Influence of spatial and temporal resolutions in hydrologic models

Transcription

1 Influence of spatial and temporal resolutions in hydrologic models Ingjerd Haddeland (University of Oslo) Dennis P. Lettenmaier (University of Washington) Thomas Skaugen (University of Oslo)

2 Outline Background, motivation Variable Infiltration Capacity (VIC) model Spatial aggregation: Rhone, Columbia and Arkansas-Red River basins Conclusions Temporal aggregation: Ohio and Arkansas-Red River basins Conclusions

3 Motivation and objective Hydrological data are just pieces... Pieces of the water balance Pieces in space Pieces in time

4 Motivation and objective Hydrological data are just pieces... Pieces of the water balance Pieces in space Pieces in time

5 Motivation and objective Representation of spatial variations in soil properties, topography and precipitation Spatial resolution of available input data and hydrologic models changes frequently Choice of spatial/temporal scale: Is often based on computational considerations, or issues related to the resolution of the observations. a1) a2) a3) b1) b2) b3) Elevation (m) a1) a2) a3) b1) b2) b3) Mean annual precipitation ()

6 Science questions How are model simulations impacted by changing the spatial resolution? How different does models evaluated at two temporal scales perform? Is it possible to reconcile simulations performed at different scales?

7 Previous studies: Examples Holmann-Dodds et al., Journal of Geophysical Research, 1999 Koren et al., Water Resources Research, 1999 Skaugen, Journal of Hydrology, 1997

8 VIC: Variable Infiltration Capacity Model N soil layers (3) N vegetation types (1) N elevation bands (1) Energy balance winter/suer Variable infiltration Nonlinear baseflow Distributed precipitation Typical scale of application: 1/8-2 degrees latitude by longitude, 1 hr to 24 hr temporal resolution

9 VIC: Variable Infiltration Capacity Model N soil layers (3) N vegetation types (1) N elevation bands (1) Energy balance winter/suer Variable infiltration Nonlinear baseflow Distributed precipitation Typical scale of application: 1/8-2 degrees latitude by longitude, 1 hr to 24 hr temporal resolution

10 Spatial aggregation: Study areas b) a) 12 W 1 W 8 W 5 N 5 N COLUMBIA The Dalles 4 N 4 N ARKANSAS - RED Elevation (m) 4 c) Little Rock Shreveport 3 N 3 N 1 km 12 W 1 W 8 W Mean annual precipitation ()

11 Aggregation method Rhone: 8*8 km One-half and one degree Time series Meteorological data (precip, temp, wind) Vegetation types Columbia and Arkansas-Red: One-eighth degree (~12.5 km*12.5 km) One-quarter, one-half, one and two degrees Static data Soil properties Elevation bands Flow direction

12 Results: RhoneAGG

13 Results: RhoneAGG

14 Results: Columbia and Arkansas-Red Spatially and temporally uniform precipitation, daily time steps Percent a1) -25 Columbia Grid resolution (degrees) Precipitation Direct runoff a2) -25 Arkansas-Red Grid resolution (degrees) Baseflow Total runoff Columbia J F M A M J J A S O N D Streamflow (1 m 3 /s)15 b1) One-eighth One-quarter b2) Arkansas-Red J F M A M J J A S O N D One-half One Two a) Percent changes in moisture fluxes, compared to the results at one-eighth degree spatial resolution, and b) Mean monthly streamflow at all spatial resolutions, for the 1) Columbia and 2) Arkansas-Red River basins, using spatially constant grid cell precipitation

15 Results: Arkansas-Red Scale sensitivity of total runoff for Arkansas-Red River basin as a function of water year precipitation -5-1 Percent TotalQ-Percent change One-quarter One-half One -3 Two Precipitation(/year)

16 Results: Arkansas-Red Effect of parameterization for spatial variability of precipitation (black) vs spatially uniform precipitation (open) Percent a) -25 Arkansas-Red Grid resolution (degrees) Precipitation Direct runoff Baseflow Total runoff Arkansas-Red J F M A M J J A S O N D Streamflow (1 m 3 /s)4 b) One-eighth One-quarter One-half One Two

17 Results: Columbia Effect of elevation bands (open symbols) vs no elevation bands (black symbols) Percent 5 a) Columbia Grid resolution (degrees) Precipitation Direct runoff Baseflow Total runoff 15 b) Columbia J F M A M J J A S O N D Streamflow (1 m 3 /s)15 One-eighth One-quarter One-half One Two

