Closed chrisroadmap closed 1 year ago
Regression based on AR6 WG3 has been done (versus CO2 FFI and period).
Slight issue in that 2020 value from regression does not quite match obs (though, obs are very uncertain). Eleanor suggests only using IMAGE scenarios.
To document here, I also tried looking at MAC curves from AR6 - they are too steep for DICE to handle in the optimizer (and quite a lot of scatter between scenarios). I also looked approximating a quadratic MAC based on this paper - but it made one "optimal" ensemble member crash, so it wasn't pursued,
Updated regression works well - don't know why I suggested IMAGE only now