H41F-1149 REAL-TIME IMPLEMENTATION OF THE PENN STATE ... · Bhatt, G. et al., PIHMgis: A "Tightly...
1
(m) (m) REAL-TIME IMPLEMENTATION OF THE PENN STATE INTEGRATED HYDROLOGIC MODEL ---THE SHALE HILLS CRITICAL ZONE OBSERVATORY Xuan Yu 1 , Christopher Duffy 1 , Gopal Bhatt 1 , Yuning Shi 2 , Lorne Leonard 1 , Mukesh Kumar 3 1. Civil & Environmental Engineering, Pennsylvania State University, 2. Meteorology, Pennsylvania State University, 3. Nicholas School of the Environment, Duke University. Works Cited Balay, S. et al., PETSc Web page, www.mcs.anl.gov/petsc/, (2009). Bhatt, G. et al., PIHMgis: A "Tightly Coupled" GIS Framework for Integrated Hydrologic Modeling, in AGU Fall Meeting Abstracts, vol. 1, (2006). Bhatt, G., M. Kumar, and C.J. Duffy, Bridging the gap between geohydrologic data and distributed hydrologic modeling, In Proceedings of International Congress on Environmental Modeling and Software, (2008). Goodrich, D. C. et al., Runoff simulation sensitivity to remotely sensed initial soil water content, Water Resources Research. 30, no. 5 (1994). Loague, K. et al., Further testing of the Integrated Hydrology Model (InHM): event‐based simulations for a small rangeland catchment located near Chickasha, Oklahoma, Hydrological Processes 19, no. 7 (2005). Qu, Y. and C. J. Duffy, A semidiscrete finite volume formulation for multiprocess watershed simulation, Water Resources Research. 43 (2007). Kumar, M. Toward a Hydrologic Modeling System, PhD Thesis, Penn State University, (2009). Kumar, M., G. Bhatt, and C.J. Duffy, The Role of Physical, Numerical and Data Coupling in a Mesoscale Watershed Model, Advances in Water Resources, (2008). Kumar, M., G. Bhatt, and C.J. Duffy, An efficient domain decomposition framework for accurate representation of geodata in distributed hydrologic models, IJGIS, (2009). Forecast Results 2009 data is used as a prototype, before we implement the actual weather forecast. The model started at 2009-1-1. Weather forecasting data came from the observation data with a slight random fluctuation in precipitation and temperature. We are adding to forcing data with one-day records every hour. The forcing data for forecast is three days longer than that of real-time simulation, which represents the weather forecasting results. So we are doing one-day real-time simulation and three-day forecast. A simple webpage of discharge rate figure shows system output. Real-time simulation starts one day ahead, and forecast model ends three days after. http://cataract.cee.psu.edu/rt/plot/webpages/index.html Hindcast Results The Shale Hills Introduction The future of river forecasting system will utilize monitoring station networks with continuous real-time measurement of hydrologic, atmospheric, and pedologic variables across diverse terrestrial water cycle. Physical model simulation is the best method to integrate abundant observation data and predict hydrologic variables directly. In recant years, physical-based distributed hydrologic models have evolved to be an up-and-coming trend of hydrology. Initial conditions have been proved crucial to hydrological models in many studies. Initial unsaturated storage and saturated storage are important in numerical modeling of rainfall and runoff response at the event scale (Goodrich, 1994). The uncertainty in initial soil-water content estimates for event-based simulation is shown to be a major limitation for physics-based models (Loague, 2005). The research provides a framework to implement a physical-based integrated hydrologic model (Penn State Integrated Hydrologic Model) for real-time hydrologic forecasting, and targets a feedback strategies gaining model improvement from forecasting results. Penn State Integrated Hydrologic Model (PIHM) PIHM(Penn State Integrated Hydrologic Model) represents a strategy for the formulation and solution of fully-coupled process equations at the watershed and river basin scales, and includes a tightly coupled GIS tool for data handling, domain decomposition, optimal unstructured grid generation, and model parameterization. PIHM