An Investigation on Machine Learning Models for the Prediction of Cyanobacteria Growth
Journal of Fundamental and Applied LimnologyGiere, Johannes; Riley, Derek; Nowling, R. J.; McComack, Joshua; Sander, Hedda
2020
Harmful algal blooms, which are a danger to the lives of humans and animals, are caused by a sudden increase in the concentration of cyanobacteria in freshwater lakes. Cyanobacteria concentrations can be reliably measured using chemical and biological indicators, but the measurement process of the indicators is either labor-intensive or very costly. These limitations do not allow the general public to measure concentrations, so local health organizations or departments regularly assume the responsibility of measuring water quality. While computational models exist to predict algal concentrations, the accuracy of these models and need for customization due to varied lake conditions make them generally not yet reliable. We find that common regression-error functions cannot sufficiently evaluate the performance of cyanobacteria prediction models because the occurrence of harmful algal blooms is rare. Therefore, we present a method of forecasting cyanobacteria concentrations in freshwater lakes based on a machine-learning model trained on a dataset from Lake Utah with automatically-measured indicators from lake buoys. We compare several models and find that a support vector machine with a radial basis function kernel for regression reliably forecasts harmful algal blooms using comparatively few and easy-to-obtain input parameters. The special feature of the model is that it exclusively uses variables that can be measured by the general public without great effort and costs, and the amount of data necessary to train such a model is relatively minimal, allowing different models to be trained to accommodate for the nuances of different lakes.
View more
Diurnal vertical migration of cyanobacteria and chlorophyta in eutrophied shallow freshwater lakes
Fundamental and Applied Limnology / Archiv für Hydrobiologie,von Orgies-Rutenberg, M., Rolfes, C., Eckel, T., Quiroz, A., Skalbeck, J., Riley, D., Sander, H.
2017
Circadian rhythms are thought of as means for adaptation helping survival fitness of a species. For algal species associated with harmful algal blooms (HAB) in eutrophied freshwater lakes usually light and nutrient availability, especially phosphate, seem to drive patterns of the vertical migration within the water column. The vertical migration patterns of species associated with HAB in freshwater lakes (Cyanobacteria) should be taken as input parameters for modelling algae. As HAB present a health risk to the public they should be monitored and predicted via simulation models, and the results of the predictions should be shared with the public using familiar tools such as smartphone apps or websites. To gather the data on which the model will be formulated, two shallow freshwater lakes (eutrophic condition: Lake Stadtgraben, Northern Germany, oligotrophic condition: Lake Russo, Wisconsin, USA in temperate climates were selected to serve as models for investigating the vertical migration in different seasonal times under natural conditions. Phosphate concentrations, as well as light and temperature over time in hourly increments at the lake surface and bottom were monitored. In addition the vertical migration pattern of Cyanobacteria and Chlorophyta populations was followed over 24 hrs in spring (May) and fall (August) in order to derive a behavior assumption as input for a model predicting HAB. In Lake Stadtgraben the vertical migration pattern was strongly influenced by light rather than by phosphate availability in spring, as phosphate was readily available at that time in all depths, while temperature was significantly different between the top and -bottom. The vertical migration pattern was dampened in fall season in both, the oligotrophic and the eutrophic lake, while temperature was not significantly different from the top to the bottom. Thus, vertical migration patterns observed may change slightly with season, which will impact on the outcome of simulation models dependent on the time of day and lake depth, at which input parameters such as Chlorophyll-a are measured.
View more
Using Data Mining in Combination with Machine Learning to Enhance Crowdsourcing of a Formal Model of Biodiesel Production
Midwest Instructional Computing SymposiumFischer, M., Riley, D.
2016
Formal modeling, simulation, and analysis of complex systems is valuable because it can provide insights into complex systems that are too expensive or difficult to analyze otherwise. In this work, we present an approach for improving simulation trajectory choices in a Monte Carlo framework using a combination of crowdsourcing, machine learning, and data mining. We apply machine learning to analysis of a formal model of biodiesel production as a method of improving the efficiency of the crowd sourced mobile simulation analysis of the model. Data is collected and data mined in a central server where machine learning is applied and recommendations from the machine learning algorithm are fed back to crowd workers via suggestions on the mobile app. Ultimately, we show that this approach can improve efficiency of optimal safe state identification in the biodiesel model analysis.
View more
Development of a Mobile Phone Application for the Prediction of Harmful Algal Blooms in Inland Lakes
Fundamental and Applied Limnology / Archiv für HydrobiologieGotthold, J.P., Deshmukh, A., Nighojkar, V., Skalbeck, J., Riley, D., Sander, H.
2016
Harmful algal blooms mainly caused by cyanobacteria in freshwater ecosystems often present a health risk to the public within eutrophied shallow lakes due to algal toxins released into the water during the final stage of an algal bloom. Thus, algal growth should be carefully monitored during the summer season, especially in fre- quented recreational areas. Traditionally, water samples must be sent to a lab to analyze the data to predict algal blooms, costing time and money. Models on a smartphone predicting harmful algal blooms from easily measurable parameters could help individuals to take precautionary measures in order to prevent health risks from drinking and bathing in water and help to raise public awareness. In this work we present a mobile smartphone application that generates a prediction of the likelihood of an algal bloom from a variety of easily-measured input parameters that could be obtained by an informed smartphone user with simple instruments. Our model was implemented in an Android mobile phone application using App Inventor. The model we use is based on the Verhulst equation and allows users to enter any of the following measurements to predict and algal bloom: surface temperature, inverse Secchi depth, dissolved oxygen (DO) at the surface, and chlorophyll fluorescence (Chl-a).
View more
Mobile Technologies in Healthcare: Approaches and Architecture
AIMS InternationalMukherjee, M., Chalasani, S., Riley, D.
Accepted 2016
Crowdsourcing Automobile Parking Availability Sensing Using Mobile Phones
UWM Undergraduate Research SymposiumVillalobos, J., Kifle, B., Riley, D., Torrero, J.U.Q.
2015
A lack of reliable knowledge about automobile parking availability in areas such as schools, work, or major cities wastes time, energy, and fuel as people try to find available parking spaces. Real-time parking monitoring phone applications exist, but keeping accurate, reliable parking availability information proves to be a difficult task due to the unreliability of real time information, especially in less densely populated areas. In this paper, we present a parking monitoring system that uses crowdsourcing in combination with mobile phone sensors to provide accurate, reliable real-time parking availability information. We present a study of the use of the application on a university campus to demonstrate its effectiveness.
View more
