Tropical cyclone precipitation in the S2S models: climatology and skill

Tropical cyclone precipitation (TCP) can contribute up to 40% of the total annual rainfall in tropical regions whereas heavy TCP over coastal regions can cause high-impact extreme events such as floods and landslides. Skillful prediction of TCP at sub-seasonal timescale can potentially increase response time for emergency preparation and planning.

Here, we assess the climatological representation of TCP by sub-seasonal to seasonal (S2S) forecast models by comparing the mean spatial distribution, seasonality and azimuthally averaged TCP amounts with observations. This analysis of the TCP climatology highlights that model biases in the TC track density are strongly related to biases in the spatial distribution of TCP. In addition, most models overestimate the mean precipitation rate in the inner-core. Model skill is first evaluated for cases in which the models accurately forecast the genesis of a cyclone at 2-3 days lead time or in which the storm was active at the initialization date using verification methods similar to those used for numerical weather prediction models. For sub-seasonal lead-times (week 1 and longer), the contribution of TCP to extreme precipitation climatology and skill scores is quantified to evaluate the role of TCP for the mean representation and forecasts of extreme events in these models.