Following on the success of its 2009 climate assessment project on British Columbia’s Coquihalla Highway, PCIC was asked to provide a similar assessment for the Yellowhead Highway between Priestly Hill and Vanderhoof. In both assessments, past and future simulations of the impact of climate change on highway infrastructure were investigated.
Using downscaled results from several climate models, present and future estimates for temperature and precipitation were made, including a probability assessment of climate extremes. This study used regional climate model (RCM) simulations provided by NARCCAP, evaluated for the highway areas and with climate parameters and meteorological events that were defined by provincial and local engineers.
Methods
Two downscaling approaches were applied in the Yellowhead project: probability mapping and statistical downscaling.
Probability Mapping
Probability mapping is used to estimate climate change induced shifts in the frequency of climate or weather extremes (e.g., localized heavy rainfall events) using RCMs. When a heavy localized rainfall event is recorded at a particular weather station, the average rainfall over a larger area such as that represented by an RCM gridbox (approximately 50 km x 50 km) will be considerably less than the heavy localized precipitation experienced at the station. The most extreme events are thus not directly represented in the RCM, which provides only area averages over entire gridboxes. To relate the amount of heavy precipitation at a station to a different (lower) threshold in the RCM that corresponds to the heavy rainfail event, the threshold for defining the event must itself be downscaled.
To illustrate this, Figures 1 and 2 depict the process for a normally distributed quantity such as temperature. The mapping is defined by identifying corresponding station and RCM quantiles (cf. the empirical transformation of Panofsky and Brier [1958]), and is applied to future RCM events.
Figures 1 and 2: Probability mapping. Local (station) and regional (RCM) scales of equal probability are identified, and future probabilities are derived directly from the RCM.
Statistical Downscaling
In addition to the extremes addressed by threshold mapping, there is a need to consider changes to rare extreme events, such as those which would be expected to recur once every 100 years, on average. It is generally not possible to estimate the actual size of a typical 100-year rainfall event for the end of the 21st century using probability mapping alone. One possible solution to the problem is to use a full statistical (empirical) downscaling model. The goal is to obtain a quantitative link between large-scale atmospheric circulation and local-scale climate or weather events. This link is usually estimated from historical records of large-scale fields using reanalyses and small-scale station observations by using a statistical model. This model is applied to simulated future atmospheric fields to obtain future climate data for the station in question.
For the Yellowhead assessment PCIC applied the Expanded Downscaling (EDS) method. EDS simulates local events as closely as possible and consistent with the prevailing atmospheric circulation (Bürger 1996). At the same time the EDS method generates local covariability that is realistic enough to be used for studying the impact of climate on weather extremes such as floods and droughts, and driving corresponding impact models.
Results
In general, the results from both probability mapping and statistical downscaling project an increase in temperature and precipitation for the study area. This trend towards a warmer and wetter future climate is highlighted in the more specific impacts highlighted below.
Probability Mapping
For the Yellowhead Highway, probability mapping reveals that rising temperatures will have the greatest impact on cold extremes. Low temperature events of -35°C or colder are projected to occur less frequently in the future, less than once per year instead of the current rate of five times per year. Similarly, an increase in the occurrence of very hot days (i.e., greater than 35°C) is projected. Most models project an increase in heavy precipitation events (more than 35 mm per day), though the actual probability (one such event every 30 years) is uncertain due to the small sampling size. All models agree that ground freeze, where maximum daily temperature is less than -5°C, will occur less often in the future.
A very important quantity for engineering purposes is snow accumulation but only one model, the Canadian Regional Climate Model (CRCM), reported results for snow accumulation, which limits confidence in future predictions for this quantity. However, the CRCM clearly projected a decrease in snowpack, a result corroborated by statistical downscaling.
Statistical Downscaling
EDS-based statistical downscaling produces daily time series of present and future climate for three variables: minimum temperature, maximum temperature and precipitation. From these variables, annual time series can be derived for the 27 Climdex indices used by the World Meteorological Organization to monitor climate extremes.
For example, future projections show a dramatic decrease in the number of frost days as well as an increase in the length of the growing season. Also estimated were the extremes for each of the three variables. Despite considerable uncertainty surrounding these estimates, a significant increase in the 100-year event for precipitation (42-66 mm per day) was projected. For maximum temperature the 100-year event increases from 39°C to 42°C, while the corresponding minimum temperature event increases from -55°C to -50°C.
References
Bürger, G, 1996: Expanded Downscaling for Generating Local Weather Scenarios. Climate Research, 7, 111-128.
Acknowledgements
BC Ministry of Transportation and Infrastructure: Dirk Nyland and Jim Barnes.