Probabilistic assessment with the Hector simple climate model

I’m performing calibration of Hector using Markov Chain Monte Carlo. Uncertain model parameters are tweaked until settling in a range that matches observations fairly well. Then, we can use those parameter ranges to create robust probabilistic hindcasts and projections of the model’s outputs.

Project code is on github

An example using this technique for temperature:

Here are two probabilistic projections for temperature under RCP8.5 (a business-as-usual greenhouse gas emissions scenario). In red, we include ocean heat, temperature, and sea level rise (SLR) observations as calibration constraints. In blue, we leave SLR out of the calibration. The shaded ranges are 95% credible intervals, the solid colored lines are the median projections. The black range to the right is the range from CMIP5.

Previous Research Project:

Using a large climate model ensemble to investigate El Nino variability in a changing climate (Nature Scientific Reports, 2017)