California Water Quality Research Articles

Machine learning for environmental justice: Dissecting an algorithmic approach to predict drinking water quality in California

The potential for machine learning to answer questions of environmental science, monitoring, and regulatory enforcement is evident, but there is cause for concern regarding potential embedded bias: algorithms can codify discrimination and exacerbate systematic gaps. This paper, organized into two halves, underscores the importance of vetting algorithms for bias when used for questions of environmental science and justice. In the first half, we present a case study of using machine learning for environmental justice-motivated research: prediction of drinking water quality. While performance varied across models and contaminants, some performed well. Multiple models had overall accuracy rates at or above 90 % and F2 scores above 0.60 on their respective test sets. In the second half, we dissect this algorithmic approach to examine how modeling decisions affect modeling outcomes – and not only how these decisions change whether the model is correct or incorrect, but for whom. We find that multiple decision points in the modeling process can lead to different predictive outcomes. More importantly, we find that these choices can result in significant differences in demographic characteristics of false negatives. We conclude by proposing a set of practices for researchers and policy makers to follow (and improve upon) when applying machine learning to questions of environmental science, management, and justice.

Investigating the relationship between agricultural water usage and water quality in California

California has a highly developed agriculture system, but it faces many challenges associated with water supply and water quality. This research examines the relationship between agricultural water use and water quality at multiple scales, including a comparison of southern and northern California, analysis of seven regions, and four counties within the Central Valley, which accounts for more than 70% of agricultural water use in the state. Statistical analyses of georeferenced data from USGS and CIMIS were used to test the hypothesis that higher agricultural water use is associated with lower water quality. Results indicate that while there is no significant difference in water quality between southern and northern California, there are significant differences among regions. Furthermore, within the Central Valley, there is a significant inverse relationship between agricultural water use and two of three water quality indicators, namely dissolved oxygen and dissolved nitrate. Although further research is necessary to establish causality, these findings suggest that policymakers need to consider the effects of water use on water quality in planning California’s future agricultural development.

Efficient Reduction of Cr(VI) with Carbon Quantum Dots.

Hexavalent chromium (Cr(VI)) pollution is a global problem, and the reduction of highly toxic Cr(VI) to less toxic Cr(III) is considered to be an effective method to address Cr(VI) pollution. In this study, low-toxicity carbon quantum dots (CQDs) were used to reduce Cr(VI) in wastewater. The results show that CQDs can directly reduce Cr(VI) at pH 2 and can achieve a reduction efficiency of 94% within 120 min. It is observed that under pH higher than 2, CQDs can activate peroxymonosulfate (PMS) to produce reactive oxygen species (ROS) for the reduction of Cr(VI) and the reduction efficiency can reach 99% within 120 min even under neutral conditions. The investigation of the mechanism shows that the hydroxyl groups on the surface of CQDs can be directly oxidized by Cr(VI) because of the higher redox potential of Cr(VI) at pH 2. As the pH increases, the carbonyl groups on the surface of CQDs can activate PMS to generate ROS, O2•–, and 1O2, which result in Cr(VI) being reduced. To facilitate the practical application of CQDs, the treatment of Cr(VI) in real water samples by CQDs was simulated and the method reduced Cr(VI) from an initial concentration of 5 mg/L to only 8 μg/L in 150 min, which is below the California water quality standard of 10 μg/L. The study provides a new method for the removal of Cr(VI) from wastewater and a theoretical basis for practical application.

