Biases in land surface models Yingping Wang CSIRO Marine and Atmospheric Research.

Slides:



Advertisements
Similar presentations
Dynamical Prediction of the Terrestrial Ecosystem and the Global Carbon Cycle: a 25-year Hindcast Experiment Jin-Ho Yoon Dept. Atmospheric and Oceanic.
Advertisements

Land Surface Evaporation 1. Key research issues 2. What we learnt from OASIS 3. Land surface evaporation using remote sensing 4. Data requirements Helen.
Using Flux Observations to Improve Land-Atmosphere Modelling: A One-Dimensional Field Study Robert Pipunic, Jeffrey Walker & Andrew Western The University.
Modelling Australian Tropical Savanna Peter Isaac 1, Jason Beringer 1, Lindsay Hutley 2 and Stephen Wood 1 1 School of Geography and Environmental Science,
Detection of a direct carbon dioxide effect in continental river runoff records N. Gedney, P. M. Cox, R. A. Betts, O. Boucher, C. Huntingford & P. A. Stott.
Reading: Text, (p40-42, p49-60) Foken 2006 Key questions:
School Research Conference, March 2009 Jennifer Wright Supervisors: M.Williams, G. Starr, R.Mitchell, M.Mencuccini Fire and Forest Ecosystems in the Southeastern.
03/06/2015 Modelling of regional CO2 balance Tiina Markkanen with Tuula Aalto, Tea Thum, Jouni Susiluoto and Niina Puttonen.
Cross-spectral analysis on Net Ecosystem Exchange: Dominant timescale and correlations among key ecosystem variables over the Ameriflux Harvard forest.
Biosphere Modeling Galina Churkina MPI for Biogeochemistry.
4. Testing the LAI model To accurately fit a model to a large data set, as in the case of the global-scale space-borne LAI data, there is a need for an.
CSIRO LAND and WATER Estimation of Spatial Actual Evapotranspiration to Close Water Balance in Irrigation Systems 1- Key Research Issues 2- Evapotranspiration.
Global Carbon Cycle Feedbacks: From pattern to process Dave Schimel NEON inc.
Remote Sensing Data Assimilation for a Prognostic Phenology Model How to define global-scale empirical parameters? Reto Stöckli 1,2
Optimising ORCHIDEE simulations at tropical sites Hans Verbeeck LSM/FLUXNET meeting June 2008, Edinburgh LSCE, Laboratoire des Sciences du Climat et de.
Laboratoire des Sciences du Climat et de l'Environnement P. Peylin, C. Bacour, P. Ciais, H. Verbeek, P. Rayner Flux data to highlight model deficiencies.
Paul R. Moorcroft David Medvigy, Stephen Wofsy, J. William Munger, M. Dietze Harvard University Developing a predictive science of the biosphere.
Real-time integration of remote sensing, surface meteorology, and ecological models.
Data assimilation in land surface schemes Mathew Williams University of Edinburgh.
Summary of Research on Climate Change Feedbacks in the Arctic Erica Betts April 01, 2008.
A process-based, terrestrial biosphere model of ecosystem dynamics (Hybrid v. 3.0) A. D. Friend, A.K. Stevens, R.G. Knox, M.G.R. Cannell. Ecological Modelling.
BIOME-BGC estimates fluxes and storage of energy, water, carbon, and nitrogen for the vegetation and soil components of terrestrial ecosystems. Model algorithms.
Results from the Carbon Cycle Data Assimilation System (CCDAS) 3 FastOpt 4 2 Marko Scholze 1, Peter Rayner 2, Wolfgang Knorr 1 Heinrich Widmann 3, Thomas.
A detailed look at the MOD16 ET algorithm Natalie Schultz Heat budget group meeting 7/11/13.
Water and Carbon Cycles in Heterogeneous Landscapes: An Ecosystem Perspective Chapter 4 How water and carbon cycles connect the organizational levels of.
Land Surface Processes in Global Climate Models (1)
Remote Sensing of LAI Conghe Song Department of Geography University of North Carolina Chapel Hill, NC
The Lagrangian time scale for turbulent transport in forest canopies, determined from measured fluxes and concentrations and modelled source distributions.
CO 2 - Net Ecosystem Exchange and the Global Carbon Exchange Question Soil respiration chamber at College Woods near Durham New Hampshire. (Complex Systems.
Scaling and Analysis of Long Time Series of Soil Moisture Content By Gabriel Katul Nicholas School of the Environment and Earth Sciences, Duke University.
Using FLUXNET data to evaluate land surface models Ray Leuning and Gab Abramowitz 4 – 6 June 2008.