has been successfully applied at different scales of watersheds. http://www.pihm.psu.edu/ R eal‐Time Hydrologic Monitoring Network NCEP Forecast Model R esults Real‐Time S imulation Module Real‐Time S imulation Module Forecast Module Forecast Module state variables state variables Output Input Input Time Step 1 Time Step 2 Updating Updating Output Input Input Real‐Time Web Publishing Forcing data and initial conditions are major input of hydrologic model. A real-time hydrologic monitoring network has been developed to support data observation. The Real- time Forecast Framework will incorporate data monitor system with hydrologic modeling system. There are two modeling systems (one for forecast, one for real-time simulation) running independently. Both the systems keep checking if the forcing data is updated. Once forcing data is updated, the corresponding system will restart from the last stop of real-time simulation with the state variable of real-time simulation as initial condition, and move forward until the end of the updated forcing data. Real-time Forecast Framework Future Work Test on simulation and forecast efficiency Incorporating feedback strategies of simulation and forecast efficiency Web design of spatial figures and temporal serials of unsaturated storage, saturated storage, evapotranspiration and discharge rate. River basin scale application with improvement of the the computational performance. PETSc is being implemented for large-scale application. AGU 2010 H41F-1149 Hydrologic Real‐Time Forecast System Real‐Time Monitoring Network Real‐Time Simulation Short‐Term Hydrologic Forecast Forcing Data Initial Condition ∙ Simulation Results ∙ Fitting Coefficient ∙ Forecast Precision ∙ Forecast Efficiency Feedback Feedback ∙ Sensitivity Analysis ∙ Network Planning Redesign Spatial forecasting result: average from 6/20/09 to 6/22/09 Spatial forecasting result: average from 6/22/09 to 6/24/09 Forecast error of groundwater level on 6/20/09 Forecast error of groundwater level on 6/22/09 Juniata River Basin Pennsylvania Shaver’s Creek Watershed Forecast error is represented by the variance between real-time simulation and forecast results. The regions with large errors can be future monitor sites.
Transcript of H41F-1149 REAL-TIME IMPLEMENTATION OF THE PENN STATE ... · Bhatt, G. et al., PIHMgis: A "Tightly...
(m) (m)
REAL-TIME IMPLEMENTATION OF THE PENN STATE INTEGRATED HYDROLOGIC MODEL ---THE SHALE HILLS CRITICAL ZONE OBSERVATORY
Xuan Yu1, Christopher Duffy1, Gopal Bhatt1, Yuning Shi2, Lorne Leonard1, Mukesh Kumar3
1. Civil & Environmental Engineering, Pennsylvania State University, 2. Meteorology, Pennsylvania State University, 3. Nicholas School of the Environment, Duke University.
Works Cited Balay, S. et al., PETSc Web page, www.mcs.anl.gov/petsc/, (2009). Bhatt, G. et al., PIHMgis: A "Tightly Coupled" GIS Framework for Integrated Hydrologic Modeling, in AGU Fall Meeting Abstracts, vol. 1, (2006). Bhatt, G., M. Kumar, and C.J. Duffy, Bridging the gap between geohydrologic data and distributed hydrologic modeling, In Proceedings of International Congress on Environmental Modeling and Software, (2008). Goodrich, D. C. et al., Runoff simulation sensitivity to remotely sensed initial soil water content, Water Resources Research. 30, no. 5 (1994). Loague, K. et al., Further testing of the Integrated Hydrology Model (InHM): event‐based simulations for a small rangeland catchment located near Chickasha, Oklahoma, Hydrological Processes 19, no. 7 (2005). Qu, Y. and C. J. Duffy, A semidiscrete finite volume formulation for multiprocess watershed simulation, Water Resources Research. 43 (2007). Kumar, M. Toward a Hydrologic Modeling System, PhD Thesis, Penn State University, (2009). Kumar, M., G. Bhatt, and C.J. Duffy, The Role of Physical, Numerical and Data Coupling in a Mesoscale Watershed Model, Advances in Water Resources, (2008). Kumar, M., G. Bhatt, and C.J. Duffy, An efficient domain decomposition framework for accurate representation of geodata in distributed hydrologic models, IJGIS, (2009).
