Scoping climate data for a risk assessment

4 minutes

Before anyone scores a risk, someone decides which climate data goes into the assessment: the baseline, the horizons, the scenarios, the models, the indices and the resolution. Those choices set what the assessment can and cannot say. Here is what to settle before the work starts, and what to ask for in the brief.

Start from the decisions, not the data

A climate risk assessment is only as useful as the thresholds it tests. A culvert is overwhelmed when rainfall exceeds its design value, a building’s cooling plant when hot days exceed its capacity, a northern road when freezing degree days fall too low to keep its foundation frozen. List the assets, services and decisions in scope first, then the threshold that matters for each. The climate indices follow from the thresholds, not the other way round.

Baseline and horizons

Climate is reported over 30-year periods. Use a recent baseline, 1981 to 2010 or 1991 to 2020, and tie the future horizons to the lives of the assets you are assessing: the 2050s (2041 to 2070) for most municipal assets, the 2080s (2071 to 2100) for long-lived infrastructure, and beyond 2100 for dams, mines and containment that must last for generations. Keep the same horizons in the climate analysis and in your risk matrix, so a likelihood score for the 2050s means the same thing everywhere.

Scenarios

Ask for at least two emissions pathways. The usual pair today is SSP2-4.5 and SSP5-8.5 from CMIP6; older datasets use RCP4.5 and RCP8.5. Read SSP5-8.5 as a high-end stress test, not the expected outcome: scenario developers now consider emissions that high implausible. The pathways barely differ before mid-century and diverge after it, so the choice matters most for the 2080s. Some assessments now use global warming levels (2 °C, 3 °C) instead of pathways, which makes results easier to compare across studies. Either is fine; what matters is that the brief says which, and why.

Models and ensembles

Never rely on a single climate model. Ask for an ensemble, with the median and a range (typically the 10th to 90th percentile) for every index and horizon. For Canada, the downscaled and bias-corrected CMIP6 datasets on ClimateData.ca are the standard starting point; NEX-GDDP-CMIP6 and regional climate models such as NA-CORDEX (built on the earlier CMIP5 models) serve where more variables (wind, humidity, radiation) are needed, or outside Canada. Whatever the source, the data should be bias-corrected against observations or a reanalysis, and the brief should name the dataset so the work can be checked and repeated.

Resolution and extremes

Gridded projections at about 10 km are adequate for means and most indices, but not for everything. Sub-daily rainfall is not in the gridded sets: for IDF curves, ClimateData.ca provides climate change-scaled values at Environment and Climate Change Canada’s IDF stations, and sites away from a station need a dedicated analysis. Freezing rain and ice accretion, permafrost temperature and lake levels are not in the standard sets and need dedicated methods or published regional studies. If those hazards matter to your assets, say so in the brief, and expect the analysis to state its confidence in both the direction and the size of the change.

Six things to put in the brief

  1. The baseline period and the future horizons, matched to asset lives
  2. The emissions pathways or warming levels, at least two
  3. The dataset and the number of models in the ensemble
  4. The indices, each with the threshold it tests
  5. The uncertainty you expect to see: ranges across models, and confidence in direction and magnitude
  6. The data and the method delivered with the report, in open formats, so your team can rerun and update them

Scoped this way, the climate analysis answers the questions the assessment will ask, and the numbers that drive each risk score can be traced back to the data behind them.

Sources

Discuss your project

Start with a 30-minute call with a climate scientist about your question, your data and your timeline.