Improving hydrologic simulations Martyn Clark (and many others)

Slides:



Advertisements
Similar presentations
PROCESS-BASED, DISTRIBUTED WATERSHED MODELS New generation Source waters and flowpaths Physically based.
Advertisements

Regional water cycle studies: Current activities and future plans Water System Retreat, NCAR 14 January 2015 Martyn Clark, Naoki Mizukami, Andy Newman,
Watershed Hydrology, a Hawaiian Prospective: Evapotranspiration Ali Fares, PhD Evaluation of Natural Resource Management, NREM 600 UHM-CTAHR-NREM.
NWS Calibration Workshop, LMRFC March, 2009 Slide 1 Sacramento Model Derivation of Initial Parameters.
How do model errors and localization approaches affects model parameter estimation Juan Ruiz, Takemasa Miyoshi and Masaru Kunii
AMS 25th Conference on Hydrology
New Directions for WRF Land Surface Modeling 1 Polar WRF Workshop – 3 November 2011 Michael Barlage Research Applications Laboratory (RAL) National Center.
Introduction The agricultural practice of field tillage has dramatic effects on surface hydrologic properties, significantly altering the processes of.
Atmospheric Analysis Lecture 3.
Key considerations for simulating Arctic weather and climate with limited area models Nicole Mölders University of Alaska Fairbanks, Geophysical Institute.
Scale Issues in Hydrological Modelling: A Review
VIC Model Status Blowing Snow and Lake Algorithms Princeton Meeting December 4, 2006.
Globally distributed evapotranspiration using remote sensing and CEOP data Eric Wood, Matthew McCabe and Hongbo Su Princeton University.
Engineering Hydrology (ECIV 4323)
Are the results of PILPS or GSWP affected by the lack of land surface- atmosphere feedback? Is the use of offline land surface models in LDAS making optimal.
A Macroscale Glacier Model to Evaluate Climate Change Impacts in the Columbia River Basin Joseph Hamman, Bart Nijssen, Dennis P. Lettenmaier, Bibi Naz,
Hydrology and Water Resources Civil and Environmental Engineering Dept. Physically-based Distributed Hydrologic Modeling.
Development of hyper-resolution large-ensemble continental-scale hydrologic model simulations AGU, San Francisco, CA 14 December 2014 Martyn Clark, Naoki.
The Calibration Process
CSIRO LAND and WATER Estimation of Spatial Actual Evapotranspiration to Close Water Balance in Irrigation Systems 1- Key Research Issues 2- Evapotranspiration.
Water Supply Forecast using the Ensemble Streamflow Prediction Model Kevin Berghoff, Senior Hydrologist Northwest River Forecast Center Portland, OR.
Land Surface Models & Surface Water Hydrology Cédric DAVID.
Improved Land Modeling for Drought Monitoring and Seasonal Hydrological Prediction Including Groundwater Mickael Ek, Rongqian Yang, Youlong Xia, Jesse.
Martyn Clark (NCAR/RAL) Bart Nijssen (UW) Building a hydrologic model: Spatial approximations, process parameterizations, and time stepping schemes CVEN.
MODELING OF COLD SEASON PROCESSES Snow Ablation and Accumulation Frozen Ground Processes.
LINDSEY NOLAN WILLIAM COLLINS PETA-APPS TEAM MEETING OCTOBER 1, 2009 Stochastic Physics Update: Simulating the Climate Systems Accounting for Key Uncertainties.
Evaporation Slides prepared by Daene C. McKinney and Venkatesh Merwade
Discussion and Future Work With an explicit representation of river network, CHARMS is capable of capturing the seasonal variability of streamflow, although.
Advancements in Simulating Land Hydrologic Processes for Land Surface Modeling (LSM) Hua Su Presentation for Physical Climatology.
CryosPheric responses to Anthropogenic PRessures in the HIndu Kush-Himalaya regions: impacts on water resources and society adaptation in Nepal DHM Centre.
Advances in Macroscale Hydrology Modeling for the Arctic Drainage Basin Dennis P. Lettenmaier Department of Civil and Environmental Engineering University.
Hydrologic model benchmarks: Synthetic test cases, CZO data, and continental-scale diagnostics CUAHSI Community Modeling Working Group, San Francisco,