Integration of biosphere and atmosphere observations Yingping Wang 1, Gabriel Abramowitz 1, Rachel Law 1, Bernard Pak 1, Cathy Trudinger 1, Ian Enting.
Some challenges of model-data- integration a collection of issues and ideas based on model evaluation excercises Martin Jung, Miguel Mahecha, Markus Reichstein,
BioVeL MS11 Workshop - Ecosystem functioning & valuation web services and workflows Recent model developments in Hungary: Biome-BGC MuSo Zoltán Barcza,
A parametric and process- oriented view of the carbon system.
Impacts of leaf phenology and water table on interannual variability of carbon fluxes in subboreal uplands and wetlands Implications for regional fluxes.
The PILPS-C1 experiment Results of the first phase of the project Complementary simulation to be done Proposition for the future.
Key information from FDOS Global distribution of plant communities as described by quantitative traits [and their association with phylogenetic composition??]
Investigating Land-Atmosphere CO 2 Exchange with a Coupled Biosphere-Atmosphere Model: SiB3-RAMS K.D. Corbin, A.S. Denning, I. Baker, N. Parazoo, A. Schuh,
Using data assimilation to improve estimates of C cycling Mathew Williams School of GeoScience, University of Edinburgh.
Space-Time Variability in Carbon Cycle Data Assimilation Scott Denning, Peter Rayner, Dusanka Zupanski, Marek Uliasz, Nick Parazoo, Ravi Lokupitiya, Andrew.
Goal: to understand carbon dynamics in montane forest regions by developing new methods for estimating carbon exchange at local to regional scales. Activities:
Evapotranspiration Estimates over Canada based on Observed, GR2 and NARR forcings Korolevich, V., Fernandes, R., Wang, S., Simic, A., Gong, F. Natural.
Spatial Processes and Land-atmosphere Flux Constraining ecosystem models with regional flux tower data assimilation Flux Measurements and Advanced Modeling,
CarboEurope: The Big Research Lines Annette Freibauer Ivan Janssens.
Model-Data Synthesis of CO 2 Fluxes at Niwot Ridge, Colorado Bill Sacks, Dave Schimel NCAR Climate & Global Dynamics Division Russ Monson CU Boulder Rob.
Edinburgh, June 2008Markus Reichstein Critical issues when using flux data for reducing Land Surfcace Model uncertainties – towards full uncertainty accounting?
Mechanistic model for light-controlled phenology - its implication on the seasonality of water and carbon fluxes in the Amazon rainforests Yeonjoo Kim.
Data assimilation as a tool for biogeochemical studies Mathew Williams University of Edinburgh.
Dr. Monia Santini University of Tuscia and CMCC CMCC Annual Meeting
Perspectives on water cycling in ecosystem models Sarah Davis June 12, 2012 Water in Bioenergy Agroecosystems Workshop.
1 Xiaoyan Jiang, Guo-Yue Niu and Zong-Liang Yang The Jackson School of Geosciences The University of Texas at Austin 03/20/2007 Feedback between the atmosphere,
Production.
1 UIUC ATMOS 397G Biogeochemical Cycles and Global Change Lecture 18: Nitrogen Cycle Don Wuebbles Department of Atmospheric Sciences University of Illinois,
Forest Research, 20 February 2009 Understanding the carbon cycle of forest ecosystems: a model-data fusion approach Mathew Williams School of GeoSciences,
Shortwave and longwave contributions to global warming under increased CO 2 Aaron Donohoe, University of Washington CLIVAR CONCEPT HEAT Meeting Exeter,
SimSphere SVAT model SimSphere is available for free from Aberystwyth University, UK:
Assimilation Modeling of CO2 Fluxes at Niwot Ridge, CO, and Strategy for Scaling up to the Region William J. Sacks David S. Schimel,
Influence of tree crown parameters on the seasonal CO2-exchange of a pine forest in Brasschaat, Belgium. Jelle Hofman Promotor: Dr. Sebastiaan Luyssaert.
Community Land Model (CLM)
Ruth Doherty, Edinburgh University Adam Butler & Glenn Marion, BioSS
Conghe Song Department of Geography University of North Carolina
Department of Hydrology and Water Resources
Marcos Heil Costa Universidade Federal de Viçosa
Ecosystem Demography model version 2 (ED2)
The absorption of solar radiation in the climate system
Adam Butler & Glenn Marion, Biomathematics & Statistics Scotland •
Coherence of parameters governing NEE variability in eastern U. S
Carbon Model-Data Fusion
Presentation transcript:

