Noah-MP: A New Paradigm

Transcription

1 Noah-MP: A New Paradigm for Land Surface Modeling Zong-Liang YANG, Fei Chen, Mike Barlage, Mike Ek, Guo-Yue Niu, Xitian Cai, Rongqian Yang, et al. Presentation at the Workshop on Land Surface Modeling in Support of NWP and Sub-Seasonal Climate Prediction, George Mason University, 5-6 December,

2 How Can We Use Sophisticated Evaluation Methods To Guide LSM Development? Two schools of thought in LSM development and evaluation LSM developers consider 1. Increasing realism in representing key processes 2. Understanding feedbacks and interactions 3. Maintaining synergism between LSM and other modules in the host GCM 4. Aiming for past, present, and future climate applications & operational weather/climate predictions 5. Generalizing parameterzations across sites LSM developers do not use automated, sophisticated evaluation tools. Model Evaluation Pyramid Atmospheric Forcing Land Surface Model (CLM, Noah, Vic, ) Model Structure Augments (gw, dv, ) LSM evaluators consider 1. Uncertainty in many subsurface parameters and other nonmeasurable parameters 2. Uncertainty in atmospheric forcing and observations used for evaluation 3. Calibration of the parameters for the augmented part only or for the entire LSM 4. Evaluation in all dimensions 5. Equifinality? LSM evaluators calibrate/evaluate LSMs that already exist. 2

3 An Effort to Reconcile Both 1) Gulden, L. E., E. Rosero, Z.-L. Yang, M. Rodell, C.S. Jackson, G.-Y. Niu, P. J.-F. Yeh, and J. Famiglietti, 2007: Improving land-surface model hydrology: Is an explicit aquifer model better than a deeper soil profile? Geophys. Res. Lett., 34, L09402, doi: /2007gl ) Gulden, L.E. et al., 2008: Model performance, model robustness, and model fitness scores: A new method for identifying good land-surface models, Geophys. Res. Lett., 35, L11404, doi: /2008gl ) Jiang, X., G. Niu, and Z.-L. Yang, 2009, Impacts of vegetation and groundwater dynamics on warm season precipitation over the Central United States, J. Geophys. Res., 114, D06109, doi: /2008jd ) Rosero, E., Z.-L. Yang, L. E. Gulden, G.-Y. Niu, and D. J. Gochis, 2009: Evaluating enhanced hydrological representations in Noah-LSM over transition zones: Implications for model development, J. Hydrometeorology, 10, DOI: /2009JHM ) Rosero, E., Z.-L. Yang, T. Wagener, L. E. Gulden, S. Yatheendradas, and G.-Y. Niu, 2010: Quantifying parameter sensitivity, interaction and transferability in hydrologically enhanced versions of Noah-LSM over transition zones, J. Geophys. Res., 115, D03106, doi: /2009jd

4 2010 NOAA/NCEP Land Modeling Workshop at Austin, Texas 4

5 Noah-MP 5

6 Generalized Land Surface Modeling Exchange processes with the atmosphere o o o o Momentum Energy (reflected shortwave, emitted longwave, latent/sensible heat) Water (precipitation, evapotranspiration) Trace gases (CO 2, CH 4, N 2 O, BVOCs)/dusts/aerosols/pollutants Exchange processes with the ocean o o o Fresh water Sediments/nutrients Salinity Land-memory processes o Vegetation phenology o Snow/ice cover o Soil moisture o Groundwater Intraseasonal to Interannual Variability Human activities o o o o Land use (agriculture, afforestation, deforestation, urbanization, ) Water use (irrigation, human withdraws, dams, ) Air pollution / Water pollution Environmental degradation 6

7 Status Quo in Land Surface Modeling A land surface model represents, mathematically and numerically, the land surface s characteristics (e.g., land cover parameters), states (e.g., soil moisture, temperature, snow, leaf area), fluxes (e.g., runoff, evapotranspiration, outgoing longwave radiation, photosynthesis), as a function of space and time. There is a rich body of literature in modeling each of the biogeophysical and biogeochemical processes. How to choose parameterizations (parts) and assemble them into an LSM (whole) depends on the developer s experience, judgment, and understanding. Examples: BATS, SiB, SSiB, VIC, Noah, CLM, JULES, CABLE, 7

