Colorado Headwaters Project


Figure 1: The WRF model domain and location of SNOTEL sites (black dots). (a) The full model domain. (b) Sub–domain focused on the 112 SNOTEL sites in the Colorado Headwaters region.

The Colorado Headwaters effort was initiated in the Spring of 2008 as a project within the RAL/Integrated Science Program Water System program. It is focused on assessing the impact of climate change on winter precipitation, snowpack and runoff processes in Colorado's headwater basins using a very high resolution fully coupled atmospheric–hydrologic model (WRF coupled with the NOAH land surface model). This work is collaborative with NCAR/MMM and researchers from the University of Colorado, the University of Washington, and the University of Texas.

The first phase of the Colorado Headwaters project applied the WRF model to produce high–resolution regional climate simulations of cold–season snowfall, snowpack, evapotranspiration and runoff in the Colorado Headwaters region. The domain of the high-resolution model is shown in Figure 1a, and the SNOwpack TELemetry (SNOTEL; ) observation sites used to evaluation the simulations in Figure 1b as the black dots. These initial simulations forced at the boundaries using the North American Regional Reanalysis (NARR) data, with model simulations conducted at grid spacings of 2 km, 6 km, 18 km and 36 km, over three 6–month periods over the winters of 2002–2003, 2004–2005, 2005–2006, and one 12-month period for the water year October 2007 – September 2008. Additional simulations were produced to assess the response of snowfall and snowpack to a pseudo climate warming scenario, in which the NARR boundary conditions were adjusted to match the future climate from the global-scale CCSM model under the IPCC SRES A1B scenario.

Analysis during the first phase of Headwaters activities focused on (i) model evaluation, (ii) model sensitivity studies on various WRF physics parameterizations, (iii) diagnostic analysis of the PGW simulations, (iv) comparison of statistical vs dynamical downscaling using the Colorado Headwaters simulations, and (v) development of an improved snowpack physics in the Noah Land Surface model. The evaluation of the WRF simulations over the four retrospective 6-month simulation of cold season precipitation showed that the model reproduced observed SNOTEL precipitation amounts within 10-15% of observations from 112 SNOTEL sites. The spatial pattern of precipitation from the high resolution model simulation also showed excellent agreement with the SNOTEL observations. Significantly, the snowfall amount from the coarser model resolution simulations (6, 18, and 36 km grid spacings) underestimated the observed season-total snowfall amount by 20-25 % due to terrain smoothing and associated weaker vertical motions. The strong dependence of simulated snowfall and snowpack on grid resolutions clearly indicates the importance and usefulness of high–resolution models in improving the future climate projections by global climate models. Model sensitivity tests on various WRF physics parameterizations, including seven microphysics parameterizations, three LSMs, five PBL parameterizations and two radiative transfer models, have shown that the model was most sensitive to microphysics parameterization, where the agreement with the SNOTEL observations was the best with the Thompson et al. 2008 and Morrison et al. 2009 schemes.

FY2011 Accomplishments


The second phase of the Colorado Headwaters project focuses on understanding the climate sensitivity of the land component of the hydrological cycle in the Headwaters region. The FY2011 activities focused on producing continuous simulations over the Colorado Headwaters domain (Figure 1), at grid spacing of 4-km, for the 8-year period October 2000 until September 2008. This longer continuous simulation period provides much more events for analysis, and the longer simulation period reduces the sensitivity of results to hydrologic initial conditions (the specification of hydrologic initial conditions at the start of the simulation period has a very minor impact on model results after approximately one year into the simulation). Moreover, the new Headwaters simulations use the Barlage et al. (2010) modifications to the Noah snow model. As with the first phase of the Headwaters project, additional simulations were produced to assess the response of snowfall and snowpack to a pseudo climate warming scenario, in which the NARR boundary conditions were adjusted to match the future climate from the global-scale CCSM model under the IPCC SRES A1B scenario.

8-year continuous simulations

Figure 2. Comparison WRF simulations and observations of accumulated precipitation, as averaged over the 112 SnoTEL sites within the Colorado Headwaters domain, for the 8-year continuous simulation period.
Colorado Headwaters
Figure 3. Detailed assessment of the WEF simulations of precipitation during the summer of 2002, showing (top plot) the precipitation time series, as averaged over all stations in the Colorado Headwaters domain; and (bottom plot) the precipitation sequence as expressed as precipitation accumulations. The error bars in the bottom plot depict the spatial variance in precipitation accumulation, as expressed as +/- one standard deviation.

The results presented in Figure 2 compare simulated precipitation from the continuous 8-year simulations against SNOTEL observations. Results show that while there is good agreement between model simulations and observations during winter, the simulations degrade during summer months with a tendency for the WRF simulations to over-estimate precipitation at the SNOTEL stations. The differences between WRF and SNOTEL estimates of end-of-year accumulated precipitation, as expressed as a percentage of the SNOTEL accumulation, range from -2.3% (in 2007-2008) to +15.4% (in 2001-2002). Figure 3 provides more detailed analysis of the precipitation sequence during the summer of 2002, showing precipitation as averaged over all SNOTEL stations and all lower-elevation climate stations. Consistent with the results in Figure 2, the WRF simulations are producing excessive precipitation during most storms. Research is ongoing to understand the cause of the summertime precipitation biases.

