USGS -- SMIG --
Surface-water quality and flow Modeling Interest Group

General-Circulation-Model Simulations of Future Snowpack in the Western United States

by Gregory J. McCabe1 and David M. Wolock2

1 U.S. Geological Survey, Denver, CO
2 U.S. Geological Survey, Lawrence, KS

Please direct correspondence to:
  Gregory J. McCabe
  U.S. Geological Survey
  Box 25046, MS 412
  Denver Federal Center
  Denver, CO 80225-0046
  Internet: gmccabe@usgs.gov
  Phone: (303) 236-7278
  FAX: (303) 236-5034

Editor's note:
This article was published recently in the Journal of the American Water Resources Association. The document available here is based on the final draft provided to the journal. Minor discrepancies between this document and the published version, therefore, may exist.

This version of the article has all of the figures converted to thumbnails with links to the larger images. A version with all of the figures inline is also available; the download time may be longer, but the inline figures may be more convenient for viewing and printing.

Citation:
McCabe, G.J. and D.M. Wolock, 1999, General-circulation-model simulations of future snowpack in the western United States, J. American Water Resources Assn., 35(6), 1473-1484.

Contents

Abstract

April 1 snowpack accumulations measured at 311 snow courses in the western United States (US) are grouped using a correlation-based cluster analysis. A conceptual snow accumulation and melt model and monthly temperature and precipitation for each cluster are used to estimate cluster-average April 1 snowpack. The conceptual snow model is subsequently used to estimate future snowpack by using changes in monthly temperature and precipitation simulated by the Canadian Centre for Climate Modeling and Analysis (CCC) and the Hadley Centre for Climate Prediction and Research (HADLEY) general circulation models (GCMs). Results for the CCC model indicate that although winter precipitation is estimated to increase in the future, increases in temperatures will result in large decreases in April 1 snowpack for the entire western US. Results for the HADLEY model also indicate large decreases in April 1 snowpack for most of the western US, but the decreases are not as severe as those estimated using the CCC simulations. Although snowpack conditions are estimated to decrease for most areas of the western US, both GCMs estimate a general increase in winter precipitation toward the latter half of the next century. Thus, water quantity may be increased in the western US; however, the timing of runoff will be altered because precipitation will more frequently occur as rain rather than as snow.

Introduction

Snowpack accumulations are an important source of runoff and water supply in the western United States (World Meteorological Organization, 1970; Gray and Male, 1981). Scientists have estimated that increasing concentrations of atmospheric carbon dioxide may cause global warming and changes in temporal and spatial distributions of precipitation (Gammon et al., 1985; Bolin, 1986; Lins et al., 1988). There is concern that global warming may adversely affect snowpack accumulations in the western United States and therefore have a negative effect on water supply (Gleick, 1987; Lettenmaier and Sheer, 1991; McCabe and Legates, 1995).

The objectives of this study are to

  1. identify and quantify the relations between observed snowpack and winter temperature and precipitation in the western US,

  2. apply and evaluate the usefulness of a conceptual snow accumulation and melt model to estimate snowpack, and

  3. use the conceptual snow model with simulations of future monthly temperature and precipitation from general circulation models (GCMs) to estimate potential changes in future snowpack in the western US.

Observed and GCM-Simulated Temperature and Precipitation

Following the protocol of the U.S. National Assessment of the Potential Consequences of Climate Variability and Change (U.S. National Assessment) simulations of future climate from the Canadian Centre for Climate Modeling and Analysis (CCC) and the Hadley Centre for Climate Prediction and Research (HADLEY) GCMs were used in this study. The CGCM1 version of the CCC model and the HADCM2 version of the HADLEY model were used.

The CGCM1 is a spectral model with triangular truncation at wave number 32 (yielding a surface grid resolution of roughly 3.75o latitude by 3.75o longitude) and 10 atmospheric levels (Boer et al., in press, Flato et al., in press). The ocean component is based on the Geophysical Fluid Dynamics Laboratory MOM1.1 model and has a resolution of roughly 1.8o of latitude by 1.8o of longitude and 29 vertical levels.

The HADCM2 is a coupled ocean-atmosphere model with a spatial resolution of 2.5o by 3.75o (latitude by longitude) (Johns et al., 1997). The atmospheric component of HADCM2 has 19 levels and the ocean component has 20 levels.

