Revealing Biotic and Abiotic Controls of Harmful Algal Blooms in a Shallow Subtropical Lake through Statistical Machine Learning

Gilmore

et al. 2020

Preprint

Abstract. Time-lapse imagery of streams and rivers provide new qualitative insights into hydrologic conditions at stream gauges, especially when site visits are biased toward baseflow or fair-weather conditions. Imagery from fixed, ground-based cameras is also rich in quantitative information that can improve streamflow monitoring. For instance, time-lapse imagery may be valuable for filling data gaps when sensors fail and/or during lapses in funding for monitoring programs. In this study, we automated the analysis of time-lapse imagery from a single camera at a single location, then built and tested machine learning models using programmatically calculable scalar image features to fill data gaps in stream gauge records. Time-lapse images were taken with a fixed, ground-based camera that is part of a documentary watershed imaging project (https://plattebasintimelapse.com/). Features were extracted from 40,000+ daylight images taken at one-hour intervals from 2012 to 2019. The algorithms removed dawn and dusk images that were too dark for feature extraction. The image features were merged with United States Geological Survey (USGS) stage and discharge data (i.e., response variables) from the site based on image capture times and USGS timestamps. We then developed a workflow to identify a suitable feature set to build machine learning models with a randomly selected training set of 30 % of the images with the remaining 70 % for a test set. Predictions were generated from Multi-layer Perceptron (MLP), Random Forest Regression (RFR), and Support Vector Regression (SVR) models. A Kalman filter was applied to the predictions to remove noise. Error metrics were calculated, including Nash-Sutcliffe Efficiency (NSE), Prediction Bias (PBIAS), RMSE-Standard Deviation Ratio (RSR), and an alternative metric that accounted for seasonal runoff. After suitable features were identified, the dataset was divided into test sets of simulated data gaps for 2015, 2016, and 2017. The training sets for each gap were features from contiguous images and sensor readings before and after the gaps. NSE for the year-long gap predictions ranged from 0.63 to 0.90 for discharge and 0.47 to 0.90 for stage. The predictions for 2015 and 2017 displayed lower prediction errors than for 2016. The 2016 discharge was significantly higher than training data, which could explain the poorer performance. First and second half-year test sets were created for 2016 along with MLP models from before/after training sets for each of the gaps that held discharge measurements similar to those in the gaps. The half-year gap models' predictions improved NSE, PBIAS and RSR. The results show it is possible to extract features from images taken with the downstream facing camera to build machine learning models that produce accurate stage and discharge predictions. The methods employed should be transferrable to other sites with ground-based cameras.

Section: Discussionmentioning

confidence: 99%

Camera-based Water Stage and Discharge Prediction with Machine Learning

Gilmore

et al. 2020

Preprint

“…Each fold was used as training data four times, and testing data once. This process was repeated five times to create a total of 25 models, similar to the approach used by Nelson et al (2018). The four folds designated to build the model underwent a nested five-fold cross validation, as specified in the trainControl function within the caret (Classification and Regression Training) R package (Kuhn, 2008;R Core Team, 2017).…”

Section: Random Forest Applicationmentioning

confidence: 99%

Determination of vadose and saturated-zone nitrate lag times using long-term groundwater monitoring data and statistical machine learning

