (NS&T) Mussel Watch Program; Biota- Sediment Accumulation Factor (BASF)
Chatterjee R. 2007 (summary of Yeager et al. 2007)	Grey-Lit	NR	-	Upper HSC	-	PCDD/F (general)	NR	Commentary for Dioxin Contribution Sources
Santschi et al. 2001	Peer-Rev	Sed.	1995	Trinity Bay	Pb; Ba; Hg; Cd	Sum 24 PAHs Sum 18 PCBs Sum DDTs	ug/g	Comparison to Natural Background Levels
Seward 2012	D/T (MS)	Sed. Core	2004	Galveston Bay; HSC; lower San Jacinto River; lower Trinity River floodplain	-	Individual 17 PCDD/F; Total PAHs	Dx/F in pg/g Total PAHs in ng/g	Historical contamination; fate & sourcing; redistribution; sorption & bioavailability
Simons et al. 2009	Peer-Rev	Sed	2000–2004	Galveston Bay	Pb; Hg; Zn; As	PAHs	Averages in ug/g	Resource for Ecological Conditions
Suarez et al. 2005	Peer-Rev	Surf Sed. (<5 cm)	2002–2003	Upper Galveston Bay/HSC	-	Individual 18 PCDD/F	TEQs per kg dry wt;	Trends & Status of PCDD/F
Suarez et al. 2006	Peer-Rev	Surf Sed. (<5 cm)	2002–2003	Upper Galveston Bay/HSC	-	Individual 18 PCDD/F	-	Sediment Flux
TCEQ 1994	Grey Lit	Sed.	1992, 1993	-	Cu; Zn; Hg; Pb; Cr	DDT Av. PCBs	Ranges in ug/kg	Trends; Sediment Quality
TCEQ SWQW 2020	Data-base	Sed.	1990–2019	Upper Galveston Bay; Lower Galveston Bay; HSC	See Suppl Table 2	See Suppl. Table 2	ug/kg; ng/g; mg/kg	Monitoring Data
University of Houston-Clear Lake and the University of Houston Houston, Texas 2003	Grey Lit	NR	-	-	-	PAHs	Zhang et al. paper highlighted	Reports & Trends
Wei 2016	D/T (PhD)	Surf Sed.	2001–2010	TCEQ Segments 1005 (HSC); 1006 (HSC); 1007 (HSC)	Pb; Cu; Hg; Zn	-	Av. in mg/kg (unclear if this value was per year or by station)	Spatio-temporal water & sediment distributions; seasonal variation of air pollutants; pollutants & health outcomes
Yeager et al. 2007	Peer-Rev	Sed. Core	2006–2007	inlet off entry to Buffalo Bayou; near Bay Shore Park; inlet by Kirby Inland Marine Oper Center; inlet off Buffalo Bayou near BB Toll Bridge; inlet south of Greens/Baffalo Bayou split; terrestrial control off of Trinity River; inlet off Upper San Jacinto Bay; inlet off Tabbs Bay	-	PCDD/F	Av. in ng/kg dry wt	Sedimentary Processes & Flux
Yeager et al. 2010	Peer-Rev	Surf Sed. (3–4 cm)	2007	Hog Island, Alexander Island	-	PCDD/F	Av. in pg/g	Remedial Actions
Yuill 1991	D/T (PhD)	Sed.	1987; 1988	-	-	-	-	Not included for analysis due to sampling occurring prior to 1990 and no focus on metals
Zhang et al. 2003	Grey Lit.	NR	-	-	-	PAHs	NR; Flux Rates Calculat-ed	Sediment Flux; Risk Management Strategies

In total, 18 articles reported metals, 31 reported organics, and 24 reported both. Of the studies reporting quantitative data, 7 were peer-reviewed articles that included latitudes and longitudes; NOAA and TCEQ databases also included latitudes and longitudes. The remaining 8 quantitative studies/reports used general location descriptors (e.g., Bear Lake, Lower San Jacinto Bay, etc.). Given the mix of exact geocoordinates and general location descriptors, the three general regions of HSC, Upper GB, and Lower GB were used to discern any spatial patters in the chemical data (Figure 2). In the peer-reviewed articles, there was a distinct difference between the frequency of metal reporting versus organic reporting. For example, most metals data were spread throughout each of the three regions, but there was more diversity in Lower GB than the HSC and Upper GB (Figure 2a). Sample numbers also varied for each research agency. The samples analyzed for organics for instance, were mainly in the Upper HSC with a cluster noted in the San Jacinto Waste Pits (SJWP) (Figure 2b). Both NOAA and TCEQ monitoring programs reported throughout GB/HSC (Figure 2c–e).