The GCM simulations used in this study are transient. Each simulation includes observed increases in atmospheric CO2 from 1900 to 1993, and subsequent increases in atmospheric CO2 of 1 percent per year to the end of the next century. The simulations also include the direct effects of sulfate aerosols.

As outlined by the U.S. National Assessment, observed monthly temperature and precipitation for 1895-1993, and GCM-simulated scenarios of monthly temperature and precipitation estimated by the CCC and the HADLEY GCMs for the period 1994-2099 were used. The GCM scenarios were computed by applying GCM-simulated changes in monthly temperature and precipitation for the 1994-2099 period to the observed mean data for 1961-90 (current conditions). In addition, as part of the U.S. National Assessment protocol the observed and GCM-simulated data were interpolated to a 5o latitude by 0.5o longitude grid by the Vegetation/Ecosystem Modeling and Analysis Project (VEMAP) (Kittel et al., 1995; 1996).

The use of observed and GCM-simulated data interpolated to the VEMAP grid, rather than using raw GCM output at GCM grid resolutions provided several advantages; the VEMAP grid provides

  1. a common grid among all data sets used in the analyses,

  2. interpolated temperature and precipitation values that include topographic effects that are more realistic than those included in GCMs, and

  3. data for relatively small grids that provide an improved representation of local temperature and precipitation than that obtained using the large grids employed by GCMs.

Conceptual Snow Model

A conceptual snow model was chosen for this study to avoid the need for extensive model calibration and the need for the generation of site-specific model parameters. Model calibration often leads to biases in model parameters (Nash and Gleick, 1991). By fitting model parameters to climatic conditions for some specified period, a bias to the climatic conditions of that period are introduced into the model. Nash and Gleick (1991) suggest that the use of calibrated models for climate-change studies can produce erroneous results if the climatic conditions used for climate-change scenarios differ significantly from the climatic conditions used to calibrate the hydrologic model. In addition, Nash and Gleick (1991) indicate that because hydrologic models often include many parameters, several combinations of parameter values often can be used to develop equally good calibration fits to observed data and can have different effects on model sensitivity to climate inputs.

The snow accumulation and melt model used in this study is based on concepts often used in monthly water balance models (van Hylckama, 1956; McCabe and Ayers, 1989, Tasker et al., 1991). Inputs to the model are monthly temperature (T) and precipitation (P) (Figure 1). The occurrence of snow is computed as

equation       (1)

where S is monthly snow fall in millimeters (mm), P is monthly precipitation in mm, Ta is monthly air temperature in degrees Celsius (oC), Train is a threshold above which all monthly precipitation is rain, and Tsnow is a threshold below which all monthly precipitation is snow. Between Train and Tsnow, the proportion of precipitation that is snow or rain changes linearly.

fig. 1
Figure 1. Schematic diagram of the conceptual snow accumulation and melt model. Ta is monthly temperature, P is monthly precipitation.

This type of model formulation has been used in previous studies (Tarboton et.al, 1991, Tasker et al., 1991). If snow occurs in a given month, it is added to the snowpack and is subject to melt if conditions are such that melting can occur. Thus, for some cases, snow, rain, and snow melt can occur in the same month. For this study, Tsnow was set to 0 oC and Train was set to 5 oC. These values were chosen based on experimentation with the snow model and the observed snowpack data used in this study.

Snow melt is computed as a fraction of the snow storage by,

equation       (2)

where MF is the fraction of snow storage that can be melted in a month. The fraction increases from 0 at Tsnow to 0.5 at and above Train. In this study, the maximum fraction of snow storage that could be melted in a month was set to 0.5 (the maximum rate used by van Hylckama, 1956). In addition, when snow storage reached 10 mm or less, the entire storage was melted.

Snowpack Data

Snowpack data used in this study were measured at 311 snow courses in the western United States (Figure 2). The snow course data were obtained from the United States Department of Agriculture and the California Department of Water Resources. These snow courses have an uninterrupted time-series of snow course data for the winters of 1947-48 through 1986-87. April 1 snow course data (in water equivalent units) were chosen because most snowpack accumulations in the western United States generally reach their peak by the beginning of April.

fig. 2
Figure 2. Locations of snow course stations used in this study. (North latitude is indicated by positive values and west longitude is indicated by negative values.)

