Project Abstract for Resubmitted Proposal 0711491 (PI Shoemaker) In order to make effective use of watershed field data, it is necessary to have a watershed model. It is also essential to have a computationally feasible method for calibrating a model and assessing the uncertainty of model predictions. The focus of this project is on computationally feasible methods for spatially distributed models of large watersheds, including nutrient transport as well as flow. In this proposal new and recently developed methods by the PI will be applied to the Cannonsville Watershed (1,200 km2), which is a source of New York City?s water supply.
The objectives of the proposed research include: 1. Automatic Calibration and Uncertainty Analysis: We aim to provide a transformative general methodology to do automatic calibration, multivariate sensitivity analysis, and uncertainty analysis for large watershed models. Earlier methods require thousands of simulations, which could take over a year of computation for a large model. The focus for this proposal is to apply the new methods for automatic calibration and uncertainty analysis for the first time to watersheds. 2. Augmentation of Sensor/monitoring Networks: We will also develop a procedure to determine the best locations to add new sensors or monitoring stations to integrate with an existing data collection network. The analysis uses a watershed model to evaluate the value of new data obtained by the new sensors and compares to the values of alternative schemes for data collection that differ in terms of constituents measured and the location or times of measurement. The analysis incorporates the tradeoff between resources and accuracy. 3. Broader Impact: We will aim to have a broad impact by a) generating methods and software that can be used internationally with many watersheds and models, b) provide better predictive tools for the Cannonsville which has a huge environmental impact and an effect on millions of people, c) continue the PI?s practice of recruiting and training underrepresented PhD students, and work with Cornell ADVANCE program to help women faculty, and d) use REUs and augment course materials. 4. Intellectual Merit: The intellectual merit is associated with the importance of the methods for watershed analysis, and the originality of the methods being developed.
PI Shoemaker, NSF EAR 0711491 Water is one of the world’s most precious commodities. Hydrology is the study of how water moves or is stored over the land surface, in the subsurface and in the atmosphere, and how this affects the movement and storage of water’s constituents, which can include pollutants. The science of hydrology helps us understand and quantify this movement and storage of water and its constituents. In this project it is our aim to develop and apply computational methods called algorithms to improve our quantitative understanding of overland and subsurface hydrology with regard to water and contaminant transport. We have looked at a number of contaminants including subsurface contaminants TNT and Carbon Dioxide plumes, and we focus on phosphorous in watershed studies. Phosphorous is an essential nutrient for plant growth and is very beneficial to crop yield. However, movement of water bourn phosphorous into lakes and rivers can cause eutrophication, which is the excessive growth of plants like algae. In this situation phosphorous is a pollutant. One of our applications in this study is the Cannonsville Watershed in upstate New York State, which is a major supplier of drinking water to New York City. Currently NYC is not required to filter this water because it is so pure. However, if phosphorous transport increases the algal population, NYC will be required to start filtering the water and a minimum estimate for the cost of a filtration plant is $8 Billion. Instead NY state and NYC environmental agencies and university scientists are using modeling of the watershed to help understand phosphorous transport and the Best management practices that can be employed to protect the water supply from excess phosphorous. We are currently finishing a study on sensor network design for this watershed to improve the procedure for data collection to better understand the hydrology. We were able to use our computational methods to evaluate the impact of two competing hydrologic processes (Hortonian flow and Variable Source Area) on phosphorous transport and flow and we showed the temporal changes in this impact. All of this analysis required optimization algorithms. In this project we have developed general computational optimization, sensitivity analysis, and uncertainty quantification algorithms that can be used in multiple watersheds around the world to help extract more information from existing data to understand the movement of water and contaminants. This analysis includes optimization algorithms for estimating the best values of parameters in a model and methods for assessing the uncertainty of those model predictions. Our focus is on algorithms that can obtain good answers with relative few simulations of the model since computer time for each run of a model can be substantial, especially for subsurface and atmospheric models. For example some previous uncertainty analysis procedures used up to 60,000 model simulation, which would take 7 years of computing if the simulations each take an hour. The focus of our algorithms is to improve upon the previous algorithms for situations in which the total number of simulations that can be done is limited. We do this through the use of iterative surrogate response surfaces and/or control of the perturbation of the dimension of the decision vector. Our papers compare results of our algorithms to other algorithms and in most cases show a definite improvement. Intellectual Merit: The intellectual merit is associated with the importance and originality of the algorithms developed for computationally efficient hydrologic analysis and the contribution to hydrologic science arising from the applications of these methods. This includes a better understanding of the dynamic interaction between Variable Source Hydrology and Hortonian Flow. . Broader Impact: The problems being studied here associated with water bourn contamination and its control have a tremendously broad impact around the world. This has a specifically important impact for the taxpayers of NYC by helping them avoid an $8 Million water filtration plant. The computational tools developed can be used on a variety of models including other hydrologic problems and other problems in science and engineering. The PI has continued to recruit and train underrepresented PHD students and currently has 5 women PhD students.