Treating Agricultural Runoff with a Mobile Carbon Filtration Unit

Several classes of pesticides have been shown to impair water quality in California, including organophosphates, pyrethroids and neonicotinoids. Vegetative treatment systems (VTS) can reduce pesticide loads and associated toxicity in agricultural runoff, but many water-soluble pesticides such as neonicotinoids are not effectively treated by VTS, and VTS installation is not always an option for growers required to remove non-crop vegetation for food safety concerns. Recent studies have shown that biochar filtration can be used to remove soluble contaminants, especially when coupled with other VTS components. We evaluated a mobile carbon filter system consisting of a trailer-mounted tank containing approximately 600L (~ 180 kg) of biochar. Input water from a 437-hectare agricultural drainage was pre-filtered and treated with biochar during two multi-week study periods. Laboratory toxicity tests and chemical and nutrient analyses were conducted on input and output water. Pesticide concentrations were initially reduced by greater than 99%. Treatment efficacy declined linearly and was expected to remain at least 50% effective for up to 34 weeks. Toxicity was assessed with Ceriodaphnia dubia, Hyalella azteca and Chironomus dilutus. Significant input toxicity was reduced to non-toxic levels in 6 of 16 samples. Some input concentrations of the neonicotinoid imidacloprid and the pyrethroid cypermethrin exceeded organism-specific toxicity thresholds and benchmarks, but the overall causes of toxicity were complex mixtures of agricultural chemicals. Nutrients were not reduced by the biochar. Results demonstrate the utility of biochar in treating agricultural runoff and provide measures of the longevity of biochar under field conditions.

Human health implications of extreme precipitation events and water quality in California, USA: a canonical correlation analysis

BackgroundPathogens and pollutants collect on the land surface or in infrastructure between strong rainfall episodes and are delivered via storm runoff to areas of human exposure, such as coastal recreational waters. In California, USA, precipitation events are projected to become more extreme and simultaneously decrease in frequency as storm tracks move poleward due to polar-amplified global warming. Precipitation extremes in California are dominated by atmospheric rivers, which carry more moisture in warmer climates. Thus, the physical driver of extreme precipitation events is expected to grow stronger with climate change, and pollutant accumulation and runoff-generated exposure to those pollutants are expected to increase, particularly after prolonged dry spells. Microbiological contamination of coastal waters during winter storms exposes human populations to elevated concentrations of microorganisms such as faecal bacteria, which could cause gastrointestinal and ear infections, and lead to exposure to pathogens causing life-threatening conditions, such as hepatitis A. The aim of this study was to quantitatively assess the effect of precipitation on coastal water quality in California. MethodsWe used a recently published catalogue of atmospheric rivers, in combination with historical daily precipitation data and levels of three indicators of faecal bacteria (total and faecal coliforms, and Escherichia coli) detected at roughly 500 monitoring locations in coastal waters along California's 840-mile coastline, to explore weekly associations between extreme precipitation events, particularly those related to atmospheric rivers, and the variability in water quality during 2003–09. We identified ten principal components (together explaining >90% of the variability) in precipitation and faecal bacteria time-series to reduce the dimensionality of the datasets. We then performed canonical correlation analysis of the principal components to identify and quantify associations between precipitation and faecal bacteria. FindingsWe found that the temporal evolutions of the leading canonical correlation patterns for precipitation and the faecal bacteria indicators were highly correlated. By looking at the Pearson correlation between precipitation and faecal bacteria canonical correlation patterns and the corresponding original time-series, we observed that the spatial pattern for all variables was strongest in southern California, and that it was mainly related to atmospheric river extreme precipitation events. Specific cases, such as peaks in precipitation and faecal pollution associated with an intense atmospheric river during January 7–12, 2005, were examined in detail. Overall, we found that precipitation over southern California explained over half of the weekly variability in faecal pollution in regional coastal waters. InterpretationThe associations between precipitation and water quality will be used to identify mechanisms of faecal pollutant delivery to coastal recreational waters via storm runoff and to track sources of pollution common in the region, such as sewage outfalls and homeless encampments near rivers and streams. This work should help to reduce vulnerability to extreme weather and delineate measures, such as an early warning system, that improve the response and resilience of human populations and ecosystems to a varying and changing climate. These results will also be used in the context of down-scaled climate change projections to estimate effects of future precipitation regime changes on coastal water quality and related health impacts. FundingUniversity of California Office of the President via Multicampus Research Programs and Initiatives.