Because of the large number of snow courses included in this study, and because previous studies have identified areas in the western United States where snowpack is spatially covariant (McCabe and Legates, 1995; Cayan, 1996), the snowpack data measured at the 311 snow courses were subjected to a cluster analysis to identify groups of snow courses with inter-correlated snowpack. In addition, the aggregation of the snowpack data provided snowpack values on a spatial scale that is more compatible with the spatial scale on which GCMs used in this study operate.

A hierarchical average-linkage clustering method was used to identify groups of inter-correlated snow courses. The clustering was based on correlations between time series of April 1 snowpack measured at each snow course. Figure 3 illustrates changes in the clustering similarity index (i.e. correlation coefficient between 2 clusters that are joined) computed for 20 to 1 clusters. In any classification procedure the desired result is to obtain the minimum number of clusters necessary to represent the data without a great loss of information. The similarity index changes as the number of clusters decreases because of the joining of clusters that are dissimilar. When strongly dissimilar clusters are joined, the change in the similarity index is large and indicates a large loss of information. In this study, the similarity index indicates the correlation between two clusters that are joined. Figure 3 illustrates relatively small decreases in the similarity index until the number of clusters decreases from four to three, the large decrease in the similarity index at this point is indicative of a relatively large loss of information. Thus, four clusters were chosen as the preferred solution of the clustering process.

fig. 3
Figure 3. Coefficient of similarity for 20 to 1 clusters from the correlation-based clustering of April 1 snowpack from 311 snow courses in the western United States.

Because hierarchical clustering methods can include observations in one cluster early in the clustering process that better fit into another cluster, an additional step was used to improve the clustering of the snowpack data. This step involved the correlation of the time series of snowpack at each snow course with the average time series for each cluster. Snowpack data for each snow course were ultimately assigned to the cluster with which they are most highly correlated. An additional requirement was that a snow course was only classified if the cluster-average snowpack of one of the clusters explained at least 50 percent of the variability in snowpack for that snow course. A final processing of the clustering results was performed to remove snow courses that were grouped in one cluster, but were spatially separated from the majority of the snow courses in a cluster (this affected only 18 snow courses). This additional processing increased the likelihood of reliable clusters. However, the additional processing also resulted in the non-inclusion of some snow courses into clusters.

The clustering procedure produced four clusters (clusters A through D in figure 4) that included 218 (70 percent) of the 311 snow courses analyzed. Cluster A includes snow courses in the Pacific Northwest and Northern Rocky Mountains (PNW, 28 snow courses), cluster B primarily includes snow courses in the central Rocky Mountains (CRM, 74 snow courses), cluster C includes the snow courses in the Sierra Nevada (SN, 89 snow courses), and cluster D includes snow courses in the southern Rocky Mountains (SRM, 27 snow courses). Cayan (1996) performed a principal components analysis (PCA) of snowpack data in the western US which produced 5 regions of inter-correlated snowpack; the Pacific Northwest, the northern Rocky Mountains, the central Rocky Mountains, the southern Rocky Mountains, and the Sierra Nevada/Great Basin region. The loading patterns of the PCA indicated a large amount of overlap between these 5 regions. The primary difference between the classification of Cayan and the classification discussed above is that, in Cayan's classification, the snow courses in the Pacific Northwest and northern Rocky Mountain regions are separated into two clusters.

fig. 4
Figure 4. Locations of snow courses included in each of the final clusters. A - Pacific Northwest, B - Central Rocky Mountains, C - Sierra Nevada, and D - Southern Rocky Mountains.

Climatology of Cluster-Average Snowpack

Average time series of snowpack were computed for each of the final clusters (Figure 5). The time series indicate that the largest and most variable snowpacks occur in the PNW and SN regions. A trend analysis (using Kendall's slope) of the time series indicated that a significant trend (at a 95 percent confidence level) exists only in the time series for the PNW. This trend is negative, largely due to a significant decrease in PNW snowpack conditions after the mid-1970's. This decrease in PNW snowpack is related to a shift in atmospheric circulation over the eastern North Pacific Ocean and the western US during the mid-1970s (McCabe and Legates, 1995).

fig. 5
Figure 5. Time series of observed cluster-average April 1 snowpack. Region labels are as follows, PNW - Pacific Northwest, CRM - Central Rocky Mountains, SN - Sierra Nevada, and SRM - Southern Rocky Mountains.