8 Science Questions Is there a synergy among the different parameterizations? Is there an optimal combination of these parameterizations? Do existing parameterizations possess typical signatures or exhibit typical patterns in simulating energy, water, and carbon balances? Is there a modular framework that o Allows for freely assembling of these parameterizations, o Diagnose differences, o Identifies structural errors, o Improve understanding, o Enhances data/model fusion and data assimilation, o Facilitate ensemble forecasts and uncertainty quantification? 8

9 What is Noah-MP? Augmented Noah LSM with Multi- Parameterization options (Noah-MP): o Key references: (Niu et al., JGR, 2011; Yang et al., JGR, 2011) o Recoded based on the standard Noah LSM o Well documented and highly modular o Improved biophysical realism (land memory processes): separate vegetation canopy and ground temperatures; a multi-layer snowpack; an unconfined aquifer model for groundwater dynamics; an interactive vegetation canopy layer 9

10 Interactive Vegetation Canopy The model includes a set of carbon mass (g C/m 2 ) balance equations for: 1. Leaf mass 2. Stem mass 3. Wood mass 4. Root mass 5. Soil carbon pool (fast) 6. Soil carbon pool (slow) Processes include: 1. Photosynthesis (S, T, θ, e air, CO 2,O 2, N ) 2. Carbon allocation to carbon pools 3. Respiration of each carbon pool (T v,θ, T root ) M t leaf Carbon gain rate: Carbon loss rate: LAI = M leaf * C area R gain R loss Dickinson et al. (1998), photosythesis * fraction of carbon partition to leaf Yang and Niu (2003) leaf turnover (proportional to leaf mass) respiration: maintenance & growth (proportional to leaf mass) death: temperature & soil moisture 10 where C area is area per leaf mass (m 2 /g).

11 Comparison of Noah-MP and Satellite LAI and % Vegetation Cover (GVF) 11 Yang et al. (2011)

12 Noah-MP water storage change compares well with GRACE Niu et al. (2011) Yang et al. (2011) Improved biophysical realism: a multi-layer snowpack an unconfined aquifer model for groundwater dynamics an interactive vegetation canopy layer 12

13 Modeled and Obs SWE at Niwot Ridge ( ) Modeled and Obs SWE at GLEES ( ) Noah VIC LEAF SAST Noah-MP CLM * All models melt snow too fast * Larger differences among models * SAST and LEAF largest melt rate * CLM and Noah-MP closest to obs * Noah, LEAF, VIC: lowest SWE during accumulation

14 Noah-MP is unique among LSMs A new paradigm in land-surface, environmental, and hydrological modeling (Clark et al., 2007; 2011) In a broad sense, o Multi-parameterization Multi-physics Multihypothesis A modular & powerful framework for o Diagnosing differences o Identifying structural errors o Improving understanding o Enhancing data/model fusion and data assimilation o Facilitating ensemble forecasts and uncertainty quantification 14

15 Noah-MP 1. Leaf area index (prescribed; predicted) 2. Turbulent transfer (Noah; NCAR LSM) 3. Soil moisture stress factor for transpiration (Noah; SSiB; CLM) 4. Canopy stomatal resistance (Jarvis; Ball-Berry) 5. Snow surface albedo (BATS; CLASS) 6. Frozen soil permeability (Noah; Niu and Yang, 2006) 7. Supercooled liquid water (Noah; Niu and Yang, 2006) 8. Radiation transfer: Modified two-stream: Gap = F (3D structure; solar zenith angle;...) 1-GVF Two-stream applied to the entire grid cell: Gap = 0 Two-stream applied to fractional vegetated area: Gap = 1-GVF 9. Partitioning of precipitation to snowfall and rainfall (CLM; Noah) 10. Runoff and groundwater: TOPMODEL with groundwater TOPMODEL with an equilibrium water table (Chen&Kumar,2001) Original Noah scheme BATS surface runoff and free drainage More to be added Niu et al. (2011) Collaborators: Yang, Niu (UT), Chen (NCAR), Ek/Mitchell (NCEP/NOAA), and others 15