Figure 2:

The maps summarize the general regional locations for metal (a) and organic (b) concentrations reported in the peer-reviewed literature and grey literature. The TCEQ samples analyzed for metals (c) and for organics (d) show additional sites to those illustrated in (a) and (b). The NOAA metals and organics data shared the same geocoordinates are therefore are all on the same map (e). All circles designate either general sampling site (a, b) or specific sampling sites as not all publications shared detailed geocoordinates (c,d,e).

The peer-reviewed articles contained data with numerous metals (Ag, Al, As, Ba, Be, Cd, Cr, Cu, Fe, Hg, Mg, Mn, Ni, Pb, Sb, Sn, Zn), with Hg being the most commonly reported metal, and PCDD/F the most common organic reported. Averages or ranges were sometimes included for individual metals (Texas Commission on Environmental Quality 1994; Santschi et al. 2001; Simons and Smith 2009; Apeti et al. 2012; Almukaimi 2016; Wei 2016; Leonard 2018; Al Mukaimi, Kaiser, et al. 2018), while other articles described observed trends (HARC and Galveston Bay Foundation 2017a; Kennicutt II 2017a; Galveston Bay Estuary Program; TCEQ, USEPA 2019)(HARC and Galveston Bay Foundation 2017a; Kennicutt II 2017b). In contrast, both NOAA and TCEQ published extensive metal and organic analytes (Supplemental Table 4). Additionally, we noted two of the eight dissertations/theses included in this study, used the same TCEQ database described earlier in the methods section (Wei 2016; Davis 2018). Therefore, the raw TCEQ data was included for these two studies rather than their averages or descriptive statistics. All other dissertations/theses contained original data.

The PAHs or PCBs are reported in the peer-reviewed articles/reports, they appeared either as general totals or specific groupings. For PAHs, the peer-reviewed articles commonly reported total levels of PAHs as averages or summed certain PAH analytes (Gardinali 1996; Qian et al. 2001; Santschi et al. 2001; Simons and Smith 2009; Seward 2010; Davis 2018; Leonard 2018), whereas individual PAHs or general PAH trends were described but not numerically represented (Carr et al. 1996; Zhang et al. 2003; University of Houston-Clear Lake and the University of Houston Houston 2003; Oziolor et al. 2014; HARC and Galveston Bay Foundation 2017b; Kennicutt II 2017b). Shorthand abbreviations for PAHs used throughout this manuscript appear in Supplementary Table 3.

Data on PCBs also were reported; however, peer-reviewed articles commonly reported the sum of all 209 PCB congeners (Lakshmanan et al. 2010; Howell et al. 2011), 18 PCBs (Santschi et al. 2001; Lakshmanan et al. 2010), and 43 PCBs (Lakshmanan et al. 2010), or included the general discussion of the presence of PCBs (Texas Commission on Environmental Quality 1994; Carr et al. 1996; Oziolor et al. 2014; HARC and Galveston Bay Foundation 2017a; Kennicutt II 2017b). Only one study included data on several individual PCBs (Aguilar et al. 2014).

The only pesticide found to be quantitatively reported in the peer-reviewed literature was DDT (Texas Commission on Environmental Quality 1994) or the sum of DDTs (Santschi et al. 2001). Four peer-reviewed articles included information about completed analyses or qualitative descriptions of detecting the following pesticides: aldrin, dieldrin, endrin, mirex, chlordanes, BHCs, DDTs, or lindane) (Carr et al. 1996; HARC and Galveston Bay Foundation 2017a; Kennicutt II 2017b).

