Alfonso Senatore

Towards fully coupled atmosphere-hydrology model systems: long-range simulation in Southern Italy A. Senatore, University of Calabria

Towards fully coupled atmosphere-hydrology model systems: long-range simulation in Southern Italy with the WRF-Hydro modeling system

Alfonso Senatore

•  CeSMMA - Centro Studi per il Monitoraggio e la Modellazione Ambientale •  Dept. of Environmental and Chemical Engineering, University of Calabria

Padua, September 23, 2015


Objectives / Outline 1. Reliability of stand-‐alone WRF-‐Hydro hydrological model in a Mediterranean catchment (comparison of observed and simulated streamflow)

2. Parameterization and evaluation of stand-‐alone WRF mesoscale model, with particular reference to precipitation

3. Comparison of stand-‐alone WRF modeling system with fully coupled WRF/WRF-‐Hydro modeling system

� Evaluating potential of fully coupled modeling, both for hydrometeorological forecasts (short-‐medium range) and hydrological impacts due to climate change (long-‐range)


Scientific motivation


Atlantic weather fronts

~ 2000 mm ~ 500 mm

~ 35 km

Study area

Crotone

Soverato Vibo

Sybaris


5 hours

160 mm

Study area

Crotone

Soverato Vibo

Sybaris Corigliano/Rossano 12.08.2015

Platì 106 mm

Reggio C. province

09.09.2015

San Luca 115 mm

Sant’Agata 113 mm


“Crati @ S. Sofia” gauging station 1281 km2 250 m horizontal resolution Hmax = 1856 m Hmean = 672 m Hmin = 49 m

mean precipitation 1200 mm mean temperature 11.9 °C

Study area (stand-alone WRF-Hydro)

45 rain gauges (10) 35 thermometers (11) 11 radiometers (3) 12 hygrometers (5) 8 anemometers (2) 6 barometers (2)

Longwave radiation à GLDAS


� Previous studies: one-‐way coupling with RCMs

Study area

HIRHAM res. 0.11°

A2 scenario RegCM

res. 20 km A2 scenario

COSMO-‐CLM res. 0.165°

A1B scenario

T

HIRHAM +3.9 °C RegCM +3.9 °C CLM +3.5 °C

P HIRHAM -‐9% RegCM -‐20% CLM -‐12%

2070–2099 vs. 1961–1990

Groundwater storage

-‐11.6%

-‐6.5%

-‐10.7%

Senatore et al., JoH, 2011 – IPCC AR5

In-‐STRHyM hydrological model 1 km res. daily time step


WRF-Hydro Basic concepts •  Linking the column structure of land surface models with the ‘distributed’ structure of hydrological models

Credits: D

avid J. Goc

his, NCAR Surface routing

• Pixel-‐to-‐pixel routing •Steepest descent or 2d •Diffusive wave/backwater permitting •Explicit solution

•  Flexible coupling architecture designed to be extensible to new hydrological parameterizations


Subsurface routing • Quasi steady-‐state, Boussinesq flow model •Exfiltration from fully-‐saturated soil columns •Anisotropy in vertical and horizontal Ksat •No ‘perched’ flow •Soil depth is uniform • Critical initialization value: water table depth

WRF-Hydro

Credits: D

avid J. Goc

his, NCAR

Basic concepts •  Linking the column structure of land surface models with the ‘distributed’ structure of hydrological models



WRF-Hydro

Credits: D

avid J. Goc

his, NCAR

Channel routing • Solution Methods: –Gridded: 1-‐d diffusive wave: fully-‐unsteady, explicit, finite-‐difference –Reach: Muskingam, Muskingam-‐Cunge • A priori function of Strahler order • Trapezoidal channel (bottom width, side slope)

Basic concepts •  Linking the column structure of land surface models with the ‘distributed’ structure of hydrological models



Methodology

Stand-‐alone WRF-‐HYDRO with observed forcing

Oct,1s

t 2

002

Nov

,1st

Dec,1s

t

Feb,1st

Mar,1s

t

Apr,1s

t

May,1s

t

Jun,1st

Jul,1

st

Aug

,1st

Sep,1st

Jan,1st 200

3

Oct,1s

t 2

003

Jan, 6

WRF-‐only

Fully-‐coupled WRF-‐HYDRO

Oct,1st 2005

Oct,1st 2005

1.


Stand-alone WRF-Hydro Noah LSM

2.5 km res

à 250 m

à 2.5 km


� Calibration procedure

Stand-alone WRF-Hydro

www.pesthomepage.org

Minimization of the objective function Φ, given by the sum of squared deviations between model-‐generated observations and experimental observations, by means of the Gauss-‐Marquardt-‐Levenberg method (non-‐linear estimation technique)

Hourly streamflow � Experimental observations?


