Hydrology Modeling in Alaska: Variable Infiltration Capacity (VIC) Hydrologic Model Documentation Focus on the British Columbia VIC Implementation by Pacific Climate Impacts Consortium Your name: Katrina E. Bennett, Hydrologist, Pacific Climate Impacts Consortium, www.pacificclimate.org Model name: Variable Infiltration Capacity Model Authors: Xu Liang, Dennis Lettenmaier, Eric Wood Source code location (if public): Most recent release (October, 2009) is VIC 4.1.1. Source code is found here: http://www.hydro.washington.edu/Lettenmaier/Models/VIC/ Citations and URLs for basic documentation: Liang, X., Lettenmaier, D., Wood, E. and Burges, S. 1994. A simple hydrologically based model of land surface water and energy fluxes for general circulation models. http://www.icaen.uiowa.edu/~hmet/Handouts/94JD00483.pdf Accessed October 26, 2009 Liang, X., Wood, E.F., and Lettenmaier, D.P., Surface soil moisture parameterization of the VIC-2L model: Evaluation and modification, Global and Planetary Change, Volume 13, Issues 1-4, Soil Moisture Simulation, June 1996, Pages 195-206, ISSN 0921-8181, DOI: 10.1016/09218181(95)00046-1. http://www.sciencedirect.com/science/article/B6VF0-3VW7Y85 F/2/faa041863a40b5e3699b69cd74ca1323 Accessed October 26, 2009 Source code language: C Model type and/or conceptual framework: The VIC model is a physically-based, large-scale, semi-distributed hydrologic model. Data needed to run the model (inputs): Daily gridded fields of precipitation, minimum temperature, maximum temperature and wind speed. These are derived for BC using the methods outlined in Hamlet and Lettenmaier (2005), and Maurer et al. (2002). The 1 Symap algorithm (Shepard, 1984) regrids station data to grid cells based on weighted distances of station locations to grid cells and the angle of the station to the grid cell. The elevation-corrected PRISM climatology is then used to adjust the temperature and precipitation data for elevation effects. Parameters and how they are derived: Vegetation, soils and a digital elevation model (DEM). Data can be derived from any number of state-wide or global data sets (i.e. http://www.geog.umd.edu/landcover/1km-map.html). However, the format of these data sets is specific to VIC (scripts are available). Model setup, including parameterization format is found here: http://www.hydro.washington.edu/Lettenmaier/Models/VIC/Documentation /HowtorunVIC.shtml. Spatial element used to lump inputs and outputs: The VIC hydrologic model is run one grid cell at a time. Results from each grid cell are grouped, analysed and routed by watersheds. Sub-models (i.e. snow or ground thermodynamics): Two-layer algorithm, surface layer solves energy exchange atmosphere, lower layer storage to simulate deeper snowpack (Stork and Lettenmaier 1999). Canopy effects (e.g. attenuation of wind and solar radiation, melt water drip, increased sublimation from canopy) incorporated, snow “aging” effects also modelled. Recent model versions include spatially-distributed (laterally) snow coverage. Recent versions (4.1.1) of the model includes: a frozen soil algorithm also available (Cherkauer and Lettenmaier 2003), permafrost enhancements, excess ice and subsidence model, and “spatial frost”. See http://www.hydro.washington.edu/Lettenmaier/Models/VIC/Overview/Mode lOverview.shtml for more information. Rainfall/runoff transformation mechanism: Soil moisture: Soil moisture is governed by interception, infiltration, evapotranspiration from the top two layers, evaporation from bare soil, gravity-driven flow from upper layers to lower layers and the ARNO baseflow formulation for drainage from bottom layer. Runoff: Excess soil moisture can occur as either quick flow response, generated in the first thin layer (generally 10 mm), or as subsurface flow generated in the second and third soil layers. Groundwater: Drainage from the second to the third layer is gravity driven and based on the amount of soil saturation (Campbell, 1974). The deep (in most cases, the third) soil layer generates baseflow according to the ARNO formulation (Todini, 1996). Deep groundwater is not available at this time but a model is in development. Runoff routing within spatial elements and to basin outlet: 2 The routing model transports grid cell surface runoff and baseflow produced by the VIC model within each grid cell into the river system. A full description of the routing model can be found in Lohmann et al. (1998a; 1996; 1998b). The routing model implicitly assumes that the only mechanism by which water is transported out of a grid cell is by river flow. Construction of the routing network at 1/16th degree resolution is a time consuming initiative; consequently, the routing network is constructed on a case-by-case basis for the BC study area watersheds. The in-grid dynamics of surface routing are addressed for the grid cell instantaneous response function (or IRF or unit hydrograph), which is intended to capture the flow of water through the sub-grid surface runoff network to the grid cell ‘outlet’. River routing within the model uses the linearized shallow water or Saint-Venant equations, based on the model river network. The river network is constructed using the main flow direction of natural streams in each grid cell selected from one of eight possible directions. The routing distance within each grid cell is based on the flow direction and grid cell size. Due to the rather coarse discretization of the model domain using 1/16th degree cells, some grid cells (i.e. those along the periphery) may not be entirely contained within the basin being routed. As such, a flow fraction file is created which contains gridded information about the fraction of each grid cell (and, therefore, fraction of runoff) that flows into the basin. In order to simplify the routing model, grid cells were classified into the following routing features: channels/rivers or lakes/reservoirs with representative diffusion and velocity parameters for each. Routing is then run as a post-processing routine once the VIC model fluxes (runoff and baseflow) have been generated. Routed flow responses at the daily, monthly and annual time step are generated by the routing model for each sub-watershed. Method for including sub-grid scale processes: In the VIC model, topography is represented via the application of elevation bands. These elevation bands are used in the model to improve model performance in locations with steep terrain. This is important in mountainous regions, where the effects of elevation on snow pack accumulation and ablation might be lost in a large grid cell. Any number of different elevation bands can be defined in the VIC model, however typically five bands are used. The elevation band file also defines the fraction of precipitation that is falling as snow for each band based on the band elevation relative to the grid cell elevation. Temperature changes within each elevation band are calculated based on a standard lapse rate (6.5C per 1000 km elevation) multiplied by the difference between the grid cell elevation and the median elevation of the band. Resolution: The model has been implemented at 2 degrees to 1/16th of a degree (most recent Climate Impacts Group (CIG)-UW implementation, and the PCIC3 BC implementation). A 1/32nd degree model has been attempted by the model developer, Xu Liang in ‘research mode’, although at this scale the lack of slope, aspect, and ‘connectivity’ in the model may begin to cause issues. Method of deriving topography: DEMs can be downloaded from various sources. For the BC application we have used the CGIAR-CSI 90m Shuttle Radar Topography Mission data set (http://srtm.csi.cgiar.org/). Scripts are available to process the DEMs into the VIC elevation band format. Calibration approaches: Calibration of both the VIC model and the routing model must be completed for modelled and observed runoff to match closely. The VIC model simulates the fluxes such as runoff, evapotranspiration, snow melt, etc. for each grid cell and the routing model translates the runoff and baseflow fluxes through the surface drainage network. Calibration in the VIC model is achieved primarily by adjusting the unknown soils parameters that were not defined during the soils classification, namely infiltration and baseflow parameters that moderate the rate and volume of water that enters and exits the soil column. Soils parameters are calibrated automatically, which will be described below. Other parameters such as albedo may have to be calibrated manually when applying the VIC model in regions where default values are no longer valid, i.e. higher latitudes. Calibration of the routing model involves testing of the routing parameters and determining the correct velocity and diffusion factors to apply for lakes and reservoir systems, as noted above. Calibration of the combined VIC model and routing is conducted by comparing observed-to-simulated daily discharge during a select period of time. Therefore, the model domain is usually divided into sub-basins based on the location of hydrometric gauge locations that have observed streamflow data suitable for calibration. Stations are considered suitable if data was available for the entire calibration/validation period, and the drainage area of the basin is greater than 500 km2 and unregulated. Basins smaller than 500 km2 are likely to provide poor modelled results because the VIC is a macro-scale application intended for use in relatively large watersheds. In these large watersheds, effects of features not explicitly included in the model, such as aspect or slopes, will cancel out over the watershed. If regulated basins are selected, flow must be naturalized prior to calibration. The calibrated model can be run for an alternate time period to validate the parameter set established in calibration (referred to as a split-sample approach to model calibration/validation, Singh et al., 2002). 4 Treatment of frozen ground: Recent versions (4.1.1) of the model includes: a frozen soil algorithm also available (Cherkauer and Lettenmaier 2003), permafrost enhancements, excess ice and subsidence model, and “spatial frost”. See http://www.hydro.washington.edu/Lettenmaier/Models/VIC/Overview/Mode lOverview.shtml for more information. Publications using this model: Recent publications include: o Adam, J., Hamlet, A., and Lettenmaier, D. 2009. Implications of global climate change for snowmelt hydrology in the twenty-first century. Hydrol. Process. 23, 962–972 Published online 29 December 2008 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/hyp.7201 o Das, T., H.G. Hidalgo, M.D. Dettinger, D.R. Cayan, D.W. Pierce, C. Bonfils, T.P. Barnett, G. Bala, and A. Mirin, 2009: Structure and Detectability of Trends in Hydrological Measures over the Western United States. J. Hydrometeor., 10, 871–892. o Sheffield, J., K. M. Andreadis, E. F. Wood, and D. P. Lettenmaier, 2009: Global and continental drought in the second half of the 20th century: severity-area-duration analysis and temporal variability of large-scale events, J. Climate, 22(8), 1962-1981. There are a lot of VIC-related publications out there! Strengths and Weaknesses in Alaska applications: Strengths: o The VIC model can be run with limited input data. o The model is widely used and applied, at global scales and in northern regions. o Open source code. Model can be time consuming to set up, and implement. 5