Monthly temperature and precipitation for each snowpack cluster were computed from observed (1895-1993) temperature and precipitation data interpolated to the VEMAP grid. VEMAP gridded data for the grids closest (within 0.25 degrees latitude and longitude) to each snow course included in each cluster were used to compute cluster-average winter temperature and precipitation. Figure 6 illustrates the VEMAP grids from which monthly temperature and precipitation data were obtained for each cluster. Correlations between cluster-average snowpack and cluster-average temperature and precipitation indicate significant correlations between snowpack and precipitation for all regions and significant correlations between temperature and snowpack for all clusters except the CRM cluster (Table 1). For all clusters, except the SRM cluster, the correlations between temperature and precipitation are non-significant, indicating no co-variability between these two variables. The correlations between snowpack and precipitation are positive, and larger than those between snowpack and temperature. Thus, the supply of winter moisture to a region is the best predictor of snowpack. Correlations between snowpack and temperature are negative, and significant for most clusters. These significant negative correlations indicate the effect of temperature on the ratio of rain to snow (i.e. for the same precipitation level, higher temperatures result in a larger ratio of rain to snow and a smaller snowpack). The non-significant correlation between snowpack and temperature for the CRM region is due, in part, to the cold temperatures for this cluster. The mean winter temperature for this cluster is the lowest of all the clusters, and, in the observed record, winter temperatures were not warm enough to significantly affect rain to snow ratios.

fig. 6
Figure 6. Vegetation/Ecosystem Modeling and Analysis Project (VEMAP) grids from which monthly temperature and precipitation data were obtained to compute cluster-average temperature and precipitation. Letters (A - D) indicate cluster designations.

table 1
Table 1. Correlations between cluster-average winter (November through March) temperature (T) and precipitation (P) and April 1 snowpack (S). (* indicates correlations that are significant at a 95 percent confidence level. Region labels are as follows, PNW - Pacific Northwest, CRM - Central Rocky Mountains, SN - Sierra Nevada, and SRM - Southern Rocky Mountains.)

Evaluation of the Conceptual Snow Model

Time series of cluster-average temperature and precipitation for each cluster were used as inputs to the conceptual snow model to simulate April 1 snow storage for the years 1948-87. The simulated time series were then compared with cluster-average observed April 1 snowpack. During this analysis it became apparent that the VEMAP precipitation data were much lower on average than cluster-average observed April 1 snowpack. Observations indicate that winter precipitation is greater than April 1 snowpack accumulations in the western US (Serreze et. al., 1999). The underestimation of winter precipitation by the VEMAP data is due, in part, to the smoothed topography represented by the VEMAP grids. The mean elevations for the VEMAP grids often are lower than the elevations of many of the individual snow courses. Orographic effects on precipitation are important in the mountainous western US and precipitation varies significantly by elevation (Hay and McCabe, 1998). The under-representation of snow course elevations by the VEMAP grids results in a general underestimation of snow course precipitation and may also result in over-estimation of temperature.

Because of the underestimation of winter precipitation (and possibly over-estimation of temperature) by the VEMAP gridded data, some correction was needed for the VEMAP data to estimate April 1 snowpack. To determine the correction needed for the VEMAP temperature and precipitation data, a range of changes in temperature (0 through -10 oC) and precipitation (0 through 100 percent increases) were applied to the cluster-average VEMAP monthly temperature and precipitation data and used with the snow model to estimate April 1 snowpack for each cluster. For each combination of changes in temperature and precipitation, model-estimated snowpack was compared with observed cluster-average snowpack for the period 1948 through 1987. To evaluate the agreement between estimated and observed April 1 snowpack the Index of Agreement (d) was computed (Willmott, 1981; Willmott, 1984; Legates and McCabe, 1999).

The Index of Agreement varies from 0.0 (poor model) to 1.0 (perfect model) and was developed to overcome the insensitivity of correlation-based measures to differences in observed and model-simulated means and variances (Willmott; 1981; Willmott, 1984). The Index of Agreement is computed as

equation       (3)

where Oi is the observed data, Pi is the model-simulated values, o-bar is the observed mean, and N is the number of cases.

