Selection and post-processing of climate model outputs to build an ensemble of standard climate scenarios

As stated in the 2013-2020 Government Strategy for Climate Change Adaptation, a growing number of sectors are showing an interest climate change adaptation, thereby leading to an increase in the demand for climate scenarios for sectors of activity that are often interrelated.

In order to satisfy this growing demand and ensure the availability of climate scenarios that meet the needs of vulnerability, impact and adaptation (VIA) studies, a standard scenarios ensemble has been produced for the province of Québec. This ensemble is of a manageable size for users and covers the range of expected changes in temperature and precipitation for Québec.

• Generate an ensemble of standard scenarios that meet VIA study needs in order to facilitate comparisons across different sectors of application;

• Reduce the number of climate scenarios necessary for VIA studies;

• Ensure optimal coverage of projected changes in temperature and precipitation for Québec;

• Improve the efficiency of climate scenario production and delivery.

• Carry out an inventory of climate scenario requests in order to identify recurring needs;

• Perform a cluster analysis to select an ensemble of climate simulations from the Coupled Model Intercomparison Project Phase 51 ;

• Calibrate a quantile correction (1D method) for the climate model time series against the gridded interpolated dataset of daily observations with a grid spacing of 10 km2 ;

• Apply the quantile correction to the climate simulation time series from 1961 to 2100.

Simulations selection

A sub-group of 22 simulations was selected from the complete ensemble of available CMIP5 simulations (RCP 4.5 & 8.5) (Table 1). The range in projected changes between the selection and the original ensemble was statistically validated over several regions in North America (figure 1).

Table 1. Simulations selected by cluster analysis

Figure 1. Uncertainty coverage validation zones (grey points and rectangles) and the area covered by data produced to date (red)

The results indicate that the distribution of projected changes provided by the selection is never statistically different than that of the complete ensemble (15 regions x 36 criteria x 2 future horizons with α =0.05). The validation variables include projected changes in maximum and minimum temperature and in total precipitation for two future horizons (2041 to 2070 and 2071 to 2100).

Post-processing

The post-processing applied to the raw climate model output aims to refine the spatial scale of the climate scenarios and correct systematic biases in the simulations using quantile-based methods. The frequency distributions of maximum and minimum temperature and total daily precipitation were adjusted to match the gridded observational data. A moving average of 30 days was used, making it possible to take into account bias variations within the annual cycle. The method preserved the long-term trends produced by each climate simulation.

Limitations

The gridded observational data used for post-processing represent an approximation of the current historical climate, and may be biased at high altitudes or in areas with a low meteorological station density. The data are also unreliable for characterizing daily precipitation extremes at a fine spatial scale.

These limitations will also be present in the climate scenarios. Moreover, the range of projected changes provided by the selected simulations in comparison to the original ensemble was validated only for changes in monthly averages for three climatic variables. This does not necessarily guarantee adequate coverage in terms of changes in extreme values or for other climate indices.

• École de technologie supérieure (ÉTS)