Evaporation What is evaporation? How is evaporation measured? How is evaporation estimated? Reading: Applied Hydrology Sections 3.5 and 3.6 With assistance.
Streamflow Predictability Tom Hopson. Conduct Idealized Predictability Experiments Document relative importance of uncertainties in basin initial conditions.
Centro de Previsão de Tempo e Estudos Climáticos (CPTEC/INPE) São Paulo, Brazil ( Integrated observed and modeled atmospheric water budget.
How does the choice/configuration of hydrologic models affect the portrayal of climate change impacts? Pablo Mendoza 1.
LL-III physics-based distributed hydrologic model in Blue River Basin and Baron Fork Basin Li Lan (State Key Laboratory of Water Resources and Hydropower.
Coupling of the Common Land Model (CLM) to RegCM in a Simulation over East Asia Allison Steiner, Bill Chameides, Bob Dickinson Georgia Institute of Technology.
Impact Of Surface State Analysis On Estimates Of Long Term Variability Of A Wind Resource Dr. Jim McCaa
Inquiry into the appropriateness of a TILE/MOSAIC approach for the representation of surface inhomogeneities B. Ritter and J. Helmert.
Evaluating hydrological model structure using tracer data within a multi-model framework Hilary McMillan 1, Doerthe Tetzlaff 2, Martyn Clark 3, Chris Soulsby.
TURBULENT FLUX VARIABILITIES OVER THE ARA WATERSHED Moussa Doukouré, Sandrine Anquetin, Jean-Martial Cohard Laboratoire d’étude des Transferts en Hydrologie.
Variation of Surface Soil Moisture and its Implications Under Changing Climate Conditions 1.
Multi-Model or Post- processing: Pros and Cons Tom Hopson - NCAR Martyn Clark - NIWA Andrew Slater - CIRES/NSIDC.
Evapotranspiration Partitioning in Land Surface Models By: Ben Livneh.
Adjustment of Global Gridded Precipitation for Orographic Effects Jennifer Adam.
Understanding hydrologic changes: application of the VIC model Vimal Mishra Assistant Professor Indian Institute of Technology (IIT), Gandhinagar
Towards development of a Regional Arctic Climate System Model --- Coupling WRF with the Variable Infiltration Capacity land model via a flux coupler Chunmei.
Experience with modelling of runoff formation processes at basins of different scales using data of representative and experimental watersheds Olga Semenova.
Results Time Study Site Measured data Alfalfa Numerical Analysis of Water and Heat Transport in Vegetated Soils Using HYDRUS-1D Masaru Sakai 1), Jirka.
Hydro-Thermo Dynamic Model: HTDM-1.0
The Islamic University of Gaza Faculty of Engineering Civil Engineering Department EENV 5326 Groundwater Modeling.
Diagnosis of Performance of the Noah LSM Snow Model *Ben Livneh, *D.P. Lettenmaier, and K. E. Mitchell *Dept. of Civil Engineering, University of Washington.
Runoff Overview Tom Hopson.
SiSPAT-Isotope model Better estimates of E and T Jessie Cable Postdoc - IARC.
GAINING CONFIDENCE IN HYDROLOGIC SIMULATIONS: AN ALTERNATIVE RESEARCH STRATEGY FOR THE DEVELOPMENT AND APPLICATION OF HYDROLOGIC MODELS Martyn Clark, Hilary.
An advanced snow parameterization for the models of atmospheric circulation Ekaterina E. Machul’skaya¹, Vasily N. Lykosov ¹Hydrometeorological Centre of.
From catchment to continental scale: Issues in dealing with hydrological modeling across spatial and temporal scales Dennis P. Lettenmaier Department of.
NOAA Northeast Regional Climate Center Dr. Lee Tryhorn NOAA Climate Literacy Workshop April 2010 NOAA Northeast Regional Climate.
Towards development of a Regional Arctic Climate System Model ---
Precipitation-Runoff Modeling System (PRMS)
Engineering Hydrology (ECIV 4323)
The Calibration Process
Utah Water Research Laboratory
Introduction to Land Information System (LIS)
Multimodel Ensemble Reconstruction of Drought over the Continental U.S
Mire parameterization
A Multimodel Drought Nowcast and Forecast Approach for the Continental U.S.  Dennis P. Lettenmaier Department of Civil and Environmental Engineering University.
Multimodel Ensemble Reconstruction of Drought over the Continental U.S
J.T. Kiehl National Center for Atmospheric Research
Presentation transcript:

