Project Results

Our team has been involved in the "Taiwan-Philippines (PH) Collaborative Project – Taiwan-Philippines VOTE Project: Improvement of Severe Weather, Marine Meteorology, and Short-term Climate Forecasting Capability - Phase I (2016-2020)" since 2016. The second phase of the VOTE project began in December 2020 and is scheduled for three years (2020-2023). Various research activities have progressed smoothly according to the plan. The main achievements are as follows:


1. Building upon the fuzzy logic classification algorithm framework developed in the first phase of the Taiwan-Philippines Bilateral Cooperation Research Project (VOTE I), this research improved the existing clutter echo fuzzy logic algorithm and expanded its application to marine echoes. This led to the development of a more comprehensive radar observation data quality control technique. The developed algorithm effectively removes common clutter echoes in low-elevation radar observations along coastal areas while retaining weather echoes.

2. The CWB (Central Weather Bureau) assisted the PAGASA (Philippine Atmospheric, Geophysical and Astronomical Services Administration) numerical prediction team in developing a data assimilation system and initiated the construction of the PAGASA-WRF background error covariance matrix (BEC). CWB provided PAGASA with a newer version of the BEC module and program, along with recommendations for setting BEC parameters in the data assimilation module.

3. PAGASA showed high interest in the typhoon intensity forecasting guidance derived from the WAIP (Wind-Pressure Integrated) technique, which was transferred from the early stages of the VOTE project. PAGASA has started applying the guidance in operational forecasting (PISTON). However, data import issues hindered the integration of observations and forecast guidance into PISTON. The CWB will arrange technical meetings to assist PAGASA in resolving data import issues.

4. Wave simulations and satellite observations were compared for different wind field data, using CFSR, NCEP combined with CWB forecast wind fields, and WRF as atmospheric wind inputs. The analysis involved calculating BIAS, RMSE, and NRMSE. In the waters around Taiwan and the Philippines, the main errors occurred in the Taiwan Strait, the Sulu Sea, and the area east of 130 degrees longitude and between 20 and 40 degrees latitude. The WRF wind field simulation results exhibited better performance, but all three showed high errors in the Sulu Sea, likely due to the influence of land on satellite observations between islands, leading to larger differences in the calculated results in the waters between the Philippine archipelago.

5.  Using observations from a disdrometer, electromagnetic wave scattering simulations were conducted to evaluate the dual-wavelength radar echoes in the X and K bands. The study assessed the error in estimating liquid water content and analyzed the relationship between the error and the median volume diameter of raindrop size distribution. The results showed that the average root mean square error in LWC estimation for the four echo combinations ranged from 0.11 to 0.17 g/m3. Regions with larger median volume diameters exhibited larger estimation errors, partly due to the attenuation effects in the X-band, which could be reduced through attenuation correction to mitigate estimation errors.

6.  Quantitative precipitation forecast evaluation and verification were conducted for three typhoons affecting the Philippines: Mangkhut (2018), Goni (2015), and Melor (2015) using cloud-resolving ensemble forecasts. The results showed that as the forecast initial time approached the actual observation time, the errors in typhoon path, maximum wind speed, and minimum pressure decreased. Moreover, the forecast path errors stayed below 150 kilometers within two days before the target rainfall period. In terms of rainfall, the intensity and spatial distribution of rainfall became closer to the observed results as the forecast time approached the target rainfall period.

7. This study clarified the relationship between typhoon paths affecting northern Philippines (15°N-19°N) and the 30-60 day intraseasonal oscillation (ISO) during June-October. Based on the relationship between northern Philippines and ISO phase correspondence, the study explored the main paths and occurrence frequencies of typhoons affecting northern Philippines and the mechanisms influencing the results. Statistical results of ISO phase and typhoons affecting northern Philippines showed two main typhoon paths corresponding to northwestward and northward turns, with corresponding ISO phases classified into phases 3-5 and phases 6-8.

8. The association between the Western North Pacific Monsoon Index (WNPMI) and typhoon forecasting techniques was tested. The dates of more pronounced monsoon troughs did not follow a regular pattern, and there was no clear interannual variation in the strength of WNPMI. Therefore, based on the cumulative percentile of WNPMI values, WNPMI was divided into five intervals. The analysis of WNPMI and the actual number of typhoon formations using historical reforecasts from GEFSv12 revealed that as WNPMI increased, the number of typhoon formations indeed increased, and when WNPMI was below average, the number of typhoon formations tended to be lower.

In summary, this project has successfully achieved its research objectives and produced multiple outcomes. In the future, based on the research foundation of this project, we will further strengthen communication between Taiwan and the Philippines. Our team will continue to provide education and training to our Philippine counterparts, hold bilateral cooperative seminars annually to enhance typhoon forecasting techniques for both parties, and work towards the common goal of disaster reduction and prevention in Taiwan and the Philippines.



Figure 1. The 1-hour rainfall estimate field obtained by the Z-R (reflectivity-rainfall) estimation formula (Z = 32.5 R^1.65) observed by the Aparri Radar in the Philippines on September 14, 2018, at 1800 UTC during Typhoon Mangkhut (a) without the removal of sea clutter echoes using the fuzzy logic algorithm and (b) with the removal of sea clutter echoes using the fuzzy logic algorithm. The red dot in the figure indicates the location of the Aparri Radar. 

Figure 2. PAGASA-WRF Domain. The left side illustrates the operational mode, while the right side depicts the data assimilation experimental framework 

Figure 3. The architecture illustrating how the TAFIS API supports the PISTON system.  

Figure 4. Wave height variations influenced by currents (2018/08/15 00:00) (Top Left: Wave height and direction without current effect, Top Right: Wave height and direction with current effect, Bottom Left: Ocean current model flow field, Bottom Right: Wave field and direction changes with current effect). 

Figure 5-1, Experimental Process 

Figure 5-2, (a) D0 field arranged and (b) X-band echo field according to D0 arrangement 

Figure 5-3, Liquid Water Content fields, (a) True values calculated after gamma fitting, and (b) to (e) estimated values 

圖6. The schematic diagram summarizes the forecast data evaluated in this study. The seasonal forecast products include the NMME seasonal predictions and WMO Global Seasonal Climate Update (maps not shown). The S2S forecast products include ECMWF and NCEP models that generate the forecast at least once a week. The forecast frequency and ensemble members of each model are noted.

Typhoon Extended Forecast Product

Figure 7. Typhoon Extended Forecast Product by the Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA).

http://bagong.pagasa.dost.gov.ph/tropical-cyclone/tc-threat-potential-forecast