Stand-alone WRF-Hydro � Parameters involved in the calibration process

Several (dozens!) preliminary simulations in order to understand sensitivity of the model to single parameters and stepwise approach for the calibration of the parameters controlling water volume (1st step) and hydrograph shape (2nd step)

•  4 Manning roughness coefficients (CHANPARM.TBL) •  Bucket model exponent (GWBUCKPARM.TBL ) •  Slope coefficient modifying the drainage out the bottom of the last soil layer (GENPARM.TBL)

•  Noah surface runoff parameter retdt (GENPARM.TBL) •  Accompanying parameter refdk (corresponding to Ksat for silty clay loam) (GENPARM.TBL – SOILPARM.TBL)

•  Ksat for sandy loam (most diffused texture in the basin) (SOILPARM.TBL) •  depth of the bottom of the first soil layer (namelist.hrldas, hydro.namelist) •  gridded values of the overland flow roughness scaling factor (OVROUGHRTFAC)


� Calibration results Stand-alone WRF-Hydro

Calibration Validation

Calibration N.S. = 0.93 Validation N.S. = 0.72 Overall N.S. = 0.80

Issues: 1.  Recession curves 2.  GW bucket model 3.  Reservoir management rules

2.


Study area (WRF, WRF-Hydro) Large domain 12.5 km hor. res. (172 x 154 grid points)

Small domain 2.5 km hor. res. (95 x 90 grid points) One-‐way nesting

Era-‐Interim reanalysis


� WRF configurations

WRF parameterization

Acronyms M2P1 M2P2 M6P1 M6P2 M8P1 M8P2 M2P2C3 M6P2C3 Microphysics 2 2 6 6 8 8 2 6 PBL 1 2 1 2 1 2 2 2 Cum. param. 1 1 1 1 1 1 3 3

•  Microphysics: 2 -‐ Purdue Lin; 6 -‐ WSM6; 8 -‐ Thompson graupel •  PBL: 1 -‐ YSU scheme; 2 – MYJ •  Cumulus parameterization: 1 -‐ Kain-‐Fritsch; 3 -‐ Grell-‐Devenyi ensemble •  Rapid Radiative Transfer Model (RRTM) for longwave radiation •  Dudhia scheme for shortwave radiation •  Unified Noah Land-‐Surface Model

Senatore et al., JoHM, Dec 2014


WRF parameterization “Wet period” Nov 2008 – Jan 2009

Obs M2P1 M2P2 M6P1 M6P2 M8P1 M8P2 M2P2C3 M6P2C3 Mean (mm) 743.5 791.6 738.1 745.9 710.4 718.7 585.5 699.6 660.8 St.Dev. (mm) 257.8 256.1 243.9 239.7 239.6 246.6 211.2 257.0 258.3

precipitation


WRF parameterization “Dry period” Oct 2001 – Sep 2002

Obs M2P1 M2P2 M6P1 M6P2 M8P1 M8P2 Mean (mm) 832.7 972.8 859.0 975.2 855.2 799.8 687.1 St.Dev. (mm) 296.6 309.9 269.8 370.5 312.5 295.8 222.7 precipitation


WRF parameterization � Maps of simulated precipitation fields

Dry period Wet period


WRF parameterization �  Sensitivity to SST and coastal SST

Tracks of pressure minima from 8 Jan 2009, 12:00 to 9 Jan 2009, 21:00, with SST,

SST-‐0.5 °C and SST+0.5 °C Daily precipitation patterns on 9 Jan 2009

7 Nov 2014

3.

Senatore et al., Journal of Advances in Modeling Earth Systems -‐ AGU, 2015


Forcing data from WRF (P, T...)

WRF-Hydro (with NOAH LSM)

SH, LH from Hydro+LSM

SH, LH from WRF

Surface

+ Hydrological

model

SH LH

P

Results: preliminary considerations One-‐way coupling vs. Fully-‐coupling

Soil moisture evolution

SM

SM


� What should we expect?