Wells¹,

Gilmore²,

Nelson³

et al. 2020

Preprint

Abstract. In this study, we explored the use of statistical machine learning and long-term groundwater nitrate monitoring data to estimate vadose-zone and saturated-zone lag times in an irrigated alluvial agricultural setting. Unlike most previous statistical machine learning studies that sought to predict groundwater nitrate concentrations within aquifers, the focus of this study was to leverage available groundwater nitrate concentrations and other environmental variable data to determine mean vertical velocities (transport rates) of water and solutes in the vadose zone and saturated zone (3.50 m/year and 3.75 m/year, respectively). Although a saturated-zone velocity that is greater than a vadose-zone velocity would be counterintuitive in most aquifer settings, the statistical machine learning results are consistent with two contrasting primary recharge processes in this aquifer: (1) diffuse recharge from irrigation and precipitation across the landscape, and (2) focused recharge from leaking irrigation conveyance canals. The vadose-zone mean velocity yielded a mean recharge rate (0.46 m/year) consistent with previous estimates from groundwater age-dating in shallow wells (0.38 m/year). The saturated zone mean velocity yielded a recharge rate (1.31 m/year) that was more consistent with focused recharge from leaky irrigation canals, as indicated by previous results of groundwater age-dating in intermediate-depth wells (1.22 m/year). Collectively, the statistical machine-learning model results are consistent with previous observations of relatively high-water fluxes and short transit times for water and nitrate in the aquifer. Partial dependence plots from the model indicate a sharp threshold where high groundwater nitrate concentrations are mostly associated with total travel times of seven years or less, possibly reflecting some combination of recent management practices and a tendency for nitrate concentrations to be higher in diffuse infiltration recharge than in canal leakage water. Limitations to the machine learning approach include potential non-uniqueness when comparing model performance for different transport rate combinations and highlight the need to corroborate statistical model results with a robust conceptual model and complementary information such as groundwater age.

Integr Envir Assess & Manag

“…Similarly, artificial neural networks have been used to select biomarkers on the basis of key response variables (Bradley 2012). Environmental metadata using a Random Forests machine learning algorithm have likewise been used to reveal nonlinear relationships and critical thresholds for cyanobacterial blooms (Nelson et al 2018), which is significant because HABs now represent the greatest water quality threat in some ecosystems (Brooks et al 2017). Environmental metadata using a Random Forests machine learning algorithm have likewise been used to reveal nonlinear relationships and critical thresholds for cyanobacterial blooms (Nelson et al 2018), which is significant because HABs now represent the greatest water quality threat in some ecosystems (Brooks et al 2017).…”

Section: Tools For Improving Risk Assessmentmentioning

confidence: 99%

“…Decision tree models based on environmental metadata have been used to predict benthic macroinvertebrate distributions (D'Heygere et al 2003). Environmental metadata using a Random Forests machine learning algorithm have likewise been used to reveal nonlinear relationships and critical thresholds for cyanobacterial blooms (Nelson et al 2018), which is significant because HABs now represent the greatest water quality threat in some ecosystems (Brooks et al 2017).…”

Section: Tools For Improving Risk Assessmentmentioning

confidence: 99%

Towards Sustainable Environmental Quality: Priority Research Questions for the Australasian Region of Oceania

Gaw

Harford

Pettigrove

et al. 2019

Environmental challenges persist across the world, including the Australasian region of Oceania, where biodiversity hotspots and unique ecosystems such as the Great Barrier Reef are common. These systems are routinely affected by multiple stressors from anthropogenic activities, and increasingly influenced by global megatrends (e.g., the food–energy–water nexus, demographic transitions to cities) and climate change. Here we report priority research questions from the Global Horizon Scanning Project, which aimed to identify, prioritize, and advance environmental quality research needs from an Australasian perspective, within a global context. We employed a transparent and inclusive process of soliciting key questions from Australasian members of the Society of Environmental Toxicology and Chemistry. Following submission of 78 questions, 20 priority research questions were identified during an expert workshop in Nelson, New Zealand. These research questions covered a range of issues of global relevance, including research needed to more closely integrate ecotoxicology and ecology for the protection of ecosystems, increase flexibility for prioritizing chemical substances currently in commerce, understand the impacts of complex mixtures and multiple stressors, and define environmental quality and ecosystem integrity of temporary waters. Some questions have specific relevance to Australasia, particularly the uncertainties associated with using toxicity data from exotic species to protect unique indigenous species. Several related priority questions deal with the theme of how widely international ecotoxicological data and databases can be applied to regional ecosystems. Other timely questions, which focus on improving predictive chemistry and toxicology tools and techniques, will be important to answer several of the priority questions identified here. Another important question raised was how to protect local cultural and social values and maintain indigenous engagement during problem formulation and identification of ecosystem protection goals. Addressing these questions will be challenging, but doing so promises to advance environmental sustainability in Oceania and globally.