Nitrogen assessment points: development and application to high‐elevation lakes in the Sierra Nevada, California

AbstractIncreased nitrogen (N) inputs to oligotrophic high‐elevation lakes are contributing to long‐term eutrophication, changes in nutrient cycles, and shifts in phytoplankton communities. State of California water quality standards do not adequately protect mountain lake ecosystems from atmospheric deposition, including lakes in national parks and other protected areas. The development of quantitative nutrient assessment points based on measurable ecological effects specific to high‐elevation lakes is an important step in long‐term protection of these ecosystems. We conducted in situ bioassay experiments during the phytoplankton growing season and modeled the response of phytoplankton to nutrient additions using algal growth models. Phytoplankton responded to nutrient additions (N and N + phosphorus) in five of seven experiments conducted in N‐limited lakes. Results were modeled using Michaelis‐Menten, Monod, and dose–response curves and used to calculate effective doses (ED) for excess phytoplankton growth. The 10%, 50%, and 90% ED estimates were determined for early and late hydrologic seasons. Results for the 10% ED were 0.33 (early hydrologic season) and 0.89 (late hydrologic season), the 50% ED were 1.0 μmol/L (early hydrologic season) and 4.0 μmol/L (late hydrologic season), and the 90% ED were 3.1 μmol/L (early hydrologic season) and 18 μmol/L (late hydrologic season). We then applied these assessment points to lake chemistry data from the National Park Service's Inventory and Monitoring Program to assess the status of lakes across Yosemite, Sequoia, and Kings Canyon national parks. Mean nitrate concentration during the growing season was 4.58 μmol/L and ranged from <0.04 to 11.8 μmol/L (n = 75). The 10% ED were exceeded by 28% (late hydrologic season) and 37% (early hydrologic season) of lakes, the 50% ED were exceeded by 18% (late hydrologic season) and 29% (early hydrologic season) of lakes, and the 90% ED were exceeded by 0.0% (late hydrologic season) and 21% (early hydrologic season) of lakes. Our results suggest that phytoplankton populations in many Sierra Nevada lakes are affected by N deposition based on exceedance of growth assessment points. The most sensitive lakes are typically found at higher elevations in watersheds with steep, north‐facing slopes and minimal vegetation.

Patterns and potential drivers of declining oxygen content along the southern California coast

Here we examine a 50+ yr data set from a regionally coordinated southern California water quality monitoring program to assess temporal trends and determine whether nearshore waters are exhibiting changes in dissolved oxygen (DO) content similar to those reported offshore. DO in sub—mixed layer nearshore waters (< 10 km from shore) have declined up to four times faster than reported for offshore waters over the last 15 yr. These trends were evident over depth, and along isopycnals. They have no precedent over the past 50 yr and do not appear to be attributable primarily to large‐scale climate variability in ocean DO. Coastal biophysical processes, including increased phytoplankton biomass in surface waters, are likely contributing to the recent elevated rate of DO decline in nearshore waters, as evidenced by higher rates of increase in apparent oxygen utilization. It is unclear whether these processes result from upwelling‐derived or anthropogenic nutrient inputs.

Socioeconomic factors and water quality in California

We investigate the relationships between water quality and socioeconomic factors in California at the county level for the years 1993–2006 using 24 water quality indicators coming from seven different types of water bodies. We estimate these relationships using three classes of models: the traditional per capita income-pollution level—Environmental Kuznets Curve (EKC)—specifications, a more inclusive model containing main socioeconomic variables such as agricultural intensity, land use, ethnic composition, population density and educational attainment, and a model that includes the socioeconomic variables while accounting for spatial correlations too. For most water quality indicators, we do not find support for EKC specifications. For pollutants like phosphorus and total suspended solids, the level of agricultural activity is a significant determinant of water quality in California, but for other surface water pollutants commonly considered agricultural pollutants, such as ammonia and nitrate, the level of agricultural activity is not statistically significant. We find that education, ethnic composition, age structure, land use, population density, and water area are all significantly correlated with various indicators of water quality.

Socioeconomic Factors and Water Quality in California

We investigate the relationships between water quality and socioeconomic factors in California at the county level for the years 1993 to 2006 using 24 water quality indicators coming from seven different types of water bodies. We estimate these relationships using three classes of models: the traditional per capita income-pollution level - Environmental Kuznets Curve (EKC) - specifications, a more inclusive model containing main socioeconomic variables such as agricultural intensity, land use, ethnic composition, population density and educational attainment, and a model that includes the socioeconomic variables while accounting for spatial correlations too. For most water quality indicators, we do not find support for EKC specifications. For pollutants like phosphorus and total suspended solids, the level of agricultural activity is a significant determinant of water quality in California, but for other surface water pollutants commonly considered agricultural pollutants, such as ammonia and nitrate, the level of agricultural activity is not statistically significant. We find that education, ethnic composition, age structure, land use, population density, and water area are all significantly correlated with various indicators of water quality.

