発表要旨
1. プロジェクトの目的
Produce long-term (50+years) high-resolution (10km horizontal) hourly dynamically downscaled climate analysis for the contiguous United States for global climate change studies.
2. 今年度当初の計画
We had finished optimizing the code for the production run on the Earth Simulator in 2005. Therefore our goal for 2006 was to compute 30 years of downscaled analysis. As the objective of the project is to produce multi-decadal dataset for long-term climate studies, the completion of the computation is essential to the success of the study. Only when several decades of downscaled analysis are available, we will be able to study climatology and decadal changes of climate at regional scales.
Another goal is to conduct extensive validation of California Reanalysis Downscaling at 10km (CaRD10) which was produced with the same dynamical downscaling system as the current project over the U.S. We expect the U.S. downscaling will bear the same quality as CaRD10 upon completion of the computation. Our plan for 2006 was to compare CaRD10 with station observations and North American Regional Reanalysis.
Third, we planned to look at the first few years of U.S. downscaled analysis that we managed to compute on the ES. Even though the currently completed length of downscaled analysis is too short for validation in climatological sense, we wanted to ensure that meso-scale climate phenomena are reasonably produced in the downscaling, as the first stage of validation.
3. 今年度得られた成果、および達成度
With the limited computing resources granted to our project, we were able to compute only 8 years of downscaling. We hope to obtain more resources in the next year to complete the computation as soon as possible.
For the second goal, we successfully completed the validation study of CaRD10. In general, the CaRD10 near-surface wind and temperature fit better with regional scale station observations than the NCEP/NCAR reanalysis used to force the regional model, supporting the premise that the regional downscaling is a viable method to attain regional detail from large scale analysis. This advantage of CaRD10 was found on all time scales, ranging from hourly to decadal scales, i.e. from diurnal variation to multi-decadal trend. CaRD10 is comparable in quality to NARR which is a more computationally expensive regional data assimilation product. CaRD10 is superior to NARR, even without direct input of observations, in many aspects where higher horizontal resolution is important.
We performed validation of meso-scale events from a few years of the U.S. downscaling. The downscaling is able to reproduce many key meso-scale phenomena - the Great Plains low level jets, southwestern monsoon, mountain/valley winds, Chinook winds, hurricanes, superstorms, Central America gap flows, and lake effect snow, among others. This encouraging result shows that the downscaled product will be very useful for studying the relationship between weather and climate.