Revealing the sources of arsenic in private well water using Random Forest Classification and Regression.

Exposure to arsenic through private drinking water wells causes serious human health risks throughout the globe. Water testing data indicates there is arsenic contamination in private drinking water wells across New Jersey. To reduce the adverse health risk due to exposure to arsenic in drinking water, it is necessary to identify arsenic sources affecting private wells. Private wells are not regulated by any federal or state agencies through the Safe Drinking Water Act and therefore information is often lacking. To this end, we have developed machine learning algorithms including Random Forest Classification and Regression to decipher the factors contributing to higher arsenic concentration in private drinking water wells in west-central New Jersey. Arsenic concentration in private drinking water wells served as a response variable while explanatory variables were geological bedrock type, soil type, drainage class, land use/cover, and presence of orchards, contaminated sites, and abandoned mines within the 152.4-meter (500 ft) radius of each well. Random Forest Classification and Regression achieved 66 % and 55 % prediction accuracies for arsenic concentration in private drinking water wells, respectively. Overall, both models identify that bedrock, soil, land use/cover, and drainage type (in descending order) are the most important variables contributing to higher arsenic concentration in well water. These models further identify bedrock subgroups at a finer scale including Passaic Formation, Lockatong Formation, Stockton Formation contributing significantly to arsenic concentration in well water. Identification of sources of arsenic contamination in private drinking water wells at such a fine scale facilitates development of more targeted outreach as well as mitigation strategies to improve water quality and safeguard human health.

Evaluating the impact of land uses on stream integrity using machine learning algorithms.

A general pattern of declining aquatic ecological integrity with increasing urban land use has been well established for a number of watersheds worldwide. A more nuanced characterization of the influence of different urban land uses and the determination of cumulative thresholds will further inform watershed planning and management. To this end, we investigated the utility of two machine learning algorithms (Random Forests (RF) and Boosted Regression Trees (BRT)) to model stream impairment through multimetric macroinvertebrate index known as High Gradient Macroinvertebrate Index (HGMI) in an urbanizing watershed located in north-central New Jersey, United States. These machine learning algorithms were able to explain at least 50% of the variability of stream integrity based on watershed land use/land cover. While comparable in results, RF was found to be easier to train and was somewhat more robust to model overfitting compared to BRT. Our results document the influence of increasing high-medium density (> 30% Impervious Surface cover (ISC)), low density (15-30% ISC) urban and transitional/barren land had in negatively affecting stream biological integrity. The thresholds generated by partial plots suggest that the stream integrity decreased abruptly when the percentage of high-medium and low density urban, and transitional/barren land went above 10%, 8%, and 2% of the watershed, respectively. Additionally, when rural residential surpassed 30% threshold, it behaved similar to low density urban towards stream integrity. Identification of such cumulative thresholds can help watershed managers and policymakers to craft land use zoning regulations and design restoration programs that are grounded by objective scientific criteria.

A comparison of three empirically based, spatially explicit predictive models of residential soil Pb concentrations in Baltimore, Maryland, USA: understanding the variability within cities.

In many older US cities, lead (Pb) contamination of residential soil is widespread; however, contamination is not uniform. Empirically based, spatially explicit models can assist city agencies in addressing this important public health concern by identifying areas predicted to exceed public health targets for soil Pb contamination. Sampling of 61 residential properties in Baltimore City using field portable X-ray fluorescence revealed that 53 % had soil Pb that exceeded the USEPA reportable limit of 400 ppm. These data were used as the input to three different spatially explicit models: a traditional general linear model (GLM), and two machine learning techniques: classification and regression trees (CART) and Random Forests (RF). The GLM revealed that housing age, distance to road, distance to building, and the interactions between variables explained 38 % of the variation in the data. The CART model confirmed the importance of these variables, with housing age, distance to building, and distance to major road networks determining the terminal nodes of the CART model. Using the same three predictor variables, the RF model explained 42 % of the variation in the data. The overall accuracy, which is a measure of agreement between the model and an independent dataset, was 90 % for the GLM, 83 % for the CART model, and 72 % for the RF model. A range of spatially explicit models that can be adapted to changing soil Pb guidelines allows managers to select the most appropriate model based on public health targets.