General corrections to the VEMAP data were desired, rather than site specific corrections, therefore an average d statistic for all clusters was computed for each combination of changes in temperature and precipitation to determine the general corrections to VEMAP temperature and precipitation data needed to estimate April 1 snowpack (Figure 7). Results of these experiments indicated that for the snow model used in this study, cluster-average VEMAP precipitation needed to be multiplied by 1.5 to reliably estimate observed April 1 snowpack for the period 1948-87, and no correction to cluster-average VEMAP temperature data was needed (Figure 7).

fig. 7
Figure 7. Average values of the Index of Agreement for all clusters and for various changes in temperature and precipitation. The Index of Agreement represents the agreement between observed April 1 snowpack for the period 1948-1987 and multiple simulations of April 1 snowpack using a range of changes in temperature and precipitation.

Observed and Estimated April 1 Snowpack

Estimates of April 1 snowpack for the years 1948-87 computed using the conceptual snow model and cluster-average monthly VEMAP temperature and adjusted precipitation were compared with observed cluster-average snowpack. Results indicate that 74 to 77 percent of the variability in observed snowpack is explained by the snow model (Figure 8). The large amount of variability explained by the model indicates that the model properly accounts for the interaction between precipitation and temperature on snowpack. Errors in estimates (based on root mean square errors) ranged from 12 to 27 percent of observed means, with the largest error occurring for the SN region. Errors in model estimates are likely due to
  1. inadequate adjustment of VEMAP gridded precipitation data,
  2. errors in VEMAP gridded temperature data, and
  3. incomplete representation of physical processes by the model.
fig. 8
Figure 8. Comparison of observed and estimated April 1 snowpack using the conceptual snow model for 1948-1987. Region labels are as follows, PNW - Pacific Northwest, CRM - Central Rocky Mountains, SN - Sierra Nevada, and SRM - Southern Rocky Mountains. (r2 - coefficient of determination, rmse - root mean square error, and bias - simulated mean minus the observed mean)

Because this analysis is performed on a monthly basis for regionally-clustered snow course data there are several caveats that must be noted. First, the cluster-average snowpack is not intended to represent individual snowpack accumulations for all locations within a region, but instead provides a characterization of snowpack for a representative watershed in each region. Thus, within cluster differences in snowpack are not considered in this study as these differences are considered to be too small to be simulated reliably by GCMs. In addition, it must be noted that some of the correlations between snow courses used to develop the snow clusters may change as climate changes. However, the majority of these changes will occur for snow courses that are marginally correlated with other snow courses, and the bulk of snow courses included in each cluster most likely will remain highly correlated.

Estimates of Future Snowpack

VEMAP gridded observed temperature and precipitation data for the period 1961-90 (representing current climatic conditions) and VEMAP-gridded CCC and HADLEY GCM-simulated temperature and precipitation for the years 1994-2099 were averaged for each snowpack cluster and used with the snow model to examine the effects of future climatic conditions on snowpack in the western US. Cluster-average precipitation data for future conditions were multiplied by 1.5 as were the historical precipitation data. Periods of specific interest were 2025-2034 (representing approximately 25 years into the future) and 2090-99 (representing approximately 100 years into the future).

CCC Simulations

The CCC model simulates large (6 oC - 8 oC) increases in winter temperature for all clusters over the next century (Figure 9). The CCC also model simulates slight decreases in winter precipitation over the next 25 years for the PNW, CRM, and SRM clusters and an increase in winter precipitation for the SN cluster. In addition, the CCC model estimates increases in winter precipitation for all clusters by the end of the next century, with a particularly large (86 percent) increase for the SN cluster.

fig. 9
Figure 9. Mean winter (November through March) temperature and precipitation simulated by the Canadian Centre for Climate Modeling and Analysis general circulation model for 1961-90, 2025-2034, and 2090-2099. Region labels are as follows, PNW - Pacific Northwest, CRM - Central Rocky Mountains, SN - Sierra Nevada, and SRM - Southern Rocky Mountains.

The combined effects of the CCC-simulated changes in temperature and precipitation indicate a decrease in April 1 snowpack for all clusters during the next century (Figure 10, Table 2), even though simulations indicate an increase in winter precipitation by the end of the century for all clusters (Table 3). The largest percent decreases (near 100 percent less) are for the SN and SRM regions. The resulting decrease in snowpack in the western US is due to the simulated increase in winter temperatures which increases the ratio of rain to snow and the melting of snowpack. The largest decrease in percent of snowpack, as simulated using the CCC temperature and precipitation time series, occurred for the SN cluster.

fig. 10
Figure 10. Ten-year moving average percent change in April 1 snowpack simulated by the Canadian Centre for Climate Modeling and Analysis general circulation model. Region labels are as follows, PNW - Pacific Northwest, CRM - Central Rocky Mountains, SN - Sierra Nevada, and SRM - Southern Rocky Mountains.