Socioeconomic Factors and Water Quality in California

We investigate the relationships between water quality and socioeconomic factors in California at the county level for the years 1993 to 2006 using 24 water quality indicators coming from seven different types of water bodies. We estimate these relationships using three classes of models: the traditional per capita income-pollution level - Environmental Kuznets Curve (EKC) - specifications, a more inclusive model containing main socioeconomic variables such as agricultural intensity, land use, ethnic composition, population density and educational attainment, and a model that includes the socioeconomic variables while accounting for spatial correlations too. For most water quality indicators, we do not find support for EKC specifications. For pollutants like phosphorus and total suspended solids, the level of agricultural activity is a significant determinant of water quality in California, but for other surface water pollutants commonly considered agricultural pollutants, such as ammonia and nitrate, the level of agricultural activity is not statistically significant. We find that education, ethnic composition, age structure, land use, population density, and water area are all significantly correlated with various indicators of water quality.

Diagrams for evaluating groundwater quality

Some special diagrams for evaluating the chemical properties of groundwater intended for human consumption are introduced, in accordance with the Italian legislation assimilating EEC Directive 91/271 concerning the treatment of urban waste water and EEC Directive 91/676 related to the protection of waters against contamination by nitrates from agricultural sources. Although the limits imposed by the legislation reflect the Italian situation, they do however conform to directives concerning the whole European Community and offer a response to international guidelines promulgated by, for example, the World Health Organization so they are also suitable for use in national contexts other than Italy. Moreover, a diagram is introduced for monitoring the quality of water destined for agriculture, in particular for irrigation and the special issue of soil protection. In this case, in the absence of specific national and/or European laws and/or Directives, reference was made to the standards of the California Water Quality Control Board. The parameters and the relative limits have been taken from such standards, to identify three quality classes of water for agricultural use with evaluation of use. Plotting such diagrams is very simple: they can be used for improving presentation and interpretation of chemical data and, most of all, to provide comparisons between different water resources or to evaluate possible changes over time. Some examples, reported in the text, will aid understanding their use and interpretation.

Association of Nitrates, Trihalomethanes, and Arsenic in Drinking Water and Testicular Cancer Risk

SAA1-O-05 Introduction: Testicular cancer is the most common cancer among young men in the United States. The disease has few known risk factors. Chemical contaminants in drinking water are suspected disrupters of the endocrine system and may affect estrogen levels. In this study, we used a large number of cases from the California Cancer Registry to determine if an association exists between chemical contaminants in drinking water and risk of testicular cancer. Methods: We used the California Water Quality Monitoring Database to obtain levels of sampled water contaminants. We frequency-matched controls to cases based on the SES characteristics of the cases’ 1990 or 2000 U.S. Census block groups. Of 7119 geo-coded cases, we frequency-matched 7095 controls. We defined the exposure period as 2 years before the diagnosis date of the cases. We developed a probability weighting methodology that took into account several data sources to link a subject to a water system. Water systems were surveyed to characterize relative concentration contributions of upstream sampling stations for each study subject during the corresponding exposure period. Logistic regression was used for multivariate analysis. Results: A strong dose-response pattern was found for nitrates in surface water only, with significant odds ratios in every category. The AOR for nitrates in surface water above the 95th percentile (compared with the first quartile) was 1.99 (95% CI, 1.28–3.10). For arsenic, significant adjusted odds ratios emerged in surface water above the 3rd quartile and above the 95th percentile, compared with the first quartile (AOR = 1.39; 95% CI, 1.03–1.88; and AOR = 2.52; 95% CI, 1.54–4.10, respectively). In surface water, we found an apparent interaction between nitrates and trihalomethanes (THMs), with a strong dose-response when THMs were present. For surface water sources only with THMs present, AORs for nitrate ranged from 1.76 for the second quintile (95% CI = 1.25–2.48) to 2.91 for the fifth quintile (95% CI = 1.69–5.01), compared with the first quartile. Discussion: The risk of testicular cancer associated with nitrate levels increased in a dose-response fashion in households receiving surface sources of water only. We found that in the presence of THMs in surface water, the dose-response pattern became even steeper, with the risk associated with the highest level of nitrate concentrations increasing to threefold. These data are suggestive of an effect of nitrates and THMs on testicular cancer risk, but the lack of an appropriate lag time, and more exact consumption data, limits the interpretability of the findings.