16 Maximum # of Combinations 1. Leaf area index (prescribed; predicted) 2 2. Turbulent transfer (Noah; NCAR LSM) 2 3. Soil moisture stress factor for transp. (Noah; SSiB; CLM) 3 4. Canopy stomatal resistance (Jarvis; Ball-Berry) 2 5. Snow surface albedo (BATS; CLASS) 2 6. Frozen soil permeability (Noah; Niu and Yang, 2006) 2 7. Supercooled liquid water (Noah; Niu and Yang, 2006) 2 8. Radiation transfer: 3 Modified two-stream: Gap = F (3D structure; solar zenith angle;...) 1-GVF Two-stream applied to the entire grid cell: Gap = 0 Two-stream applied to fractional vegetated area: Gap = 1-GVF 9. Partitioning of precipitation to snow- and rainfall (CLM; Noah) Runoff and groundwater: 4 TOPMODEL with groundwater TOPMODEL with an equilibrium water table (Chen&Kumar,2001) Original Noah scheme BATS surface runoff and free drainage 2x2x3x2x2x2x2x3x2x4 = 4608 combinations Process understanding, probabilistic forecasting, quantifying uncertainties 16

17 Global 36 Ensemble Experiments GLDAS forcing, global 1 1 resolution 17 Yang et al. (2011)

18 Global 36 Ensemble Experiments Runoff options as the dominant player in forming clusters: SIMTOP (bottom sealed) produces the wettest soil and greatest ET BATS (greatest surface runoff) produces the driest soil and smallest ET Noah lies between SIMTOP and BATS SIMGM results in best soil moisture simulations 18 Yang et al. (2011)

19 Regional High Resolution Experiments Static Data Forcing Data Validation Data Lat-Lon mask, land mask, soil type, soil color, land use, greenness vegetation fraction (GVF) NLDAS2 forcing: precipitation, temperature, specific humidity, air pressure, downward longwave and shortwave radiation, wind USGS streamflow and groundwater, GRACE water storage change, CMC snow, MODIS LAI, Mississippi River Basin 6 HUC2 regions ~3.3 million km ~22,378 grid cells 19

20 USGS Gage Stations 20

21 Regional 14 Ensemble Experiments 21

22 Basin Mean Monthly Runoff ( ) For the entire Mississippi Basin All EXP1-14 capture seasonal and interannual variations. BATS and Noah runoff schemes produce largest runoff amplitudes. TOPMODEL and SIMGM runoff schemes show smallest amplitudes, in best agreement with observations. 22

23 Sub-Basin Mean Monthly Runoff For Missouri Sub-basin, TOPMODEL and SIMGM runoff schemes produce best simulations. For Ohio-Tennessee Sub-basin, BATS and Noah runoff schemes produce best simulations. 23

24 Statistics of Monthly Runoff ( ) For the entire Mississippi Basin All EXP1-14 show ~90% correlation BATS and Noah runoff schemes produce largest runoff amplitudes TOPMODEL and SIMGM runoff schemes shows smallest amplitudes 24

25 Statistics of Monthly Runoff ( ) Different subbasins have different best performing combinations. Runoff options dominate over other options. 25

26 Mississippi River Basin ( ) Runoff options as the dominant player in forming clusters: SIMTOP produces wetter soil and greater ET BATS produces drier soil and smaller ET Noah performs in a way similar to BATS SIMGM simulations produce wetter soil and greater ET 26

27 GPP Soil Water in Sub-Basins Missouri Ohio-Tennessee Runoff options as the dominant player in forming clusters: SIMTOP produces wetter soil and greater GPP BATS produces drier soil and smaller GPP Noah performs in a way similar to BATS SIMGM simulations produce wetter soil and greater GPP 27

28 Taylor Plots for Different Options (36 experiments) DV: on is more realistic than off; Ball-Berry exhibits more variability than 28 Jarvis; β shows least difference; Runoff options are the dominant options influencing energy, water, and carbon balances.