As PCDD/F were the most common organics, most peer-reviewed articles contained data for 18 individual PCDD/F (Suarez et al. 2005; Suarez et al. 2006; Dean et al. 2009; Louchouarn et al. 2018) or the sum of all 18 individual PCDD/F (Yeager et al. 2007; Yeager et al. 2010; Hieke et al. 2016). Only one article included qualitative descriptions of PCDD/F trends within GB/HSC (Rhitu 2007), which also referenced Yeager et al.’s (2007) findings. For articles containing data on PCDD/F in GB/HSC, the outcomes were often related to sediment quality, bioaccumulation, fate, transport, and trends within a given time period. Sediment quality outcomes were common for both metals and organics data, but ecological monitoring was often included for metals data discussion. Additional summary statistics for the peer-reviewed articles and reports are summarized in Supplemental Tables 5 and 6.

In contrast to the peer-reviewed articles and reports, the two national and regional agencies, NOAA and TCEQ, reported between 89 to 280 individual chemicals. For some chemical classes additional totals were included. For example, some PCB congeners were grouped together (e.g., PCB 13/12, PCB 59/62/79, PCB 61/70/74/75) versus a summative total of all 209 congeners. To understand the diversity of the individuals detected and their corresponding concentration ranges, the summary statistics for both NOAA and TCEQ are summarized in Supplemental Tables 7–13.

The search of published dissertations/theses showed that research on chemical contamination of both metals and organics exists in undergraduate (Leonard 2018), master’s (Dobberstine 2007; Seward 2010; Davis 2018), and doctoral research (Yuill 1991; Gardinali 1996; Almukaimi 2016; Wei 2016; Oziolor 2017). Of these dissertations/theses, four contained data on metals (Yuill 1991; Almukaimi 2016; Wei 2016; Leonard 2018), five contained data on organics (Gardinali 1996; Seward 2010; Oziolor 2017; Davis 2018; Leonard 2018), and one contained research on both select metals and organics (Dobberstine 2007).

Of the nine dissertations/theses included, one had an embargo; therefore, only this document’s title/abstract were only considered for inclusion (Oziolor 2017). Another dissertation/thesis was excluded based on the full-text screening, as the sampling years occurred before 1990 (Yuill 1991), although, this study did provide an interesting historical timeline for significant events within GB/HSC (e.g., 1900 hurricane, formation of petroleum companies in the HSC, dredging).

Two of the remaining eight included dissertations/theses contained data from publications already considered within this project (Seward 2010; Almukaimi 2016). For example, in Seward (2010), the data included in this thesis had been previously published by Yeager et al. (2007). Thus, the data included within these dissertations/theses were not re-extracted. For the other dissertations/theses included in this study, data containing totals or individual chemical data for PCDD/F and Hg were extracted. As two dissertations/theses did not provide PCDD/F or Hg data, their summary statistics were reported in Supplementary Tables 14–16 (Gardinali 1996; Dobberstine 2007).

To capture reporting trends for both metals and organics reporting trends, the heatmap in Figure 3 compares sampling years and publication years based on whether metals, organics, or both were sampled. Two hurricanes, Hurricane Ike and Hurricane Harvey, mark two significant environmental events within Galveston Bay. Therefore, samples collected before either event are useful to consider for potential chemical spatial and temporal trends. In many instances data came from samples collected 2–5 years before the article was published (Figure 3). For instance, Agular et al.’s (2014) sample collection occurred in 2010, while their results were published in 2014. In contrast to this collection versus publication difference, Yeager et al. (2007) collected samples in 2006 and 2007, with multiple publications utilize these same samples (Louchouran et al. 2018 and Heike et al. 2016). Conversely, some dissertation/theses also confirmed the peer-reviewed data they had published or utilized (Seward 2010; Almukaimi 2016), while other dissertation/theses contained data from the same databases identified in this project (Wei 2016; Davis 2018).

Figure 3:

Heatmap illustrates the sparsity of reported sampling times in comparison to the publication year. The colored boxes and the colored circles signify whether metals (purple), organics (red), both metals and organics (yellow) or no chemicals were reported (black). All references shown in this figure are from the database and grey literature search with most of the publication originating from peer-reviewed journals.