table 2
Table 2. Changes in April 1 snowpack estimated using the conceptual snow model and simulated monthly temperature and precipitation from the Canadian Centre for Climate Modeling and Analysis (CCC) and Hadley Centre for Climate Prediction and Research (HADLEY) general circulation models for the periods 2025-2034 and 2090-2099. Changes are differences (in millimeters and percent [in parentheses]) in mean April 1 snowpack from the 1961-1990 mean. (** indicates significant differences in mean April 1 snowpack from current conditions (1961-90) at a 99 percent confidence level, and * indicates significant differences at a 95 percent confidence level. Region labels are as follows, PNW - Pacific Northwest, CRM - Central Rocky Mountains, SN - Sierra Nevada, and SRM - Southern Rocky Mountains.)

table 3
Table 3. Trends in winter (November through March) temperature (T) and precipitation (P), and estimated April 1 snowpack (S) for the period 2000-2099 expressed as correlations with time (year) simulated by (A) the Canadian Centre for Climate Modeling and Analysis (CCC) general circulation model (GCM), and (B) Hadley Centre for Climate Prediction and Research (HADLEY) GCM. (Region labels are as follows, PNW - Pacific Northwest, CRM - Central Rocky Mountains, SN - Sierra Nevada, and SRM - Southern Rocky Mountains. * indicates correlations that are significant at a 99 percent confidence level.)

HADLEY Simulations

The HADLEY model simulates changes in winter temperature (4 oC - 5 oC, Figure 11) that are smaller than those simulated by the CCC model (Figure 8). Similar to the CCC model, the HADLEY model simulates the largest increase in precipitation for the SN cluster. The HADLEY model simulates increased temperatures and precipitation during the next century for all clusters. The effects of the HADLEY-simulated changes in temperature and precipitation indicate a decrease (55 to 90 percent) in snowpack for the PNW, SN, and SRM regions by the end of the next century (Figure 12, Table 2). For the CRM region, the overall effect is no significant change in snowpack conditions (Figure 12, Tables 2 and 3).

fig. 11
Figure 11. Mean winter (November through March) temperature and precipitation simulated by the Hadley Centre for Climate Prediction and Research general circulation model for 1961-90, 2025-2034, and 2090-2099. Region labels are as follows, PNW - Pacific Northwest, CRM - Central Rocky Mountains, SN - Sierra Nevada, and SRM - Southern Rocky Mountains.

fig. 12
Figure 12. Ten-year moving average percent change in April 1 snowpack simulated by the Hadley Centre for Climate Prediction and Research general circulation model. Region labels are as follows, PNW - Pacific Northwest, CRM - Central Rocky Mountains, SN - Sierra Nevada, and SRM - Southern Rocky Mountains.

The results for the PNW computed using the HADLEY scenarios are similar to results obtained by Hamlet and Lettenmaier (in press) for the Columbia River basin. Hamlet and Lettenmaier used a macro-scale hydrology model driven by climate scenarios derived from GCMs to estimate future hydrologic conditions in the Columbia River basin. They found, using the HADLEY scenario, that by the year 2045 winter precipitation in the Columbia River basin will increase significantly, but because of increased temperatures March 1 snowpack will be decreased by as much as 35 percent (the study presented here shows a 24 percent decrease in April 1 snowpack for the 2025-34 period, Table 2). Hamlet and Lettenmaier also found that winter runoff will increase by approximately 50 percent, and spring/summer runoff will decrease by 10 percent. By 2095 Hamlet and Lettenmaier found that the HADLEY scenario suggests that the Columbia River basin will shift from a snow-melt dominated system to a transient snow system.

Statistical Significance of Estimated Changes

Using a conceptual snow model, monthly temperature and precipitation simulated by the CCC and HADLEY GCMs indicate what appear to be substantial changes in mean April 1 snowpack for future climatic conditions. These changes, however, need to be evaluated with respect to natural climatic variability to determine whether they are statistically significant (McCabe and Wolock, 1991; 1997). To accomplish this task, two-tailed Student t-tests were used to compare the distributions of estimated cluster-average snowpack for the 1961-90 period to distributions of cluster-average snowpack estimated using GCM-simulations for the periods 2025-2034 and 2090-2099. The t-tests were performed to determine if the mean values of cluster-average April 1 snowpack simulated for future conditions were significantly different (at a 95 percent confidence level) from the means for the 1961-90 period, given the variability in the data.