(353) University of California Water Quality Program for Floriculture and Nursery Growers in Ventura and Los Angeles Counties

Nursery growers must implement “Best Management Practices” (BMPs) to mitigate runoff and leaching of pollutants. However, there is little data on the effectiveness of BMPs, and further research is needed. Growers require training to minimize runoff and capital to comply with evolving regulations. We collected irrigation efficiency data at 26 nursery sites using catch cans. Uniformity value was calculated as: DU = 100 × (Average of the “Low Quarter”/Average of All Measurements). Results showed that one-third of the nurseries had irrigation uniformities below 67%, and only one nursery exceeded 82%. Continuous flow monitoring at 10-minute intervals was conducted at three nursery sites to determine daily and monthly variation. One hundred samples of the runoff were taken four times at each nursery and analyzed for nitrate, chloride, and sulfate. We developed a 29-page checklist of BMPs to reduce runoff, which we used to survey 53 growers in Ventura County. After completing the survey, 20 growers applied for funds from a 1.2 million dollar cost-share program that we initiated. This program provides funds to implement improved technologies to reduce runoff and/or conserve water; funds were awarded to 18 nurseries. We are collecting monitoring data from each cooperating nursery implementing improvements. These data measure the number of BMPs used by growers, provide a current “snapshot” of the industry, and document the effectiveness of future BMP implementation. We offer on-farm consulting, and conducted eight water quality/irrigation educational programs—four in Spanish. We elevated grower awareness concerning regulations and options for reducing runoff, and the data will be useful in evaluating future improvements.

New genotypes and factors associated with Cryptosporidium detection in mussels ( Mytilus spp.) along the California coast

A 3 year study was conducted to evaluate mussels as bioindicators of faecal contamination in coastal ecosystems of California. Haemolymph samples from 4680 mussels ( Mytilus spp.) were tested for Cryptosporidium genotypes using PCR amplification and DNA sequence analysis. Our hypotheses were that mussels collected from sites near livestock runoff or human sewage outflow would be more likely to contain the faecal pathogen Cryptosporidium than mussels collected distant to these sites, and that the prevalence would be greatest during the wet season when runoff into the nearshore marine environment was highest. To test these hypotheses, 156 batches of sentinel mussels were collected quarterly at nearshore marine sites considered at higher risk for exposure to livestock runoff, higher risk for exposure to human sewage, or lower risk for exposure to both faecal sources. Cryptosporidium genotypes detected in Haemolymph samples from individual mussels included Cryptosporidium parvum, Cryptosporidium felis, Cryptosporidium andersoni, and two novel Cryptosporidium spp. Factors significantly associated with detection of Cryptosporidium spp. in mussel batches were exposure to freshwater outflow and mussel collection within a week following a precipitation event. Detection of Cryptosporidium spp. was not associated with higher or lower risk status for exposure to livestock faeces or human sewage sources. This study showed that mussels can be used to monitor water quality in California and suggests that humans and animals ingesting faecal-contaminated water and shellfish may be exposed to both host-specific and anthropozoonotic Cryptosporidium genotypes of public health significance.

Metal attenuation at the abandoned Spenceville copper mine

Acid mine drainage from the abandoned Spenceville Cu mine contributes toxic levels of metals to nearby streams. The majority of metal contamination, however, is removed from the system over a short spatial distance. The primary mechanism responsible for this geochemical scavenging is adsorption onto Fe, Mn and Al-hydroxide precipitates that are clearly visible in the streambed. A 5-step sequential extraction procedure was performed on the streambed sediments in order to characterize solid phase speciation of metals in this area. Two fractions (carbonate and amorphous Fe-Mn hydroxide) were considered to be bioavailable, with the possibility of the metals in these phases being remobilized under changing ambient conditions. Copper and Zn were found to have the highest bioavailability, followed by Cd, Pb and Al. Although most metal in the acid mine drainage was rapidly removed from the system, aqueous concentrations of Cu and Zn exceeded California water quality criteria 0.5 km downstream of the mine site.