18 Results: Columbia Effect of elevation bands (open symbols) vs no elevation bands (black symbols) Effect of parameterization of precipitation as a function of elevation Percent 5 a) Columbia Grid resolution (degrees) Percent 5 a) Precipitation Direct runoff Baseflow Total runoff -25 Columbia Grid resolution (degrees) Precipitation Direct runoff Baseflow Total runoff 15 b) Columbia J F M A M J J A S O N D Streamflow (1 m 3 /s)15 15 b) One-eighth One-quarter One-half One Two 3 Columbia J F M A M J J A S O N D Streamflow (1 m 3 /s)15 One-eighth One-quarter One-half One Two

19 Conclusions: Spatial aggregation In general: Form of hydrographs preserved, runoff decreases as spatial scale increases Snowmelt dominated areas: Interaction precipitation/temperature Elevation bands Drier areas: Interaction precipitation/vegetation Subgrid precipitation and soil moisture, canopy evaporation Wet areas: Decrease in direct runoff is compensated by an increase in baseflow

20 Temporal scale

21 Temporal scale: Background 12 W 1 W 8 W The backdrop: Models evaluated at one temporal scale (time step) may perform much differently at another 5 N 4 N 3 N 15 3 Elevation (m) 12 W Canada USA 5 km Arkansas-Red 1 W Ohio 8 W 5 N 4 N 3 N 12 Arkansas-Red 12 Ohio WB Mean annual runoff in the Arkansas-Red and Ohio River basins. Daily water balance mode (24.WB) and 3 hourly energy balance mode ()

22 Temporal scale effects Sub-daily (/year) Sub-daily (/year) Runoff 2 3.WB Daily (/year) Transpiration 2 3.WB Daily (/year) Sub-daily (/year) Sub-daily (/year) 1 8 Evapotranspiration WB Daily (/year) Canopy evaporation Surface temperature WB Daily (/year) Sub-daily (C) WB Daily (C) Sub-daily (W/m 2 ) Net radiation 3.WB Daily (W/m 2 ) Spatially and temporally uniform precipitation. Daily water balance (24.WB) runs compared to 3 hourly water balance (3.WB) and 3 hourly energy balance () runs.

23 Model differences: 24.WB Energy balance, water balance Surface temperature Net radiation 5 3.WB Daily (C) Parameterization of canopy evaporation Daily time steps: Evaporation can include current time step s precipitation Sub-daily time steps: Evaporation cannot include current time step s precipitation Sub-daily (C) Surface temperature Sub-daily (W/m 2 ) Net radiation 3.WB Daily (W/m 2 ) So how can we easily reconcile model simulations?

24 Rescaling parameters for time step differences Search for parameters (interception capacity factor and minimum stomatal resistance): ( R ) ( ) new Rorig k1, n min months = ( C new ) n, i = ( Corig ) * k n, i 2, n 2 2 {( EC EC ) + ( TV TV ) } n, i * n, i sub daily 24. WB sub daily 24. WB SCEM-UA algorithm (Vrugt et al., Water Resources Research, 23) Search done across transect at one degree interval, evaluation at 1/8 degree (parameters interpolated for intermediate grid cells) Reproduce daily water balance results from 3 hr energy balance runs

25 Results: Transects 3 hourly energy balance compared to 24 hourly water balance runs Sub-daily (/year) Runoff Daily (/year) 1 Transpiration Daily (/year) 1 Evapotranspiration 12 1 a) Runoff c) Canopy evaporation Daily (/year) 12 2 a) Runoff 1 9 Canopy evaporation c) Canopy evaporation Daily (/year) 24.WB b) Evapotranspiration 1 d) Transpiration 2 b) Evapotranspiration 2 d) Transpiration

26 12 W 1 W 8 W 5 N Canada USA 5 N Results: Transects 4 N 3 N 5 km Precipitation ( year -1 ) N 3 N 12 W 1 W 8 W 3 hourly energy balance runs matched to 24 hourly water balance runs Sub-daily (/year) Sub-daily (/year) Runoff 2.k Daily (/year) Transpiration 2.k Daily (/year) Sub-daily (/year) Sub-daily (/year) 1 Evapotranspiration k Daily (/year) 1 Canopy evaporation k Daily (/year) 12 1 a) Runoff c) Canopy evaporation a) Runoff c) Canopy evaporation b) Evapotranspiration 1 d) Transpiration 2 b) Evapotranspiration 2 d) Transpiration WB.k

27 Results: River basins 5 N 12 W Canada USA 1 W 8 W 5 N 12 A Runoff 12 O Runoff 4 N Ohio 4 N N 5 km Arkansas-Red 3 N Elevation (m) 12 W 1 W 8 W A Evapotranspiration 15 O Evapotranspiration 1 1 A: Arkansas-Red O: Ohio A Soil moisture O Soil moisture 3 hourly energy balance runs and 24 hourly water balance runs WB