Results: preliminary considerations

WRF vs. WRF-‐Hydro

Soil layers

Re-‐infiltration

more soil moisture

more runoff generation


�  Averaged precipitation maps (Nov 2002 – Sep 2005)

Results - precipitation WRF-‐only WRF-‐Hydro

2118 mm 2110 mm 3502 mm 3422 mm

WRF-‐only -‐ WRF-‐Hydro


� Performance indices of WRF-‐only and fully-‐coupled WRF/WRF-‐Hydro modeled precipitation fields

Results - precipitation

Bias lower with WRF-‐Hydro in 25

stations, with WRF in 13 stations

RMSE lower with WRF-‐Hydro in 22

stations, with WRF in 16 stations


�  Averaged precipitation evolution in the Crati catchment Results - precipitation

3319 3207

3434

Absolute differences with obs daily precipitation > 10 mm in: •  7.3% of cases with WRF •  6.9% of cases with WRF-‐Hydro


�  Days with absolute precipitation differences between WRF-‐only and fully-‐coupled WRF/WRF-‐Hydro higher than 10 mm

Results - precipitation

Day Observed (mm)

WRF (mm)

WRF-‐Hydro (mm)

Differences WRF -‐ WRF-‐Hydro

15/11/2004 24.6 29.1 45.5 -‐16.4 05/06/2004 5.9 14.0 27.4 -‐13.4 24/12/2003 4.8 38.5 23.9 14.6 26/01/2003 6.5 40.4 24.9 15.5 10/12/2002 5.5 17.4 0.5 16.9 14/11/2003 0.0 74.5 45.8 28.7 26/07/2004 4.8 100.2 27.9 72.3


�  Jul 26, 2004 Results - precipitation

WRF WRF-‐Hydro

Western station

Eastern station


�  Jul 26, 2004 Results - precipitation

IWV (kgm-‐2), sea level pressure (contours; hPa), and 10-‐m winds (barbs; ms-‐1) on 26 Jul 2004, 0900 UTC with a) WRF-‐only and b) fully-‐coupled WRF/WRF-‐Hydro simulations. c) Ts differences in the skin temperature for the two models (WRF-‐only minus fully-‐coupled WRF/WRF-‐Hydro) at the same date and time


Results �  Surface runoff (a) and deep drainage (b)

1068 mm

641 mm

893 mm

723 mm


�  Soil moisture Results

ZSOIL(1) 0-‐0.05 m

ZSOIL(4) 0.7-‐1.5 m


Results �  Soil moisture 1st layer -‐ Aug 31, 2003, 13:00

WRF-‐only WRF-‐Hydro

Shannon enthropy (as a measure of spatial variability): •  WRF = 0.70

•  WRF-‐Hydro = 0.77

Mean SMC = 0.126 Mean SMC = 0.143


Results �  Soil moisture 1st layer -‐ Feb 20, 2004, 13:00


Mean SMC = 0.265 Mean SMC = 0.265




Results � Heat fluxes

H

LE


Results � Heat fluxes

Dec 2002 Aug 2005

DJF MAM JJA SON

Ts WRF (K) 286.5 277.1 284.5 296.6 288.3 WRF-‐Hydro (K) 286.4 277.1 284.4 296.4 288.2

H WRF (W m-‐2) 49.6 -‐1.0 50.6 102.0 43.6 WRF-‐Hydro (W m-‐2) 46.7 -‐1.2 49.2 95.5 40.0

λE WRF (W m-‐2) 49.8 22.0 72.9 70.9 24.0 WRF-‐Hydro (W m-‐2) 53.6 23.3 74.8 78.6 28.4


Results �  LE -‐ Aug 31, 2003, 13:00





Results � Soil moisture and fluxes validation?

Bonis FLUXNET

station Soil moisture Sensible heat flux


Results �  Soil moisture feedback on precipitation


Results � Models vs. MODIS LST – Aug 31, 2003

WRF WRF-‐H

MODIS

Inland cells

WRF WRF-‐H MODIS µ 307.3 306.4 306.1 c.v. 0.0120 0.0124 0.0144 E 0.69 0.72 0.77


Results �  Streamflow

Total volumes at the outlet: 1426×106m3 1334×106m3 Observed WRF-‐Hydro

Ratio to precipitation: 0.36 0.34

Nash-‐Sutcliffe 0.27


Conclusions (1/2)

� Reliability of stand-‐alone WRF-‐Hydro hydrological model in a Mediterranean catchment

� WRF vs. fully-‐coupled WRF-‐Hydro à differences due to soil water lateral redistribution and re-‐infiltration: �  Largest differences for surface runoff and deep drainage �  Soil moisture and heat fluxes differences (summertime λE 8 Wm-‐2)

�  Slightly better performance for precipitation


Conclusions (2/2)

� An answer not only to the question “if”, but also “to what extent” does the fully coupled modeling alter mesoscale model performance in terms of land surface variables and precipitation, at least in the analyzed region

� More evident improvements for continental interior regions (no flat areas / not too humid climate)?

� Results achieved mainly relevant to long-‐range simulations (short-‐term under investigation…)


Acknowledgements: G. Mendicino (Unical), D.J. Gochis, Wei Yu, D.N. Yates (NCAR), H. Kunstmann, B. Fersch, T. Rummler (IMK-IFU)

Science

Alfonso Senatore