29 Coupled WRF/Noah-MP Seasonal Pairs of six-month 30-km simulations starting Feb and 2010 Spin-up soil for one year using offline HRLDAS IC/BC from NARR CAM radiation; YSU; Thompson Climate Simulations Barlage et al. (2013)

30 Transpiration Efficiency Default Noah LSM LSM Setup Noah-MP with OPT_RUN=3: free drainage comparable to Noah runoff scheme Noah-MP with OPT_RUN=5: Miguez-Macho & Fan groundwater with equilibrium water table Noah-MP with OPT_BTR=1(Noah) and 2 (CLM) Soil Saturation Noah CLM Barlage et al. (2013)

31 Miguez-Macho & Fan water table dynamics in NOAH-MP Equations: Mass balance in groundwater storage: Darcy s Law for groundwater river exchange: Darcy s Law for lateral groundwater flow: Q dsg x yr 8 Qn Q dt r 1 Q r = rc wtd - riverbed n w wtd Kn dz wtd K dz n ( ) conductivit y 2 hn h s Cell ij R Qr Water table depth (wtd) width of flow cross section Transmisivity Head difference divided by distance (water table slope) h i,j S g Q 8 Q 4 Mean sea level Cross section view Plan view Q 8 Q 1 Q 2 Q 3 i, j Q 7 Q6 Q 4 Q 5 i,j w Fan et al, JGR 2007 Miguez-Macho et al., JGR 2007

32 Regional Groundwater Recharge: 2002 Noah Noah-MP R3 Noah-MP R5 Barlage et al. (2013)

33 Regional Deep Soil Moisture: 2002 Noah Noah-MP R3 Noah-MP R5 Barlage et al. (2013)

34 Regional Root Soil Moisture: 2002 Noah Noah-MP R3 Noah-MP R5 Barlage et al. (2013)

35 Regional Latent Heat Flux: 2002 Noah-MP R3 Noah-MP R5 Barlage et al. (2013)

36 Noah-MP Reduces Seasonal 2-m Temperature Bias in WRF 36 Barlage et al. (2013)

37 Noah-MP Improves Seasonal Temperature Forecasts in CFS 37 Rongqian Yang et al. (2013)

38 WRF Simulated & Observed Monthly and Seasonal Mean Precipitation in Central Great Plains WRF with interactive canopy (DV) improves summertime rainfall in the central Great Plains. DV + dynamic water table = DVGW improving the simulation even more. Reason: improved coupling between soil moisture and precipitation through lowered lifting condensation level (see next slide). Incorporating vegetation and groundwater dynamics into a regional climate model would be beneficial for seasonal precipitation forecast in the transition zones. Jiang et al. (2009) J. Geophys. Res. 38

39 Lifting condensation level (LCL) height versus soil moisture index (SMI) in the soil layers Jiang et al. (2009) J. Geophys. Res. 39

40 Summary Improved realism in snowpack, groundwater, and vegetation phenology improves performance. The multi-parameterization (MP) framework allows for multi-hypothesis testing and understanding of parameterization interaction. Runoff parameterizations are the dominant options influencing energy, water, and carbon balances. Different sub-basins have different best performing combinations. The groundwater-climate interaction improves seasonal prediction of 2-m temperature and precipitation. 40

41 Summary (continued) More studies, both offline and coupled, covering multiscales (meters to tens of km and daily through seasonal to interannual), are warranted to realize the full potential of the MP framework. 41

42 Path Forward: Terrestrial Hydrological Model Intercomparison Testbed: Multi- Data Cal/Val Data Assimilation Benchmarking Scaling Cloud Modeling LSMs High-performance Computing Everything Coupler SSMs CLM JULES Noah-MP SiB VIC RTMs CATHY ParFlow OGS PAWS PIHM 42

43 Thank you! KAUST TACC NSF NASA NOAA 43