Biases in land surface models Yingping Wang CSIRO Marine and Atmospheric Research

Model residuals Differences between predictions and data, and result from errors –Data: representation and precision –Model formulation –State (time varying) –Parameters (time independent)

Errors in state space Total errors: Errors due to model structure: Errors due to incorrect state and parameters values

How big are those errors? Abramowitz et al Averaging window size (day) Parameter error Systematic error Random error

Some errors can not be accounted for by parameter tuning Use the improved CBM (CABLE) Eight parameters varied within their reasonable ranges Grey region shows PDF of ensemble predictions From Abramowitz et al. 2008

Model errors If systematic model errors are not modeled the SE of optimized model parameters can are too optimistic; Estimates of model parameters can be biased;

Systematic model errors Inaccurate inputs Missing processes Low sensitivities Incorrect formulations

Incorrect inputs of LW to the model

Why does CABLE predict incorrect energy partitioning ? Haverd unpublished data

Modeling variance in the data statistically Braswell et al of 11 optimized photosynthesis parameters are well constrained. But the model still failed capturing a significant fraction of seasonal and inter- annual variations in NEE data.

Analyzing errors in frequency domain From Braswell et al Inter-annual Seasonal Daily

S ET/ET m Katual et al Incorrect response to soil water

Explaining the variance in the data Any variability that can not be modeled deterministically.. must be.. modeled statistically (Enting 2002) Analysis model residuals in both time and frequency domains

Analyzing model residual in t and f domains Time domain (t)Frequency domain (f) Residual plots SOFM Wavelet analysis Intuitive Clues for when and why models failed Separation of what the models should and should not explain at different time scales Difficult to resolve some complex interactions at different time scales Little information about why the models fail

Conclusions How many models should be calibrated? One or many? –Many. How do we address the initial condition problem? –Treat initial state values as model parameters. How do we detect and address model flaws? –SOFM, –State-space formulation –Analysis model residuals in both t and f domain, –data-model fusion as a sensitivity analysis tool –etc.

Deficiencies in land surface models Inadequate representation of canopy and soil Inaccurate formulations

Deficiencies in land surface models Overestimate heat fluxes, and because of –insulation by litter –canopy heat storage Incorrect response to soil water, and because of –incorrect model parameters –model structure

The Kalman-gain (g) Kalman gain (g)  cov( x )/cov( z ); Larger errors in data give smaller g; Lower sensitivity to z to x gives smaller gain; We need to separate model structural errors and from state and parameter errors Errors must be accounted by statistical models

Fast vs slow process Variance in EC data is dominated by the variation at diurnal and seasonal scales. Fitting LSM to EC data then gives better constraints on parameters for fast process than those for slow processes

Analyzing model residuals in frequency domain: the Bayesian approach

A consistent framework for studying model residuals

Fast biophysical processes Canopy conductance photosynthesis, leaf respiration Carbon transfer, Soil temp. & moisture availibity Slow biogeographical processes Vegetation dynamics & disturbance Land-use and land-cover change Vegetation change Autotrophic and Heterotrophic respiration Allocation Intermediate timescale biogeochemical processes Phenology Turnoover Nutrient cycle Solution of SEB; canopy and ground temperatures and fluxes Soil heat and moisture Surface water balance Update LAI, Photosyn- thesis capacity Physical- chemical forcing T,u,Pr,q, R s, R l, CO 2 Radiation water, heat, & CO2 fluxes dayyears Biogeo- chemical forcing Time scale of biosphere-atmosphere interactions Atmosphere hour

Limitations of current land surface models What is PFT? Do all plants in the same PFT truly have same parameter values? Mismatch between model and data, soil T and q for example. Spatial heterogeneity in canopy and soil Litter layer

Why EC data cannot constrain soil BGC processes? Sensitivity of turnover rate of slow pools to C fluxes is low; Soil C has a spectrum of turnover rate as substrate quality changes with time; Soil C has long memory (disturbance history, weather history etc) The parameters you obtained have limited applicability in predicting response to future climate change

State and parameter estimation

What eddy flux data can constrain effectively? Sensitivity of Biogeochemical processes (particularly slow pools) Plant phenology Vegetation dynamics

Schimel’s Figure