Mean cluster snowpack simulated using the CCC model data were significantly different from the 1961-90 means for both the 2025-2034 and 2090-2099 periods for all clusters (Table 2). The HADLEY-simulated data, however, indicated non-significant changes in mean snowpack for the CRM and SRM clusters for the 2025-2034 period, and non-significant changes in mean snowpack for the CRM cluster for the 2090-2099 period (Table 2).

These results suggest that in the short term (i.e. the next 25 years) there may be significant decreases in snowpack in the PNW and SN clusters. By the end of the next century, the GCM simulations suggest a significant decrease in snowpack in all regions, except possibly in the CRM region.

An interesting result of these simulations is that snowpack conditions may decrease in the future in spite of GCM estimates of a general increase in winter precipitation toward the latter half of the next century. This result occurs primarily because temperatures are estimated to increase enough to increase the ratio of rain to snow. Thus, more winter precipitation will occur as rain rather than as snow which alters the timing of runoff and reduces April 1 snowpack in most of the western US.

Conclusions

April 1 snowpack accumulations measured at 311 snow courses in the western United States (US) were grouped using a correlation-based cluster analysis into four regions of inter-correlated snowpack. A conceptual snow accumulation and melt model and monthly temperature and precipitation for each snowpack region computed using gridded data from VEMAP were used to estimate cluster-average April 1 snowpack for each region. The conceptual snow model reliably simulated the response of snowpack to variations in temperature and precipitation for the historical period 1948-87.

The conceptual snow model was subsequently used to estimate future snowpack by using simulations of monthly temperature and precipitation from the CCC and HADLEY GCMs. Results for the CCC model indicate that although winter precipitation is estimated to increase in the future, increases in temperatures will result in large decreases in April 1 snowpack for the entire western US. Results for the HADLEY model also indicate large decreases in April 1 snowpack for most of the western US, but the decreases are not as severe as those estimated using the CCC simulations. Although snowpack conditions are estimated to decrease for most areas of the western US into the future, both GCMs estimate a general increase in winter precipitation toward the latter half of the next century. These results suggest that although water quantity may be increased in the western US, simulated temperature increases will result in precipitation occurring more frequently as rain rather than as snow, April 1 snowpack will be reduced, and the timing of runoff in the western US will be altered.

Literature Cited

Boer, G.J., G.M. Flato, and D. Ramsden, in press. A Transient Climate Change Simulation With Historical and Projected Greenhouse Gas and Aerosol Forcing: Projected Climate for the 21st Century. Climate Dynamics.

Bolin, B., 1986. How much CO2 will remain in the atmosphere? in The Greenhouse Effect, Climate Change and the Ecosystem, B. Bolin, B.R. Doos, J. Jaeger, and R. Warrick (Editors). John Wiley, New York.

Cayan, D.R., 1996. Interannual Climate Variability and Snowpack in the Western United States. Journal of Climate 9, 928-948.

Flato, G.M., G.J. Boer, W.G. Lee, N.A. McFarlane, D. Ramsden, M.C. Reader, and A.J. Weaver, in press. The Canadian Centre for Climate Modelling and Analysis Global Coupled Model and its Climate. Climate Dynamics.

Gary, D.M. and D.H. Male, 1981. Handbook of Snow. Pergamon Press, New York.

Gammon, R.H., E.T. Sundquist, and P.J. Fraser, 1985. History of carbon dioxide in the atmosphere. in Atmospheric Carbon Dioxide and the Global Carbon Cycle, J.R. Trabalka (Editor). U.S. Department of Energy, Carbon Dioxide Research Division, DOE/ER-0239, pp. 28-62.

Gleick, P.H., 1987. Regional hydrologic consequences of increase in atmospheric CO2 and Other Trace Gases. Climatic Change 10, 137-161.

Hamlet, A.F. and D.P. Lettenmaier, in press. Effects of Climate Change on Hydrology and Water Resources Objectives in the Columbia River Basin. Journal of the American Water Resources Association.

Hay, L.E. and G.J. McCabe, 1998. Verification of the Rhea-Orographic-Precipitation Model. Journal of the American Water Resources Association 34, 103-112.