Improving hydrologic simulations Martyn Clark (and many others)

Outline Introduction: Why is there a problem? Approach: A more controlled approach to model development and parameter identification Discussion: Strategy to meet project deliverables

Subjectivity in model selection: How does the choice of model equations impact simulations of hydrologic processes? Missing processes, inappropriate parameterizations? Subjectivity in selecting/applying models Define a-priori values for model parameters Decide what model parameters we adjust, if any Decide what calibration strategy we implement, if any  Choice of objective function  Choice of forcing data and calibration period Model parameters Decide which processes to include Define parameterizations for individual processes Define how individual processes combine to produce the system-scale response Solve model equations Model structure Subjectivity in parameter identification: How does our choice of model parameters impact simulations of hydrologic processes? Compensatory effects of model parameters (right answers for the wrong reasons)? Climate change studies commonly involve several methodological choices that might impact the hydrologic sensitivities obtained. In particular:

Current approaches to model development: Are they adequate? Scrutiny during model development –Ideally, a discerning model developer will carefully scrutinize each modeling decision and thoughtfully evaluate alternatives –However, although multiple alternatives may be considered when a model is developed, it is typical that only one approach is implemented and tested (or one approach is reported). Model evaluation along the axis of complexity –Top-down approach, etc. –Effectively restricts the investigation to a single branch of the model development tree Rejectionist frameworks, e.g., GLUE –Typically an uncontrolled approach to model evaluation Model inter-comparison experiments –Weak methods for model evaluation (not focused on processes) –Difficult to attribute inter-model differences to specific processes Key community objectives: Improved representation of observed processes More precise representation of model uncertainty Key community objectives: Improved representation of observed processes More precise representation of model uncertainty

Current parameter identification approaches: Are they adequate? Deterministic model calibration –The calibration process is often poorly constrained (e.g., a single objective function) –Parameters for individual model sub-components may be assigned unrealistic values during calibration in order to compensate for unreaslitic parameters in another part of the model or weaknesses in structure and uncertainty in model forcing Regionalization –Basin-by-basin calibration produces parameter sets in different basins that are fitted to the noise in the input-response data –It is difficult to establish regional relationships between calibrated model parameters and basin characteristics A-priori parameter estimation –Many model parameters are not directly observable Key community objective: Physically realistic parameter estimates from headwater catchments to continental scales Key community objective: Physically realistic parameter estimates from headwater catchments to continental scales

Outline Introduction: Why is there a problem? Approach: A more controlled approach to model development and parameter identification Discussion: Strategy to meet project deliverables

Advocate pursuing the method of multiple working hypotheses Scientists often develop “parental affection” for their theories T.C. Chamberlain Chamberlin’s method of multiple working hypotheses “…the effort is to bring up into view every rational explanation of new phenomena… the investigator then becomes parent of a family of hypotheses: and, by his parental relation to all, he is forbidden to fasten his affections unduly upon any one” Chamberlin (1890)

many options The modeling decisions include –Choice of processes to include/exclude –Choice of parameterizations for individual processes –Choice of model architecture (how different methods combine to produce the system-scale response) precipitationevaporation vertical percolation surface runoff For example, a possible state equation for the unsaturated zone is VIC parameterization... TOPMODEL parameterization Two popular models: Understanding differences among models

PRMSSACRAMENTO ARNO/VICTOPMODEL pe S1TS1T S1FS1F S2TS2T S 2 FA S 2 FB q if qbAqbA qbBqbB q 12 ep q sx qbqb q 12 S2S2 S1S1 GFLWR S2S2 S1S1 ep q 12 q sx qbqb ep S 1 TA q sx q if qbqb S 1 TB S1FS1F q 12 S2S2 Clark, M.P., A.G. Slater, D.E. Rupp, R.A. Woods, J.A. Vrugt, H.V. Gupta, T. Wagener, and L.E. Hay (2008) Framework for Understanding Structural Errors (FUSE): A modular framework to diagnose differences between hydrological models. Water Resources Research, 44, W00B02, doi: /2007WR FUSE: Framework for Understanding Structural Errors

The multiple-hypothesis framework: A “more controlled” approach to model evaluation 10 Isolate hypotheses Accommodate different decisions regarding process selection Accommodate different options for model architecture Separate the hypothesized model equations from their solutions Evaluate hypotheses Sensitivity analysis (understand reasons for inter-model differences) Extensive evaluation using research data (test internal components of the model) Clever use of routine observing networks (“large sample” hydrology, but not as you know it).