MTBE in California's Drinking Water: A Comparison of Groundwater Versus Surface Water Sources

During the last decade, the fuel oxygenate methyl tertiary butyl ether (MTBE) has received widespread attention as a potential threat to water quality, primarily due to leaking underground gasoline storage tanks and watercraft with two-stroke engines. In this article, we examine the annual detection frequency, number of new source detections, and concentration of MTBE detected in California's public drinking water groundwater and surface water sources from 1995 to 2002. This work builds on our previous evaluations of California's water quality monitoring database. However, it is unique in that it includes separate evaluations for groundwater and surface water sources that are of greatest concern to regulators, and which are likely being used for current public consumption. Our evaluations also include full-year data for 2002 (which have not been published previously) and an analysis of how the sampling and reported detections of MTBE vary by geographic location. We find that MTBE was generally detected (at any level) in approximately 0.5-0.9% and 0.2-0.4% of all groundwater sources assuming a one-detection and two-detection criterion, respectively. The overall detection frequency for MTBE in surface water sources is significantly higher than for groundwater sources, although these surface water detections appear to have substantially declined since 1996 (e.g., 7-9% for all surface water sources during 1996 to 1999 and 4% for all surface water sources during 2000 to 2002, assuming a one-detection criterion). The detection frequency of MTBE concentrations at or above the state drinking water standards in all drinking water sources (both groundwater and surface water sources) and the subset of drinking water sources that are likely to currently be delivered to consumers is markedly lower (and often zero). Despite the significant increase in water sampling over time, the number of new drinking water sources found to contain MTBE in California has not increased at the same rate and appears to have remained relatively stable or to have decreased since 1998. The data also show that nearly all of the 58 counties in California have routinely sampled at least some of their groundwater and surface water sources for MTBE over the last 8 years. Geographical evaluations show that MTBE has been detected (at least once) in groundwater sources in 34 counties and in surface water sources in 18 counties but has only been detected routinely (i.e., for 3 or more years) in 16 and 7 counties, respectively. Detected concentrations of MTBE are also generally below state drinking water standards, particularly for surface water sources. In short: (1) MTBE is rarely found in California groundwater or surface water sources that are of greatest concern to regulators or the public, and (2) drinking water detections of MTBE are expected to decline in the future due to the pending phase-out of MTBE and recent regulatory programs aimed at controlling gasoline releases from underground storage tanks and two-stroke-engine watercraft.

Deregulation, Chemical Waste, and Ground Water: A 1949 Debate

When did scientists, regulators, and industry first realise that industrial waste threatens the quality of ground-water? The conventional wisdom holds that widespread knowledge of this problem only dates back to the 1970s. In this view, what knowledge did exist earlier was confined to small circles of technical specialists and not generally known in industry. The passage of the Dickey Act, which established California's Water Quality Control Boards in 1949, provides documentation of events and ideas against which this hypothesis can be tested directly. A legislative committee, after lengthy public hearings, produced a detailed scientific report about ground-water and stream pollution. Bitter public controversy then erupted among politicians, industry, regulators and scientists about whether and how to protect ground-water supplies from contamination by industrial wastes. This controversy in the most populous region of the western United States demonstrates that many well-informed policy-makers knew that industrial wastes could pollute ground-water.

Impact of energy cost and water resource availability on agriculture and ground water quality in California

This paper discusses the simultaneous effects of water scarcity and energy price increases on farm-level decisions such as water-related technology substitution and cropping patterns, and on groundwater quantity and quality, and farm income. A dynamic model is used. The model maximizes lomg-run private profits by incorporating dynamic relationships of water level and quality in aquifer, with management and technology change. The model is applied on a regional scale to conditions in Kern Country, California. This region has been hit by a six-year drought, and faces a proposed increase in electricity rates for agricultural-class customers. The simulation results suggest that over time, given the changes in water availability and energy prices, the proportion of groundwater in irrigation water will increase, cropping patterns will consist of higher shares of salt-tolerant crops, more energy and water saving pumping and irrigation equipment will replace existing technologies, agricultural net income will drop significantly, the level of groundwater in the aquifer will drop, and its quality will degrade. In extreme scenarios, agriculture may cease in locations depending heavily on groundwater.