The effects of the urban built environment on the spatial distribution of lead in residential soils.

Lead contamination of urban residential soils is a public health concern. Consequently, there is a need to delineate hotspots in the landscape to identify risk and facilitate remediation. Land use is a good predictor of some environmental pollutants. However, in the case of soil lead, research has shown that land use is not a useful proxy. We hypothesize that soil lead is related to both individual landscape features at the parcel scale and the landscape context in which parcels are embedded. We sampled soil lead on 61 residential parcels in Baltimore, Maryland using field-portable x-ray fluorescence. Thirty percent of parcels had average lead concentrations that exceeded the USEPA limit of 400 ppm and 53% had at least one reading that exceeded 400 ppm. Results indicate that soil lead is strongly associated with housing age, distance to roadways, and on a parcel scale, distance to built structures.

Comparison of the A-Cc curve fitting methods in determining maximum ribulose 1.5-bisphosphate carboxylase/oxygenase carboxylation rate, potential light saturated electron transport rate and leaf dark respiration.

A review of the literature revealed that a variety of methods are currently used for fitting net assimilation of CO2-chloroplastic CO2 concentration (A-Cc) curves, resulting in considerable differences in estimating the A-Cc parameters [including maximum ribulose 1.5-bisphosphate carboxylase/oxygenase (Rubisco) carboxylation rate (Vcmax), potential light saturated electron transport rate (Jmax), leaf dark respiration in the light (Rd), mesophyll conductance (gm) and triose-phosphate utilization (TPU)]. In this paper, we examined the impacts of fitting methods on the estimations of Vcmax, Jmax, TPU, Rd and gm using grid search and non-linear fitting techniques. Our results suggested that the fitting methods significantly affected the predictions of Rubisco-limited (Ac), ribulose 1,5-bisphosphate-limited (Aj) and TPU-limited (Ap) curves and leaf photosynthesis velocities because of the inconsistent estimate of Vcmax, Jmax, TPU, Rd and gm, but they barely influenced the Jmax : Vcmax, Vcmax : Rd and Jmax : TPU ratio. In terms of fitting accuracy, simplicity of fitting procedures and sample size requirement, we recommend to combine grid search and non-linear techniques to directly and simultaneously fit Vcmax, Jmax, TPU, Rd and gm with the whole A-Cc curve in contrast to the conventional method, which fits Vcmax, Rd or gm first and then solves for Vcmax, Jmax and/or TPU with V(cmax), Rd and/or gm held as constants.

Statewide mapping and assessment of vernal pools: a New Jersey case study.

In 2001, the New Jersey Department of Environmental Protection (NJDEP) adopted rules specifically protecting vernal pool habitat for the first time. Vernal pools are small isolated temporary bodies of water that provide critical breeding habitat for a number of amphibian species. To implement these rules and ultimately afford vernal pools protection, the NJDEP first needed to assemble a statewide database of vernal pool locations. In response, the Rutgers University Center for Remote Sensing and Spatial Analysis (CRSSA) was funded to develop a cost effective technique to map vernal pool locations statewide. The objective of CRSSA's mapping effort was to develop a complete potential vernal pool database to be able to identify individual isolated vernal pools as well as areas of high local density, or 'hotspots'. CRSSA used visual interpretation of leaf-off color infrared digital orthophotography in a computerized GIS environment to identify and map over 13,000 potential vernal pools. Using the 1m scale imagery, we determined the minimum detectable pool size to be on the order of 0.02 ha in size. Subsequent field checking has revealed a 12% error of commission that was due to our inclination towards erring on the side of inclusion in mapping many water features as potential vernal pools. For a vernal pool to receive regulated protection, it must be 'certified' that it serves as habitat for obligate or facultative vernal pool amphibian species. To aid in these efforts, CRSSA developed an interactive internet mapping site to assist NJDEP and its citizen volunteer corps in locating and navigating to their survey areas and to facilitate the on-line submittal of survey observations.

Modeling the ecological consequences of land-use policies in an urbanizing region.