28 Results: River basins 5 N 12 W Canada USA 1 W 8 W 5 N A Runoff O Runoff 4 N 3 N 5 km Arkansas-Red 15 3 Elevation (m) 12 W 1 W Ohio 8 W 4 N 3 N A Evapotranspiration 15 O Evapotranspiration 1 1 A: Arkansas-Red O: Ohio A Soil moisture O Soil moisture 3 hourly energy balance runs matched to 24 hourly water balance runs WB.k

29 Results: Spatial images a) Runoff Original results Corrected results 3 hourly energy balance runs compared to 24 hourly water balance runs /24.WB b) Evapotranspiration Original results Corrected results /24.WB

30 Results: Spatial images a) Runoff Original results Corrected results 3 hourly energy balance runs matched to 24 hourly water balance runs /24.WB b) Evapotranspiration Original results Corrected results /24.WB

31 NLDAS (North American Data Assimilation System) Relative runoff bias WY , evaluated at USGS gauges with minimal management effects. Lohmann et al., 24: Streamflow and water balance Maurer et al., 22: A long-term hydrologically-based intercomparisons of four land surface models in the North data set of land surface fluxes and states for the American Land Data Assimilation System project, J. conterminous United States, J. Climate, 15, Geophys. Res., 19, D7S91, doi:1.129/23jd3517

32 NLDAS: Arkansas-Red 1 Evapotranspiration 8 1.EB (/year) 1.EB (/year) Arkansas-Red results Lohmann et al. (1.EB), vs Maurer et al. () 1 Runoff EB 1.EB (/year) 1 Transpiration Canopy evaporation 8 1.EB (/year) 1.EB: 1 hr energy balance runs (spatially and temporally disaggregated precipitation) 1 1.EB (/year) : 3 hr energy balance results (no spatial or temporal disaggregation of precipitation) (/year) EB 1.EB (/year) (/year)

33 Results: NLDAS 12 9 A a) Runoff 12 9 O a) Runoff 12 W 1 W 8 W N Canada USA 5 N 15 A b) Evapotranspiration 15 O b) Evapotranspiration 4 N 3 N 5 km Arkansas-Red 15 3 Elevation (m) 12 W 1 W Ohio 8 W 4 N 3 N A c) Canopy evaporation O c) Canopy evaporation 1 hourly energy balance runs, spatially and temporally disaggregated precipitation, matched to 3 hourly energy balance runs, temporally and spatially uniform precipitation A d) Transpiration A e) Soil moisture O d) Transpiration 25 5 O e) Soil moisture EB 1.EB.k

34 Results: NLDAS Evapotranspiration (/year) Arkansas-Red Ohio 1.EB Disaggregation method Uniform Temporal Spatial Temporal and spatial Runoff (/year)

35 Conclusions temporal aggregation Moisture fluxes simulated by the VIC model are sensitive to the time step used, to the assumptions made regarding closure of the surface energy budget, and to the method of temporal and spatial disaggregation of precipitation. Simulated canopy evaporation differences are the main reason for the discrepancies between simulated model results. Sensitivity analyses performed at sub-daily time steps (3 hours and 1 hour) indicate that temporal disaggregation of precipitation is the most significant factor controlling canopy evaporation at sub-daily time steps. Simulation results at different model setups can to a large extent be reconciled by introducing correction factors that adjust the canopy interception capacity and canopy resistance. It is possible to calibrate the model in the computationally efficient daily water balance mode and thereafter introduce correction factors to the sub-daily energy balance simulations without having to recalibrate the model.

36 References Boone, A., F. Habets, J. Noilhan, E. Blyth, D. Clark, P. Dirmeyer, S. Fox, Y. Gusev, I. Haddeland, R. Koster, D. Lohmann, S. Mahanama, K. Mitchell, O. Nasanova, G.-Y. Niu, A. Pitman, J. Polcher, A.B. Shmakin, K. Tanaka, B. van den Hurk, S. Verant, D. Verseghy, P. Viterbo, and Z.-L. Yang, 24, The Rhone-Aggregation Land Surface Scheme Intercomparison Project: An Overview, Journal of Climate 17(1), Haddeland, I., B.V. Matheussen, and D.P. Lettenmaier, 22, Influence of spatial resolution in a macroscale hydrologic model, Water Resources Research, 38(7), doi:1.129/21wr854 Haddeland, I., D.P. Lettenmaier, and T. Skaugen, 26, Reconciling simulated moisture fluxes resulting from alternate hydrologic model time steps and energy balance closure assumptions, Journal of Hydrometeorology (in press)