Forecast Results 2009 data is used as a prototype, before we implement the actual weather forecast. The model started at 2009-1-1. Weather forecasting data came from the observation data with a slight random fluctuation in precipitation and temperature. We are adding to forcing data with one-day records every hour. The forcing data for forecast is three days longer than that of real-time simulation, which represents the weather forecasting results. So we are doing one-day real-time simulation and three-day forecast. A simple webpage of discharge rate figure shows system output. Real-time simulation starts one day ahead, and forecast model ends three days after. http://cataract.cee.psu.edu/rt/plot/webpages/index.html
Hindcast Results
The Shale Hills Introduction The future of river forecasting system will utilize monitoring station networks with continuous real-time measurement of hydrologic, atmospheric, and pedologic variables across diverse terrestrial water cycle. Physical model simulation is the best method to integrate abundant observation data and predict hydrologic variables directly. In recant years, physical-based distributed hydrologic models have evolved to be an up-and-coming trend of hydrology. Initial conditions have been proved crucial to hydrological models in many studies. Initial unsaturated storage and saturated storage are important in numerical modeling of rainfall and runoff response at the event scale (Goodrich, 1994). The uncertainty in initial soil-water content estimates for event-based simulation is shown to be a major limitation for physics-based models (Loague, 2005). The research provides a framework to implement a physical-based integrated hydrologic model (Penn State Integrated Hydrologic Model) for real-time hydrologic forecasting, and targets a feedback strategies gaining model improvement from forecasting results.
Penn State Integrated Hydrologic Model (PIHM) PIHM(Penn State Integrated Hydrologic Model) represents a strategy for the formulation and solution of fully-coupled process equations at the watershed and river basin scales, and includes a tightly coupled GIS tool for data handling, domain decomposition, optimal unstructured grid generation, and model parameterization. PIHM has been successfully applied at different scales of watersheds. http://www.pihm.psu.edu/
R eal‐T ime HydrologicMonitoring Network
NC E P F orecast Model R esults
Real‐T ime S imulationModule
R eal‐T ime S imulationModule
F orecas t Module
F orecas t Module
s tate variables
s tate variables
Output
Inpu tInput
T im e S te p 1
T im e S te p 2
Updating
Updating
Output
Inpu tInput
R e a l ‐T im e We b P ub l i s h in g
Forcing data and initial conditions are major input of hydrologic model. A real-t ime hydrologic monitor ing network has been developed to support data observation. The Real-t i m e F o r e c a s t F r a m e w o r k w i l l incorporate data monitor system with hydrologic modeling system. There are two modeling systems (one for forecast , one for real - t ime simulation) running independently. Both the systems keep checking if the forcing data is updated. Once forcing data is updated, the corresponding system will restart from the last stop of real-time simulation with the state variable of real-time simulation as initial condition, and move forward until the end of the updated forcing data.
Real-time Forecast Framework
Future Work Test on simulation and forecast efficiency Incorporating feedback strategies of simulation and forecast efficiency Web design of spatial figures and temporal serials of unsaturated storage, saturated storage,
evapotranspiration and discharge rate. River basin scale application with improvement of the the computational performance. PETSc is being
implemented for large-scale application.
AGU 2010 H41F-1149
H ydro log ic R e a l‐T im e F o re c a s t S y s tem
R e a l‐T im e Mon ito ring N e tw ork
R e a l‐T im eS im u la t ion
S hort‐T e rmH ydro log ic F o re c a s t
F orc ing D a ta In it ia l C ond it ion
∙ S im u la t ion R e s u lts∙ F itt ing C oe ffic ie n t
∙ F o re c a s t P re c is ion∙ F o re c a s t E ffic ie nc y
F e e dba c k F e e dba c k
∙ S e ns it iv ity A na ly s is∙ Ne tw ork P la nn ing
R e de s ig n
Spatial forecasting result: average from 6/20/09 to 6/22/09 Spatial forecasting result: average from 6/22/09 to 6/24/09
Forecast error of groundwater level on 6/20/09 Forecast error of groundwater level on 6/22/09
Juniata River Basin
Pennsylvania
Shaver’s Creek Watershed
Forecast error is represented by the variance between real-time simulation and forecast results.
The regions with large errors can be future monitor sites.