The results discussed in this section indicate that, although there are historical sediment sample data are available, the timeframes in which they were collected are inconsistent and data on metals and organics is sparse. However, there appears to be more GB/HSC organics data available through NOAA and TCEQ, while there is limited organics data contained in peer-reviewed articles. A similar trend is observed for the metals data. Considering the sample general locations (Figure 2) and the heatmap (Figure 3) together indicates that data for any individual chemical or chemical class are sparse both spatially and temporally.

3.2. Trends for Chemical Concentrations

Chemical data differed between the peer-reviewed articles and the NOAA and TCEQ databases. Some of these differences may be attributed to variable analytical and sampling techniques. For concentration specific analysis, these two chemicals were investigated for their variability across time and within the HSC, Upper GB, and Lower GB regions (Figure 4). Hg levels reported in the Upper GB varied compared to Hg levels in Lower GB. However, a nested one-way ANOVA indicates that no true variability in Hg levels exists between time groups (p=0.63), although it did reveal variability between each regional group (p<0.0001; Chi-square, df: 104.6,1). To confirm this test, sampling time was ignored, and the Hg data was grouped based on the region (HSC n=260; Upper GB n=334; Lower GB n=161); a one-way ANOVA verified the nested one-way ANOVA regional variability (p<0.0001; R²=0.1474).

Figure 4:

Boxplots contrast sampling year versus the concentration of Hg (a) and Dx/F (b) from all data extracted from peer-reviewed literature, grey literature, and databases. Sampling year groups were further grouped by general region (HSC, Upper GB, Lower GB) where results were analyzed by t-test or ANOVA.

From 1990–2019 the general median Hg values ranged between 1.6–2.0 on a log10 scale(ng/g). There were two time periods, 2000–2004 and 2005–2009 that show extreme Hg concentration values and this is likely due to samples collected in Patrick Bayou. This bayou is a known point source for Hg in the HSC and has been sampled both by researchers from the peer-reviewed literature (Al Mukaimi, Kaiser, et al. 2018) and TCEQ, but not NOAA. Additionally, the site is located in an industrialized section of the HSC and is also a Superfund site, which is a hazardous waster site listed under the Comprehensive Environmental Response, Compensation and Liability Act (CERCLA).

If the NOAA data alone is considered, the median Hg concentration for this dataset is 0.042 mg/kg, which is well below the ERL and ERM SQG values of 150 and 170 mg/kg. The same observation is seen in the TCEQ data, except the median Hg concentration is 0.0826 mg/kg. While, sampling techniques and instrumentation will vary between these two-monitoring program, the data presented here provide a general range for Hg within GB/HSC to consider. Interestingly only four peer-reviewed articles reported Hg data (Santschi et al. 2001; Simons and Smith 2009; Apeti et al. 2012; Al Mukaimi, Kaiser, et al. 2018), while the rest of the Hg data came from NOAA and TCEQ that spanned several years (NOAA: 1993–96; 2000–06; 2010 and TCEQ: 1990–2019).

In comparison to the Hg reported within GB/HSC, PCDD/Fs appeared to vary in distribution throughout the region and across time (Figure 4b). However, a nested one-way ANOVA indicated variability between each time group (p=0.0348; F, Dfn, Dfd = 4.445, 4, 8) and each regional group (p<0.0001; Chi-square, df: 346.5, 1). To confirm the regional variability, sampling time was ignored and the PCDD/F data were grouped based on region (HSC n=4,606; Upper GB n=873; Lower GB n=1784). A one-way ANOVA verified the nested-one-way ANOVA regional variability (p<0.0001; R²=0.03459).

There is a notable lack of PCDD/F data prior to 2000, but more data was reported from 2000–2014 (Figure 4b). Of the data reported nine peer-reviewed articles (Aguilar et al. 2014, Hieke et al. 2016, Lakshmanan et al. 2010, Louchourarn et al 2018, Santschi et al. 2001, Simons and Smith 2009, Suarez et al 2006, Yeager et al 2010, Yeager et al. 2007) contained data from 1995, 2000–2003, 2005–2006, and 2010. NOAA only contained data from 2006 and Gardinali (1996) contained data from 1993. Given the San Jacinto Waste Pits (SJWP) are a known point-source of dioxins, most of the peer-reviewed articles research this site. In contrast the two monitoring programs assessed the general GB/HSC region.

