A Revised Snow Scheme and Its Performance in the Next Seasonal Prediction System at JMA (CPS4)

The Japan Meteorological Agency (JMA) operates the Coupled Prediction System version 3 (CPS3; Hirahara et al. 2023) to support a wide range of seasonal forecasts. CPS3 consists of an atmosphere/land surface/ocean/sea-ice-coupled model; the atmospheric and land surface model has a horizontal resolution of TL319 (approx. 55 km) with 100 vertical layers (up to 0.01 hPa), and the ocean and sea-ice model has a horizontal resolution of 0.25°x0.25° with 60 vertical layers. Various improvements implemented in 2022 led to significantly better performance of seasonal prediction than the previous system. However, there is still room for improvement, e.g. systematic biases of near-surface air temperature over land. We are working on the next seasonal prediction system at JMA (CPS4) to obtain further improvements in the prediction skills.

In terms of snow-covered areas in the Northern Hemisphere, CPS3 has systematic negative and positive biases in surface albedo near snow-free areas for winter and spring, respectively. They relate to systematic positive and negative biases of near-surface air temperature over the areas. In order to address the biases, snow-related parametrization in the land surface model has been revised. The current snow cover fraction (SCF) parametrization represents less SCF for shallow snow and larger SCF for melting snow than the observed relationship. The revised SCF parametrization depending on snow depth and snow density can represent a better relationship for SCF and reduce the biases in surface albedo. Combination of the revisions including modified snow albedo, snow density and rain-snow partitioning scheme reduces excess snow depth and improves systematic biases in SCF and surface albedo. As a result, prediction skills are improved for not only temperature in the lower troposphere but the westerly jet stream in East Asia. We will present the performance of the revised snow scheme and its impacts on seasonal prediction in CPS4.