GCM climate predictions

Downscaling

For many climate change studies, scenarios of climate change derived directly from global climate models (GCMs) are of insufficient spatial and temporal resolution. Spatial downscaling techniques (Hewitson & Crane 1996; Wilby & Wigley 1997) are used to derive finer resolution climate information from coarser resolution GCM output. The fundamental assumption behind all these methods is that the statistical relationships, linking observed time series to GCM variables, will remain valid under future climate conditions.

Fig. 2. (a) Borehole locations and depths in Grand Forks valley. Only boreholes with lithologs in BC water well database are shown, (b) Fence diagram of hydrostratigraphic units in the Grand Forks valley.

GCMs do not accurately predict local climate, but the internal consistency of these physically based climate models provides most likely estimates of ratios and differences (scaling factors) from historical (base case) to predicted scenarios (Loaiciga et al. 1996) for climatic variables, such as precipitation and temperature.

Climate scenarios for modelled present and future conditions were taken from the Canadian Global Coupled Model (CGCM1) (Flato et al. 2000) for the IPCC IS92a greenhouse gas plus aerosol (GHG + A1) transient simulation. CGCM1 predictions are valid for Canada and fall in the average of other GCMs. These include relative and absolute changes in precipitation and temperature. Precipitation variables were: mean, median, maximum, variance, dry/wet spell length, and percentage wet days in the month. Temperature statistics included: mean, median, minimum, maximum, variance, and interquartile range. Four daily data sets for CGCM1 were obtained from the Canadian Institute for Climate Studies

(CCIS 2004) for a grid location nearest to Grand Forks (Y = 11, latitude 50.09°N and X = 16, longitude 120°W; Grand Forks is at 49.1N and 118.2W). Three were CGCM1 scenarios, each with data for a number of potential predictor variables. The 'current climate' scenario was generated by CGCM1 for the period 1961-2000. The subsequent 'future climate' experiments using CGCM1 with GHG + A1 were for the 2020s, 2050s and 2080s.

The fourth data set was a calibration data set, which contains observed daily data for 1961 -2000, derived from the NCEP (National Centre for Environmental Prediction) re-analysis data set (Kalnay et al. 1996) for the period 1961-2000. This dataset provides large-scale climate variables that can be used to define analogues with GCMs for climate modelling purposes. Most climate modelling experiments in North America use the NCEP datasets for calibration of downscaling models. Monthly means and other statistics were calculated from mean daily values, and the NCEP dataset had 10% or smaller bias to observed precipitation at Grand Forks (compared monthly means), thus we have high confidence in using NCEP data for calibration of downscaling model. The NCEP dataset includes relative humidity, whereas CGCM1 datasets do not, so specific humidity was used when calibrating the model.

The downscaling of CGCM1 results was accomplished using two independent methods: (1) statistical downscaling model (SDSM) software (Wilby et al. 2002) and (2) principle component K-nn method (e.g. Yates et al. 2003; Whitfield & Cannon 2000); the results of both were compared. Four climate scenarios (30 years of daily weather)

were generated using each calibrated downscaling model: current climate (1960-1999), 2020s climate (2010-2039), 2050s climate (20402069), and 2080s climate (2070-2099).

Downscaled daily temperature time series were analysed for (1) mean, and (2) standard deviation. Predictions for mean monthly temperature are almost identical between SDSM and K-nn and calibration bias is small (Fig. 3). Similarly, both methods agree in the magnitudes and directions of temperature change, and represent an increase of approximately 1 °C per 30 years for all months. Downscaled daily precipitation time series were analysed for: (1) mean monthly precipitation, (2) standard deviation in daily precipitation, (3) percentage wet days, (4) dry series length, and (5) wet series length. SDSM results for precipitation differ from K-nn downscaling results, with mean precipitation variation being somewhat better represented by SDSM compared to observed, especially for winter and spring seasons (Fig. 4). SDSM predicts an increase in summer precipitation in future climate, but K-nn predicts a decrease at the same time. K-nn also predicts larger precipitation increases in winter than SDSM, and more precipitation variability into the future.

Clearly, the choice of downscaling method is very important for interpreting predictions of GCM models, as GCMs do not directly model precipitation at a local site. It should be noted that at Grand Forks, the local climate is not modelled very well in the CGCM1 grid cell, probably due to local convective precipitation and valley-mountain-rainshadow effects, which have a strong influence on local precipitation. The poor

1961

2000 Observed

—•—

1961

-2000 SDSM

—e—

2010

-2039 SDSM

2040

2069 SDSM

- - -D - ■

2070

2099 SDSM

1961-2000 Observed -»—1961-2000 PCA K-nn ■e—2010-2039 PCA K-nn

1961-2000 Observed -»—1961-2000 PCA K-nn ■e—2010-2039 PCA K-nn

Month

Fig. 3. Mean monthly temperature at Grand Forks, BC: observed and downscaled from CGCM1 model runs for current and future climate scenarios using (a) SDSM and (b) K-nn.

Month

Month

Fig. 3. Mean monthly temperature at Grand Forks, BC: observed and downscaled from CGCM1 model runs for current and future climate scenarios using (a) SDSM and (b) K-nn.

1961-2000 Observed » 1961-2000 SDSM —6—2010-2039 SDSM 2040-2069 SDSM ---El--2070-2099 SDSM

85

1961-2000 Observed 1 1961-2000 PCA K-nn —e—2010-2039 PCA K-nn 2040-2069 PCA K-nn ■ ■ -D - ■ 2070-2099 PCA K-nn

r

1961-2000 Observed » 1961-2000 SDSM —6—2010-2039 SDSM 2040-2069 SDSM ---El--2070-2099 SDSM

Month

Fig. 4. Mean monthly precipitation at Grand Forks, BC: observed and downscaled from CGCM1 model runs for current and future climate scenarios using (a) SDSM and (b) K-nn.

Month

Month

Fig. 4. Mean monthly precipitation at Grand Forks, BC: observed and downscaled from CGCM1 model runs for current and future climate scenarios using (a) SDSM and (b) K-nn.

downscaling results for precipitation did not allow us to use these data directly in a recharge model. Our approach was to compute change factors (absolute for temperature and relative for precipitation), and redistribute them to daily time series using a stochastic weather generator. An important assumption is made that the GCM can predict these absolute and relative changes, which then can be used to perturb current weather to arrive at future weather conditions. Although this uncertainty limits the predictive aspect of this (and similar) studies, it does not detract from the study's usefulness as a realistic sensitivity analysis to potential climate change, whatever the actual climate changes in each month will be in the future.

For the purpose of this study, only SDSM down-scaled results were selected for further modelling of weather and recharge as inputs to groundwater flow models. Details of the SDSM downscaling, calibration and a comparison of the results are provided by Allen et al. (2004b). CGCM1 downscaling was also used to predict basin-scale runoff for the Kettle River upstream of Grand Forks (Whitfield & Cannon 2000) as described later in this paper.

0 0

Post a comment