In contrast to the Hg median concentration values, the PCDD/F medians varied within each time period as well as each region. For example, on a log10-scale (ng/g) the period between 2000–2004 had a general median PCDD/F concentration of −1.921 while 2010–2014 had a median value of −2.621. The wide ranges observed in each region, with the HSC and the Upper GB the most common regions in which PCDD/F data was reported. Unlike the Hg data, the PCDD/F data ranges showed wider variability (Figure 4b), which can be attributed to different sampling locations, with many articles focusing within the SJWP site.

The Hg data were divided based on the following two timeframes to assess for variability between two distinct time periods: 2005–2009 HSC and 2010–2014 HSC. A two-tailed t-test indicated these time groups did not differ (p=0.23), which implies that from 2005–2014 most Hg concentrations detected in both regions may have remained the same across time. The same temporal grouping was applied for Hg in Upper GB, but neither group differed (p=0.8693).

The PCDD/F data was also divided into the same temporal and regional groups with the HSC and Upper GB regional grouping means differed between the two timeframes ((HSC: p<0.0001; t=5.989, df=1121); (Upper GB: p<0.0001; t=4.839, df=649). These results imply the PCDD/F concentrations from 2005 to 2014 were not consistent. Thus, overall PCDD/F appeared to be in a state of flux, with the concentrations varying between regions and across time.

Both Hg and PCDD/F remained the most prevalent chemicals studied, despite diverse chemicals analyzed through monitoring programs and peer-reviewed articles. Santschi et al. (2001) and Simons and Smith (2009) reported on both PAH totals, and varying PCBs (e.g., 209 PCBs, 43 PCBs, 18 PCBs, or certain individual PCBs) being reported by Aguilar Lakshmanan et al. (2010), Santschi et al. (2001), and Simmons and Smith (2009). Only Santschi et al. (2001) reported on total DDTs, and both NOAA and TCEQ reported additional pesticides (Supplementary Table 4).

Additional descriptive statistics for metals and organics contained in peer-reviewed articles appear in Supplementary Tables 5 and 6. The NOAA and TCEQ data for individual metals appear in Supplementary Tables 7 and 8, and data for individual PAHs, pesticides, Dx/F, and PCBs are reported in Supplementary Tables 9–13. In each supplemental table, available SQG values, ERL and ERM, are also listed. When observing reported concentration ranges and interquartile ranges for each dataset considered, they were all below the SQG guidelines. Overall trends for the GB/HSC region are difficult to discern for both metals and organics, despite diverse chemical profiles and multiple sampling years.

4. Discussion

4.1. Major Findings & Knowledge Gaps

As SEMs expand more into environmental sciences and environmental health, they will continue to provide unique insights to broad research questions. A few relevant research questions may consider the following topic areas: role of monitoring programs for chemical exposures, the likelihood of using an intervention to mitigate a particular exposure or address additional exposures relevant for disaster response research (DR2). By identifying relevant knowledge clusters, both historical data can be referenced and knowledge gaps can be identified for investment into additional research (Bilotta et al. 2014; Aiassa et al. 2015; James et al. 2016; Miake-Lye et al. 2016; Munn, Peters, et al. 2018; Munn, Stern, et al. 2018b; Saran and White 2018; Wolffe et al. 2020). Therefore, SEMs can facilitate answers to research questions by addressing the fate, transport, biotransformation, and exposure of chemicals so a clean-up or reference value could be established.

Estuaries are only one common environment across the globe; thus, the CoCoPop methodology presented in this project can be applied to other environments, other pollution related questions, and DR2 questions. Two studies in which SEMs have been applied regard the polar environment (Mangano et al. 2017) and freshwater systems with microplastics (Yao et al. 2020). These two studies illustrate the utility of SEMs in uncovering relevant knowledge gaps while also improving reporting methods for environmental research. However, the current project showed that some regions, although historically contaminated, may lack consistent sampling data. With sparse environmental background data, disaster research considering before/after effects or exposures within a given environment will be limited. One area that could therefore be affected is fate/transport modeling and pathway analysis. Consequently, the findings in this study establish baseline references as well as identify relevant historical articles on environmental conditions in GB/HSC.