Build multiple-hypothesis representation of “treetop to stream” domain snow soil Facilitates experimenting with.. 1) Different constitutive functions & parameters Albedo, turbulent heat transfer Soil hydraulic properties 2) Model architecture Surface water – groundwater interactions Sub-grid variability and lateral flow of water

Modeling approach Numerical implementation Fine spatial discretization Adaptive sub-stepping with numerical error control (tight tolerance) Fine-grain modularity, with numerical solutions clearly separated from model physics Most subroutines return fluxes and their derivatives, which are used in solver routines Limited use of existing multi- physics codes (e.g., Noah- MP) 12

Example modeling decisions 13 Parameterizations –Snow Different snow albedo parameterizations Different thermal conductivity parameterizations Different compaction parametrizations –Turbulent heat transfer Different atmospheric stability parameterizations –Transpiration (from Noah-MP) Different soil stress and stomatal resistance functions –Storage and transmission of liquid water in soil Different forms of Richards’ equation Flexibility in the choice of hydraulic conductivity profile Flexibility in choice of lower boundary condition –Vegetation traits Different parameterizations for veg roughness and displacement height Architecture –Groundwater parameterizations Non-interactive VIC-style, interactive Topmodel style, mixed form of Richards’ equation –Overall model architecture Representation of spatial variability, linkages among components > 100 model parameters

Example: Turbulent exchange coefficients

Example: Transmission and storage of liquid water within the snowpack

Example simulations for Reynolds Creek, Idaho

17 Datasets from: Reba et al. (WRR, 2011) Flerchinger et al. (JHM, 2012)

18

Simulations of longwave fluxes above the Aspen grove Comparison of 1)combined surface-atmosphere and canopy-atmosphere longwave radiation fluxes (FUSEv2 model) 2)above-canopy upward longwave observations (Flerchinger, 2012) MODEL OBS (missing data)

Simulations of below-canopy windspeed Uses serially-complete forcing from exposed site to enable multi-decade simulations Simulated below-canopy windspeed (red) compared with observed below-canopy windspeed (blue)

Partitioning of energy between sensible and latent heat Total sensible heat fluxTotal latent heat flux FUSEv2 simulations Eddy flux observations Two issues: 1)Parameterization uncertainty: impacts of seasonally frozen ground on surface runoff during the melt season and plant-available water in the growing season 2)Architectural uncertainty: non-local sources of soil moisture in the growing season

Spatial variability and hydrologic connectivity 22 Hydrologic response units –Different meteorological forcing –Different frozen precipitation multipliers –Different vegetation and terrain properties Hydrologic connectivity –Fluxes in each HRU computed individually –Use dynamic TOPMODEL and DHVSM concepts # to compute flow between HRUs # Modeling approach:  No prognostic water table Baseflow computed based on ratio of total water storage in the soil column to total storage capacity Net baseflow flux (outflow – inflow) added as a sink term to Richards’ equation  Use of HRUs instead of a high-resolution grid With connectivity

Distributed simulations – without connectivity

Distributed simulations – with connectivity

Outline Introduction: Why is there a problem? Approach: A more controlled approach to model development and parameter identification Discussion: Strategy to meet project deliverables

Summary of model structure analysis Status: Built a comprehensive multiple-hypothesis “process-based” hydrologic model for the domain treetops to stream –Framework useful to identify a sub-set of “satisfying” modeling options and improve simulations of hydrologic processes –Framework useful for physics-based estimates of uncertainty Multi-physics models (multiple parameterizations for individual processes) not be necessary to quantify model uncertainty – it’s the parameters, stupid! Differences in model architecture are critical Ongoing work: Understand impact of the (subjective) decisions made during model development –Extensive analysis using data from research basins –Attribute inter-model differences to choice of both model parameterizations and model architecture Medium-term goal: Use framework for ensemble continental-scale hydrologic simulations –Improve simulations of hydrologic processes –Quantify model uncertainty from a physical perspective

Planned steps for parameter estimation Low dimensional multi-response inference –Identify the “mapping” between different model parameters and different diagnostic signatures of hydrologic behavior –Decompose the high-dimensional problem (prone to compensatory errors) into a set of lower-dimensional sub-problems –Use a mix of local-scale and large-scale signatures to avoid over-fitting to the idiosyncrasies of individual watersheds Focus attention on the parameters in pedotransfer functions (and other transfer functions), rather than the model parameters themselves

Key deliverable: multi-model simulations of climate change impacts Model fidelity –Incremental progress: Improve estimates of parameters in a small set of existing models –More noteworthy advance: Improve representation of physical processes using modeling options available in FUSEv2 Model uncertainty –Incremental progress: Use inter-model difference as a proxy for model uncertainty –More noteworthy advance: Quantify uncertainty using a mix of parameter perturbations and model structural choices