Dear Editor,
On behalf of the Korean members of the US National Weather Service (NWS), I would like to describe the way in which operational forecasts are produced in the NWS, and to offer an opinion on the flooding issue in Korea.
We have seen many comments in the Korean news media that the KMA needs to have satellites, radars and many observing stations. Of course, they are important to increase the knowledge of the current weather and hence to improve nowcasting. However, the scope in which these remotely-sensed data are directly used for weather forecasting is very limited (at least for now). In principle, heavy rain can be produced within 2 hours in clear sky, sometimes less than 30 min. Obviously, the current NWP models cannot handle events at such a fine scale. Beyond that scale, the forecast comes almost exclusively from the numerical weather prediction (NWP) and its statistical post-processor, model output statistics (MOS). In other words, what you see in the weather channels are based entirely on NWP models, except for the description of the current weather.
Precise monitoring of current weather using data from radar, satellite, and surface stations can help understand the current weather and allows, together with NWP forecasts, nowcasting up to 6 hours (mostly 1-3 hours). Again, forecasts beyond that temporal scale depend entirely upon the NWP.Remotely-sensed observations from satellete and radar have been used in research to better-understand the mesoscale structure of severe weather events. Knowledge from these observations provides some insights into model physics in th NWP models, which in turn contribute to improve weather forecasts. In my opinion, however, the use of remotely sensed data in the NWP has a long way to go: it is not only a matter of data quality but also a problem with data assimilation which requires technological and theoretical advances/breakthroughs. For example, the NEXRAD data in the U.S. are yet to be used in operational NWP (at NCEP, they have been used only experimentally from 5 years ago).
At the NCEP, predictability of precipitation is about 30%. We evaluate both the amount and timing. Suppose that there was precipitation of 100 mm in Seoul and a model forcasted 100 mm in Suwon. Then, the forecast is wrong, and we have zero predictability. Likewise, suppose that the model forecasted the same amount in Seoul but with a few hours delay. The forecast is wrong again, and we again have zero predictability. We evaluate precipitation forecast 24h-48h in advance. Note that, if one applies the above scoring system, a 30% skill is not too bad. If one looks at the total precipipation for the day, the model forecast and the observation are remarkably comparable in general. Some Korean news media said that the KMA has forecast accuracy of 83% and is close to the US (85% as they say). I don't have an idea on what these values mean. In terms of anomaly correlation for 5-day forecast, the NCEP global model has a predictability ranging 70-75% (summer) to 80-85% (winter). Predictability of synoptic pattern ranges over 90% for day 1 and day 2 forecasts.
The difficulty in forecasting heavy rainfall in Korea can be in line with thundrstorm forecasts over the US. In the US, predictability (of severe weathers) is relatively high (in my opinion) since the scientific community has been working on the problem for so long times (including numerous field experiments). Accordingly the resulting NWP models (mesoscale and global models) are tailored for weather systems in this country. On the other hand, the scientific community in the US for obvious reasons is not terribly interested in weather systems in other countries. As pointed out by some Korean papers, physical mechanisms for heavy rain in Korea substantially differ from those for MCCs in the US. Convective precipitation physics in the mesoscale models of the US have not proven to be successful in reproducing heavy rains over Korea. A number of different alternative approaches, however, do exist (refer to some Korean jounals on heavy rain). I must point out, however, that observations must accompany modeling: without observations, all the theory and model improvements are just guesses.
Forecasters at the NCEP not only look at the precipitation forecasts (indeed this is only a piece of information) but also have good insight into the model atmosphere. The way they issue severe weather forecasts is as follows; watch the model output if there is a signal as early as 2 days in advance. Then, keep watching the latest model output and update the forecast. If the forecast consistently indicates severe weather down to the lead time of 12 hours, appropriate (evacuation, if deemed necessary) actions are initiated in collaboration with the FEMA. (Note that no radar and satellite data are used to initiate these actions.) As an example, there was an extremely heavy rainfall (in excess of 500 mm in Virginia) a few years ago. The sky was clear in the morning when many residents were evacuated: heavy rain hit that night. This is an example of a mesoscale convective system embedded within a synoptic-scale cyclone, but the predictability for a locally generated convective system accompanying severe weathers is still far behind. For example, the typical lead time for tornado outbreak is less than 10 min, which entirely depends upon nowcasting.
Given that the degree of difficulty associated with forecasting heavy rainfall in Korea is comparable to that with forecasting severe thunderstorms or tornadoes in the US, a large room for improvement exists in the NWP operations of the KMA. The lack of progress is not surprising since no field experiment has been carried out. Also, not enough attention has been paid to the problem by the Korean scientific community. Basic research through special field experiments is crucial (KORMEX is encouraging in this regard). Let me reiterate that nobody outside Korea is interested in the Korean weather. Also, heavy rain in Korea is different from that in the rest of the world. In this regard, heavy precipitation is different from other meteorological variables: for example, the anomalous climate change has been studied by many scientists all over the world.
Song-You Hong, Environmental Modeling Center/NCEP, National Weather Service, NOAA Dr. Song-You Hong is also a member of the KASA (Korean Atmospheric Scientists in America).