The results of this systematic evidence mapping show that available legacy contaminant data in the GB/HSC region are limited in terms of spatial, temporal, and chemical coverage, particularly within peer-reviewed articles. Even after dividing the data into three broad geographical regions (HSC, Upper GB, Lower GB) and 5-year intervals, there was no chemical class consistently sampled; however, when the monitoring databases were considered, additional data on metals and organics were recorded for future reference.

While there were notable data ranges for Hg and PCDD/F in particular (Figure 4), the two time periods of 2005–2009 and 2010–2014 were observed to be distinct for PCDD/F data. In contrast, the Hg data was variable between regions rather than across time. The other individual metals and organics medians varied between each article as illustrated in Supplementary Tables 5–16. Some of the differences observed between articles are tabulated in Table 3, where we saw select lists of chemicals were of interest. In some cases, chemical selection may have been determined based on what the study was interested as a reported outcome (e.g., sedimentary processes, remedial actions, fate and transport, bioaccumulation). The observed differences could also be to different sampling techniques, instrumentation, and analytical methods. In some cases, the different publication years may be indicative to which methods were utilized as more recent years may use updated analytical methods or newer instruments.

Another rational for inconsistent chemical analysis is limited resources since sample collection and sample analysis are resource intensive. For example, Dx/F analysis roughly costs $525 per sample (Personal communication). As a result, the peer-reviewed articles considered in this SEM may have had finite resources, which limited the number of chemical analyses conducted. In contrast to the peer-reviewed articles, both NOAA and TCEQ considered extensive chemical lists. Perhaps due to differing resources and a specific interest in monitoring, both agencies were able to consider additional chemical analyses compared to academic researchers.

Despite these differential chemical reporting, the most common chemicals recorded are Dx/Fs and Hg; however, this observation was common for peer-reviewed articles. In contrast to the peer-reviewed articles, both NOAA and TCEQ regularly reported both organic and metals data in certain timeframes (e.g., early 2000s, late-2000s, or early 2010s). There is still a knowledge gap for reported legacy contaminants in GB and HSC and therefore limited knowledge regarding concentration distributions of multiple chemicals in this region. Although monitoring databases were not considered in the original search strategy, the data reported by NOAA and TCEQ provided temporal and spatial data comparable to data from the peer-reviewed articles. Thus, the data presented here only provides an overview of regions within GB/HSC that may be of interest for additional studies. Our results also identify that inconsistent reporting methods of chemical data within GB/HSC remains a knowledge gap in the region.

With baseline data often of interest for pre-/post- comparisons, datasets containing comprehensive physio-chemical values before a natural disaster or seasonal flood are highlight sought after. A recent example that sought to develop such a dataset occurred in Sydney, Australia’s estuarine system (Birch and Lee 2018). As there was already an interest in developing baseline data for Sydney, resources were readily available for the level of monitoring data required. However, in many environmental studies, researchers may face difficulties regarding site access, sampling trip costs, shifting environmental conditions, or even emerging contaminants. The limited analyses for the peer-reviewed articles also posed a problem for establishing any relevant baseline or reference database in this study. Similarly, if researchers focus on individual chemicals rather than how they interact as environmental mixtures or in combination with other chemicals, a limited understanding remains for exposure risk. Therefore, with estuaries serving as the final sink for many anthropogenic inputs, there remains a need to characterize chemicals and their metabolites from a systems biology standpoint (e.g., adverse outcome pathways) (Cuevas et al. 2018) as well as from a mixtures standpoint.