Climate sensitivity of the land component of the hydrological cycle

The first phase of the Colorado Headwaters project emphasized our limited understanding of how the climate sensitivity of the land component of the hydrological cycle depends on the spatial resolution and configuration of regional and global climate models. In 2011, focused research in hydrology mechanistically compares and contrasts 2 different scenarios of mid-21st century future hydroclimate; one with the global NCAR Community Climate System version 3 (CCSM3-AR4 generation) model, and one with the continuous, 8-year, very high resolution (4km) WRF regional climate model. The results show dramatic differences in several hydrooclimatic fields that are attributable to fundamental changes in the seasonality of precipitation and the elevational control of precipitation and evaporative demand. Differences in magnitude and sign of several variables are surprising given that the climate change scenario for the regional model is based on the CCSM climate change signal.

Specifically, the regional model exhibits a much stronger future increase in wintertime precipitation and cool season (Nov-May) runoff production than does CCSM. The increase in winter precipitation is driven by increased precipitation efficiency, which in turn is driven by an increase in condensate loading over orographic barriers under the warmer and wetter future climate conditions. Changes in wintertime precipitation frequency in both models are found to be minor by mid-century implying minimal changes in storm track by that time. The input of increased precipitation at high elevations in the regional model at a relatively cool time of year result in future increases in cool season and annual runoff production in the regional model compared to the CCSM. Future changes in summertime precipitation between the two models show that the global model continues to show weak to modest increases in precipitation but the regional model lacks such a signal over high-elevation terrain. Taking into account widespread increases in warm season evaporative demand, reduce summertime and early autumn runoff production in both models.

Hydrologic sensitivity in the 4-km WRF simulations 
Figure 4. Hydrologic sensitivity in the 4-km WRF simulations, showing (a) change in precipitation; (b) change in runoff; and (c) change in runoff fraction.

Figure 4 shows changes in annualized, basin-averaged water budget terms such as precipitation runoff and runoff efficiency between the current and future climates of the 4km WRF model runs. While notable differences in the magnitudes of the values exist between basins, the sign of the changes are relatively consistent showing increases in precipitation, decreases in runoff and runoff efficiency. This signal highlights the role of increased warming and evaporative demand on regional water budgets.


Journal papers resulting from the Colorado Headwaters project include:

  • Ikeda, K., R. Rasmussen, C. Liu, D. Gochis, D. Yates, F. Chen, M. Tewari, M. Barlage, J. Dudhia, K. Miller, K. Arsenault, V. Grubišić, G. Thompson, E. Guttman, 2010: Simulation of seasonal snowfall over Colorado. Atmos. Research, 97, 462-477.
  • Barlage. M., F. Chen, M. Tewari, K. Ikeda, D. Gochis, J. Dudhia, R. Rasmussen, B. Livneh, M. Ek, and K. Mitchell, 2010: Noah Land Surface model modifications to improve snowpack prediction in the Colorado Rocky Mountains. J. Geophys. Research – Atmospheres, 115, Article Number: D22101, DOI: 10.1029/2009JD013470
  • Rasmussen, R., C. Liu, K. Ikeda, D. Gochis, D. Yates, F. Chen, M. Tewari, M. Barlage, J. Dudhia, W. Yu, K. Miller, K. Arsenault, V. Grubišić, G. Thompson, E. Gutmann, 2010: High resolution coupled climate-runoff simulations of seasonal snowfall over Colorado: a process study of current and warmer climate. J. Climate, 24, 3015-3048, DOI: 10.1175/2010JCLI3985.1.
  • Gutmann, E. D., R. M. Rasmussen, C. Liu, K. Ikeda, J. Gochis, M. P. Clark, J. Dudhia, G. Thompson, 2010: A comparison of statistical and dynamical downscaling of winter precipitation over complex terrain. J. Climate, in press.

FY2012 Plans

The Colorado Headwaters team has just begun a collaborative project with the US Bureau of Reclamation and the US Army Corps of Engineers, to improve understanding of the climate sensitivity of the land component of the hydrologic cycle. Plans during 2012 include:

  • Produce WRF simulations over the Headwaters domain spatial resolutions of 50 km, 36-km, 12-km, for the 8-year simulation period for which 4-km simulations exist (2000-2008)
  • Evaluate the accuracy of dynamical downscaled output from the large-domain WRF simulations over the contiguous USA (36-km and 12-km) and the multi-model NARCCAP simulations (50-km) with respect to their representation of precipitation processes.
  • Produce non-dynamical downscaled simulations at multiple spatial resolutions across the western USA, with specific attention to the Colorado Headwaters domain
  • Produce multi-model hydrology simulations at multiple spatial scales across the western USA, and for specific basins in the Colorado Headwaters domain
  • Evaluate the fidelity of dynamical and non-dynamical downscaled results in representing regional climate features over the western USA and the Colorado headwaters domain, such as precipitation gradients across topographic barriers, intensification and decay of storms, atmospheric rivers, and precipitation associated with different weather types
  • Evaluate the fidelity of different hydrological models in representing hydrological processes, especially with respect to the choice of model physics
  • Attribute differences in dynamical and non-dynamical downscaled results to methodological aspects, including choice of model (NARCCAP), perturbation of model boundary conditions (large-domain WRF vs. Headwaters WRF), spatial resolution (Headwaters simulations), and details of the non-dynamical downscaling methodology
  • Evaluate the sensitivity of hydrological impacts to the downscaling resolution and methodology