SBIR-STTR Award

The objective of this study is to develop methods to adequately characterize the chemical and physical environments of lakes and reservoirs for the CONUS operational implementation within SHARC. Over the course of this project we will explore the novel and innovative integration of deep machine learning, data analytics, computational fluid mechanics and in-situ/overhead empirical data. Specifically, we will utilize self-organizing map techniques to classify physiochemical and geomorphological conditions within lakes and reservoirs, leverage the classification of the geomorphological conditions to develop characteristic wind-driven flow patterns within lakes and reservoirs, incorporate the derived information into SHARCs existing database frameworks and leverage all the above information to develop the relevant questions and workflow to allow users of SHARC to conduct hazard assessment modeling within lakes and reservoirs.waterborne hazard assessment,lakes and reservoirs,machine learning,Self-Organizing Maps,SHARC

The objectives of this Phase II study are to develop a complete end-to-end capability for the operational hazard assessment modeling within limnological systems for all of CONUS and to develop preliminary techniques to determine limnological water quality parameters of OCONUS systems, allowing for the eventual hazard assessment modeling in denied or restricted regions. Achieving these objectives will require the acquisition and analysis of a significant amount of geospatial, in-situ physicochemical datasets and hyperspectral overhead imagery. The sheer volume of information and the success of the Phase I effort necessitate the use of machine learning techniques such as SOMs and supervised Artificial Neural Networks (ANN). Specifically, we will: leverage the techniques and lessons learned from Phase I to develop a complete database of the water quality parameters for all watersheds within CONUS; incorporate the results into the SHARC framework for use as initial conditions for all limnologic systems greater than 3 acres; and utilize the large amount of in-situ water quality data in conjunction with overhead hyperspectral imagery, such as LANDSAT, to train ANNs to predict parameters such as pH and TSS within OCONUS limnological systems.