4.2. Implications for Research, Management, Policy, and Practice

For biomedical sciences, guidelines on systematic reviews and evidence mapping have been developed over many decades (PRISMA 2015; Cochrane Library 2020). These guidelines encourage authors to provide as much information as possible related to the methods, materials and study design of the items reviewed. In contrast, no such protocols existed until the mid-2010s for environmental management and environmental sciences (Haddaway et al. 2018; Haddaway and Macura 2018; Macura et al. 2019). Through continued efforts by organizations such as the Collaboration for Environmental Evidence (CEE), use of systematic reviews and systematic maps as tools in environmental research continues.

The current systematic evidence map demonstrates the need to use more uniform environmental parameter reporting for sediments and other environmental media. Although chemical concentration data and geocoordinates are sometimes reported, these values were not consistently reported and there were data gaps for chemical classes. This inconsistent reporting and inadequate use of topic labels within environmental health sciences is a common problem (Randall et al. 2015; Bernes et al. 2017; Mangano et al. 2017; Behnke et al. 2020; Yao et al. 2020; Nevalainen et al. 2021). Therefore, in future environmental publications, specific key terms should be used in titles and abstracts to promote database searchability. Use of CEE guidance documents for environmental science research will also continue to improve reporting methodology. Another tool that could improve SEMs are “knowledge graphs,” which are interactive visual summaries of available research (Wolffe et al. 2020). These graphs in turn could improve reporting methods, the recording of data, and the visibility of SEMs for future research questions related to policy, management, and research practices.

To aid and supplement reporting efforts, especially for baseline and historical considerations, sediment core data also needs to be incorporated into environmental assessments. Since many estuarine sites are located within active industrial areas, there are often limited sites that remain undisturbed due to activities such as dredging and land development. Sedimentation rates are also variable within industrial areas (Al Mukaimi, Dellapenna, et al. 2018) as well as in areas prone to flooding events (Dellapenna et al. 2020; Owca et al. 2020). Therefore, a sediment core needs to contain fine-grained sediment, be from a site that remains undisturbed, and have a moderately quick sediment rate (Valette-Silver 1993). These factors combined help ensure a sediment core can be evaluated for historical trends and contamination levels.

Depending upon the region, sediment within a shipping channel may maintain SQG levels since frequent dredging occurs thereby limiting the presence of contaminants (Chen et al. 2016). In contrast, flood events such as Hurricane Harvey, can redeposit sediment (Dellapenna et al. 2020), while regular flood events can suspend sediments with limited movement (Owca et al. 2020). Due to these variable conditions, sediment cores collected within a marine or estuarine environment can provide a stable, chronological, and historical reference (Valette-Silver 1993; Chatterjee et al. 2007; Yeager et al. 2007; Louchouarn et al. 2018; Al Mukaimi, Dellapenna, et al. 2018; Owca et al. 2020).

If more studies included sediment cores, then improvements can continue to be made in monitoring programs seeking to reference historical contamination records (Roach and Walker 2017). Collectively however, uniform environmental reporting methods, additional use of index terms, and incorporation of sediment core data can further our understanding of historical contamination in marine and estuarine environments. Use of evidence maps as tools summarizing available evidence can also facilitate collection of relevant data for various environments with contamination issues.

4.3. Limitations of the Search, Evidence, Strengths

Like any search strategy, ours may have missed some studies with relevant information. When developing the scope of this study, we found general terms such as “Gulf of Mexico” to be too broad, but if we specifically searched for “Houston Ship Channel,” there were few studies. To ensure that we were broad yet specific enough, we search terms such as “Texas,” “Galveston,” “Gulf Coast,” and “United States” to name a few. However, even with these search terms, we may have missed articles that did not contain any of these terms in their titles, abstracts, or lists of key words.

How authors described or listed the legacy contaminants in their titles and abstracts also limited our search. In most cases, the general classes of dioxin/furans and pesticides as well as acronyms of PAHs and PCBs were enough to identify articles. Yet, some articles may have listed the legacy contaminant differently or not at all within their titles, abstracts, or key words; this situation appeared to occur with earlier publications before 2006. In this instance, this difference in article searchability could potentially be attributed to shifting character counts for the title page and abstracts. If authors were to used common terminology and specify regional and specific locations in their keywords, more studies and reports could be readily identified for future reference.