Insight into future land use and effective ways to control land-use change is crucial to addressing environmental change. A variety of growth-control policies have been adopted by municipal and regional governments within the United States to try to minimize the ecological impact of continued urbanization, but it is often unclear if those policies can meet the stated ecological goals. Land-use-change models provide a way to generate predictions of future change, while exploring the impact of different land-use policies before irreversible transformations occur. In this article, an approach to modeling land-use policies that focuses on their ecological consequences is described. The policy simulation approach was used to predict future land use in the Barnegat Bay and Mullica River watersheds, in southeastern New Jersey, USA. Four commonly used policies were considered: down-zoning, cluster development, wetlands/water buffers, and open space protection. The results of the analysis suggest that none of the policies modeled were able to alter future land-use patterns, raising questions about the effectiveness of commonly adopted land-use policies. However, the policy modeling approach used in this study proved to be a useful way to determine if adoption of a given policy could improve the likelihood of meeting ecological goals.

Landscape changes in the Mullica River basin of the Pinelands National Reserve, New Jersey, USA.

In 1979, the Pinelands Commission was established as a regional land-use planning and regulatory agency charged with the implementation of a Comprehensive Management Plan (CMP) for the Pinelands National Reserve (New Jersey, USA). The CMP was created to balance land preservation and development interests in the Reserve. Because water-quality degradation from developed and agricultural landscapes is associated with changes in the composition of biological communities, monitoring landscape changes provides one of the most direct measures of the impact of land-use policies on the Pinelands ecosystem. In this study, we prepared detailed, land-cover maps within randomly selected aerial-photograph plots for a major watershed in the Reserve. We used these land-cover maps to quantify changes in landscape composition and structure (i.e., patch size, patch area, and number of patches) and characterize land-cover transitions in the basin between 1979 and 1991. Because the study period represented the first 12 years of the regional-planning effort in the Reserve, we evaluated the relationship between land-cover transitions and Commission management-area designations which permit different land-use intensities. Although the landscape composition was similar in 1979 and 1991, we found an increase in the total area and number of developed-land, managed-grassland, and barren-land patches. An increase in the number of patches and a decrease in the total area and median and third-quartile patch sizes for forest land and for all patches regardless of cover type indicated that fragmentation of forest land and the landscape as a whole occurred during the study period. The major land-cover transitions that occurred during the period were the loss of forest land to development and associated cover types and the conversion of one agricultural type to another. Overall, land-cover transitions during the period were found to be consistent with the Commission management-area designations, which indicated that the regional-planning effort has been successful in directing human activities to appropriate areas of the basin.

A Spatial Model of Atmospheric Deposition for the Northeastern U.S.

Spatial patterns of atmospheric deposition across the northeastern United States were evaluated and summarized in a simple model as a function of elevation and geographic position within the region. For wet deposition, 3-11 yr of annual concentration data for the major ions in precipitation were obtained from the National Atmospheric Deposition Program/National Trend Network (NADP/NTN) for 26 sites within the region. Concentration trends were evaluated by regression of annual mean concentrations against latitude and longitude. For nitrate, sulfate, and ammonium concentrations, a more than twofold linear decrease occurs from western New York and Pennsylvania to eastern Maine. These trends were combined with regional and elevational trends of precipitation amount, obtained from 30-yr records of annual precipitation at >300 weather stations, to provide long-term patterns of wet deposition. Regional trends of dry deposition of N and S compounds were determined using 2-3 yr of particle and gas concentration data collected by the National Dry Deposition Network (NDDN) and several other sources, in combination with estimates of deposition velocities. Contrary to wet deposition trends, the dominant air concentration trends were steep decreases from south to north, creating regional decreases in total deposition (wet + dry) from the southwest to the northeast. This contrast between wet and dry deposition trends suggests that within the northeast the two deposition forms are received in different proportions from different source areas, wet deposited materials primarily from areas to the west and dry deposited materials primarily from urban areas along the southern edge of the region. The equations generated describing spatial patterns of wet and dry deposition within the region were entered into a geographic information system (GIS) containing a digital elevation model (DEM) in order to develop spatially explicit predictions of atmospheric deposition for the region.