Johns T.C., R.E. Carnell, J.F. Crossley, J.M. Gregory, J.F.B. Mitchell, C.A. Senior, S.F.B. Tett, and R.A. Wood, 1997. The Second Hadley Centre Coupled Ocean-Atmosphere GCM: Model Description, Spinup and Validation. Climate Dynamics 13, 103-134.

Kittel, T.G.F., N.A. Rosenbloom, T.H. Painter, D.S. Schimel, and VEMAP Modeling Participants, 1995. The VEMAP Integrated Database for Modeling United States Ecosystem/Vegetation Sensitivity to Climate Change. Journal of Biogeography 22(4-5), 857-862.

Kittel, T.G.F., N.A. Rosenbloom, T.H. Painter, D.S. Schimel, H.H. Fisher, A. Grimsdell, VEMAP Participants, C. Daly, and E.R. Hunt, Jr., 1996. The VEMAP Phase I Database: An Integrated Input Dataset for Ecosystem and Vegetation Modeling for the Conterminous United States. CDROM and World Wide Web (URL= http://www.cgd.ucar.edu/vemap/).

Legates, D.R. and G.J. McCabe, 1999. Evaluating the "Goodness-of-Fit" Measures in Hydrologic and Hydroclimatic Model Validation. Water Resources Research 35, 233-241.

Lettenmaier, D.P. and D.P. Sheer, 1991. Climatic Sensitivity of California Water Resources. Journal of Water Resources Planning Management 117, 108-125.

Lins, H.F., E.T. Sundquist, and T.A. Ager, 1988. Information on Selected Climate and Climate-Change Issues. U.S. Geological Survey Open-File Report 88-718. U.S. Geological Survey, Reston, Virginia.

McCabe, G.J. and M. Ayers, M., 1989. Hydrologic Effects of Climate Change in the Delaware River Basin. Water Resources Bulletin 25, 1231-1242.

McCabe, G.J. and D.M. Wolock, 1991. Detectability of the effects of a hypothetical temperature increase on the Thornthwaite moisture index. Journal of Hydrology 125, 25-35.

McCabe, G.J. and D.R. Legates, 1995. Relations Between Atmospheric Circulation and Snowpack Accumulations in the Western United States. International Journal of Climatology. 15, 517-530.

McCabe, G.J. and D.M. Wolock, 1997. Climate change and the detection of trends in annual runoff. Climate Research 8, 129-134.

Nash, L.L. and P.H. Gleick, 1991. Sensitivity of Streamflow in the Colorado Basin to Climatic Changes. Journal of Hydrology 125, 221-241.

Serreze, M.C., M.P. Clark, and R.L. Armstrong, 1999. Characteristics of the Western United States Snowpack from Snowpack Telemetry (SNOTEL) Data. Water Resources Research 35, 2145-2160.

Tasker, G., M. Ayers, D. Wolock, and G. McCabe, 1991. Sensitivity of Drought Risks in the Delaware River Basin to Climate Change. Proceedings, Technical and Business Exhibition and Symposium, Huntsville Association of Technical Societies, Huntsville, Alabama, pp. 153-158.

Tarboton, D.G., M.J. Al-Adhami, and D.S. Bowles, 1991. A Preliminary Comparison of Snowmelt Models for Erosion Prediction. Proceedings, 59th Annual Western Snow Conference, Juneau, Alaska. pp. 79-90.

van Hylckama, T.E.A., 1956. The Water Balance of the Earth. Publications in Climatology 9, 1-117.

Willmott, C.J., 1981. On the validation of models. Physical Geography 2, 184-194.

Willmott, C.J., 1984. On the evaluation of model performance in physical geography. in Spatial Statistics and Models, G.L. Gaile, and C.J. Willmott (Editors). D. Reidel, Dordrecht, 443-460.

World Meteorological Organization, 1970. Guide to Hydrometeorological practices. WMO-No. 168.TP.82, Geneva.

Back to the SMIG Features Page

button bar

SMIG Home Mailing List Features Conferences Classes Reading Room Model Archives Feedback Home | Mailing List | Features | Conferences | Classes | Reading | Model Archives | Feedback

Stewart Rounds, SMIG coordinator <sarounds@usgs.gov>
U.S. Geological Survey
http://smig.usgs.gov/SMIG/features_0300/snowpack_gcm.html
Last modified Thursday, 10-Mar-2005 17:40:52 EST
Privacy Statement · Disclaimer · FOIA · Accessibility