Another limitation in this SEM was the level of detail provided by included articles. For instance, articles often contained limited geospatial information related to their samples and only a few included relevant general regional descriptors (Yeager et al. 2007; Apeti et al. 2012; Aguilar et al. 2014; Hieke et al. 2016; Louchouarn et al. 2018). However, the monitoring programs data reported by NOAA and TCEQ included specific geocoordinates and relevant GB/HSC regional descriptors. As DR2 is likely to require spatial analysis for assessment of both pre- and post-disaster conditions, current researchers should consider how they report their sample geocoordinates. Although when environmental samples are taken near private lands or near commercial agricultural areas, general descriptors of the sampling area will be a useful substitute. This alternative geographical description would consequently protect any private property that may be implicated by a study’s research findings. Overall, if one understands chemical distribution in urbanized estuaries aside from GB/HSC, then the effects of local natural disasters within these ecosystems could be studied further.

Another limitation of this study was the update made for the Carrot2.org search engine prior to this systematic evidence map report. When the search terms used for both Google and Carrot2.org were put into the newest Carrot2.org version, the search strings used yielded fewer hits than the search recorded in this study. Should this SEM be redone, a specific online search within the grey literature will be developed to identify relevant theses, reports, white papers, and other grey literature. However, this review detailed and documented internet searches (Evidence 2019), with all grey literature having a full text document retrieved. As additional SEM and systematic review tools become available (Kohl et al. 2018), this SEM could be updated in the future.

Finally, as with any search strategy, we cannot guarantee 100% coverage of all published studies. For instance, due to resource constraints, we did not screen additional databases, such as Web of Science or Aquatic Sciences and Fisheries Abstracts, which would have added an additional 1600 references for screening. Given that the vast majority of the quantitative data we located is from NOAA and TCEQ, we felt that the diminishing return from searching additional databases would not have warranted the substantially increased level of effort required.

5.0. Conclusions

This systematic evidence map (SEM) presented a survey of available legacy contaminant data for Galveston Bay (GB) and the Houston Ship Channel (HSC) sediments. We identified relevant grey literature in addition to peer-reviewed articles from databases, as the former may have contained additional historical data. We categorized the data by location and time, as well as by types of contaminants and sample matrices, and aimed to identify relevant trends if present. The extracted data contained few trends to note, largely because of the inconsistent sample collection timeframes and lack of sample location diversity. Most of the literature reported on dioxins/furans (Dx/F) and mercury (Hg), but few data consistently recorded other organics and metals. This gap can be addressed by using the NOAA and TCEQ data, but specific chemicals may need to be examined separately to discern any trends.

Additionally, several peer-reviewed articles did not consistently record chemical data which resulted in data sparsity for certain regions of GB/HSC. In turn, this data sparsity makes it difficult to understand how environmental events, such as Hurricane Ike and Harvey, or activities such as dredging may have historically affected contaminant distribution in GB/HSC. Some of these data gaps may be addressed through monitoring data provided by regional and national agencies such as NOAA and TCEQ. Future researchers studying the GB/HSC should continue addressing the following knowledge gaps: frequency of contaminant reporting, spatial data collection, and sample sizes.

Additionally, future SEMs and systematic reviews of environmental contaminants would benefit from more consistent use of key words with titles and abstracts for search identification. More uniform reporting of geolocations and chemical data would also be beneficial. This project also helps establish initial efforts to characterize and understand available historical data regarding estuaries, which are one of many environments that SEMs could be applied to. Therefore, this project shows not only where reporting methods can improve for environmental studies, but also how other researchers could use similar approaches as presented here to expand the global understanding of chemical contamination in marine and freshwater environments. With support from the scientific community, improving reporting methods can also lead to the creation of standardized datasets and protocols. With these standards in place, data-driven studies can follow along with expanded applications of systematic review and systematic mapping protocols for environmental science and DR2 research.

Highlights.

• Characterizing baseline environmental chemical contamination remains a challenge.

• Systematic evidence mapping can be a tool for understanding baseline data gaps.

• Our study of an urban estuary found sparse data both spatially and temporally from peer-reviewed articles.

• Our results suggest a need of more uniform chemical sampling and data reporting methods.