Literature Review:Literature Review:Simulating landSimulating land--surface surface processes in transition zones processes in transition zones with enhanced versions of with enhanced versions of Noah LSMNoah LSMEnrique RoseroDepartment of Geological SciencesJackson School of GeosciencesGEO387H Physical ClimatologyFall 2007Motivation: Motivation: The importance of turbulent fluxes in transition zones The importance of turbulent fluxes in transition zones for summer time convection initiationfor summer time convection initiationSimulating land memory mechanisms with the Simulating land memory mechanisms with the Noah Land Surface Model:Noah Land Surface Model:Role of vegetation: Role of vegetation: StomatalStomatalcontrol on evaporationcontrol on evaporationRole of soil moisture: High resolution initializationRole of soil moisture: High resolution initializationParameterization issues and inverse modeling Parameterization issues and inverse modeling SummarySummaryOutline:Relevance of transition zonesKoster et al. 2004; Koster et al. 2003; Trenberth et al. 2003Hotspots for land-atmosphere coupling: anomalies in surface states propagate to atmosphere and influence precipitationSummertime convection initiationWeckwert and Parsons 2006; Wilson and Roberts 2006Afternoon, high intensity showers are driven by convergent boundaries (triple point), drylines, and bores. In capped moist boundary layers, lifting is sensitive to low level wind shear.Model biases in precipitation simulationTrenberth et al. 2003Premature initiation of convection and drizzling are known shortcomings of GCMs. Model biases are related to land surface states and fluxes.Accurate representation is key to simulate timing, duration and intensity of precipitation.Convective potential energy (CAPE), convective inhibition (CIN)Courtesy: Courtesy: Dave Dave GochisGochisIHOP_2002: Collocated, multi-sensor observationsWeckwerth et al. 2004; LeMone et al. 2007Ek et al. 2003; Mitchell et al. 2004Land memory: the role of vegetationShuttleworth 2007; Jarvis 1976; Ball and Berry 1987; Niyogi and Raman 1997Stomatal response to environmental variations control partition of net radiation and regulate soil moisture in the root zone.Jarvis parameterizes stress factors, while for Ball-Berry, considers transpiration as the cost of photosynthesis. Ball-Berry better represents observed values of canopy resistance.Coupled simulations with GEMNoah-WRF is enhanced with photosyntesis-based ET. Canopy resistance is 500% larger than Jarvis’, transpiration rate is reduced by 60% and soil moisture is increased by 10%. Holt et al. 2006Changes in surface fluxes after calibration:Coupled simulated latent heat flux after off-line calibration reduces by 20 w/m2 respect to default parameter values runChen et al. 2007Land memory: the role of soil moistureNoah -HRLDAS increases the resolution of the assimilation system from 12 to 4 km. Overestimates point measured LE,larger soil moisture amplitude , better captures spatial patterns.Inverse modeling of van Genuchten soil parameters:Noah runs with optimal parameters for both, Clap and Hornberger and van Genuchten provide nearly identical results. Latent heat flux RMSE is reduced by up to 60 W/m2 at all sites.Gutmann and Small 2007Enhancements to Noah physics:Dynamic Vegetation after Dickinson et al., 1998NCEP/NCAR/UT – AGU Joint Assembly 05.25.07 – 1.d;ks;kdGroundwater after Niu et al., 2005, 2007.Courtesy: Lindsey GuldenHypothesis: increasing realism yields a model less sensitive to choice of parametersSuccessive estimation of parameters for 4 models: Noah-STANDARD (20 parameters: 10 soil, 10 vegetation) c Noah-DV (+8 parameters, 28 total) d Noah-GW (+4 parameters, 24 total) Noah-DVGW (+12 parameters, 32 total)Spin-up: NLDAS-forced run for ~2.5 years, no scoring. Multi-objective, 5 constraints: Sensible heat (H), latent heat (LE), ground heat (G), first layer temperature STC(1) and first layer soil moisture SMC(1), scoring only 45 days.Method: Markov Chain Monte Carlo sampler (MOSCEM, Vrugt et al., 2003). 250 par. sets approximate multivariate posterior distribution. 10,000 max number of function evaluations.Experiment setup:Calibration reduces only parameter uncertainty, exposes model error:When parameter uncertainty is minimized, model performances are indistinguishableAt optimal state, model performances are indistinguishable;‘working to characterize real contribution with other metricsSummary and Conclusions• Convection initiation by convergence is the main trigger for summertime precipitation in the Southern Great Plains. • Noah LSM does a competent job of representing the diurnal cycle of surface fluxes and states.• IHOP_2002 provides a wealth of data to evaluate land-surface parameterizations’ ability to simulate fluxes that affect boundary layer processes.• Photosynthesis-based evaporation schemes (dynamic phenology modules), subsurface hydrology (groundwater interaction) are in the forefront of LSM development.• Land-atmosphere coupling is strong in transition zones. • Calibration of effective parameters in complex models renders significant improvement.Summary and Conclusions (cont.)• Estimation of new model parameters only (“piecewise calibration”) is often insufficient. (Only GW benefits from piecewise calibration.)• Despite physical realism, parameters are still tunable coefficients: new default values needed.• When given optimum parameters Noah-STD, Noah-DV, Noah-GW, and Noah-DVGW perform equivalently well: model error remains the major component of the total uncertainty.• Added components alter host model structure; point of optimality shifts to accommodate new parameter
View Full Document
Unlocking...