**4. 2012, 2013 cases: forecasting and discussion**

**Figure 4.** Boxplots for R24 and R72 rainfall at SMS from 1953 to 2011 due to TTYs (a and b), TTSs (c and d), and TDs (e and f). The vertical black thick line in the middle of each subplot denotes SMS. A on the *x*-label refers to the landfalling area, which is to the west of SMS, and B on the *x*-label refers to the landfalling area, which is to the east of SMS. The

From **Figure 4**, it can be seen that the shape of the rainfall boxplots is not symmetric compared to the boxplots on the west and on the east of SMS. Generally, the TCs that make landfall on the west of SMS within the distance of 400 km would generally bring more rainfall to SMS than those TCs that make landfall on the east of SMS with similar landfalling intensity and distance. The median (Q2) records of R24 and R72 for TTY and TTS (**Figure 4a**–**d**) generally decrease with the increase of the landfalling distance to SMS, especially for TCs landfalling on the west of SMS. However, when the landfalling distance is outside 400 km, Q2 does not decrease as clearly as its change within 400 km. From **Figure 4a**–**d**, it can be seen that TCs that make landfall at B5 (with the distance of 400–500 km on the east of SMS) and B6 (with the distance of 500– 600 km on the east of SMS) sometimes might cause very heavy rainfall at SMS. The reason might be that after those TCs land at B5-B6, they do not dissipate soon and continue to move west or east. During their continuous movement, they will probably interact with other weather system, such as southwesterly monsoon and mid-latitude trough, and bring plenty of moisture and energy over Southeast China. If SMS is on the passageway or near the passageway to transport such moisture and energy, heavy rainfall would occur in Shenzhen.

numbers after A and B on the *x*-label refer to the distance to SMS.

224 Recent Developments in Tropical Cyclone Dynamics, Prediction, and Detection

To apply the statistical boxplot scheme to forecast the potential maximum R24 and R72 rainfall due to a landfalling TC, the TC-landfalling location and intensity need to be identified first. Nowadays, operational forecasting models can usually predict the TC's track well, especially the track within 24 h [15]. Compared to the track forecast, the TC intensity forecast by NWP models is not so accurate at present. However, with the real-time observations from satellite and radar, forecasters have the ability to predict the intensity of the landfalling TC 12 h before TC landfall by rule of thumb, as it usually takes at least 12 h for a TC to change intensity or motion appreciably [1]. With the approximate TC-landfalling location and intensity available 12 h before the TC landfall, the statistical boxplot scheme can be then applied to forecast the rainfall of R24 and R72 at SMS.

In this section, the TCs that made landfall along the Southeast China coast in 2012 and 2013 will be used to test the performance of the boxplot scheme. In the typhoon seasons of these 2 years, there are 11 TCs landfalling within the distance of 700 km to SMS along the Southeast China coast. The detailed TC information is summarized in **Table 3**.

Based on the latest forecast for these 11 TCs' landfalling direction, location, and intensity, the boxplots for the corresponding TC categories are picked out. **Figure 5** shows these boxplots for the rainfall of R24, as well as the real maximum daily rainfall observation at SMS (black thick solid horizontal line) during the TC-landfalling period (rainfall within a couple of days before and after TC landfalling) due to these TCs. By comparison, the latest rainfall forecasts (brown solid horizontal line) before the TCs' landfall by European Center for Medium-Range Weather Forecasts (ECMWF) are plotted in **Figure 5** as well. ECMWF is renowned worldwide for providing the most accurate medium-range global weather forecasts up to 10 days ahead, monthly forecasts, and seasonal outlooks to 6 months ahead. It has been widely used for operational forecast and research purpose around the world [28–31]. ECMWF model is reported to well produce the medium-range forecasts of the Northwest Pacific subtropical high and South Asian high which have pronounced influences on the summertime persistent heavy rainfall in China [32]. The track of the TC in western Pacific and South China Sea is strongly affected by the area of the western Pacific subtropical high [33, 34]. Therefore, ECMWF's forecasts of TC's track for the coming 12 and 24 h are usually reliable. SZMB has been relying on ECMWF for daily operational forecast since 2009. The finest resolution of ECMWF system used in SZMB is 0.125° × 0.125°. Xu et al. [28] reported that ECMWF model could provide valuable information for rainfall forecast up to the next 10 days; however, the ability for storm rainfall forecast was not reliable. For tropical cyclone forecast, ECMWF can usually accurately predict the TC's landfalling intensity and location by the latest forecast around 12 h before TC's landfall, but it cannot predict well the storm rainfall induced by TCs.


Time format for landfalling time: month/day/hour; distance refers to the distance between the landfalling center of TC and SMS. The unit for latitude is degree north (°N) and for longitude is degree east (°E). The unit for distance is kilometer (km).

**Table 3.** Information about the 11 TCs landfalling within the distance of 700 km to SMS in 2012 (the first three TCs) and 2013 (the last eight TCs).

From **Figure 5**, it can be seen that the observed R24 rainfall records at SZMB for the 11 TCs are most of the time between the historical observed maximum and minimum rainfall for the corresponding category of landfalling TC, except for TY Trami and STY Usagi. TY Trami landed in Fujian province, China on August 22, 2013. It continued to move west to Jiangxi Province, until it finally disappeared in Hunan Province. During its process in mainland, Trami interacted with the strong southwest monsoon and brought immense downpours in Southeast China. It set a new rainfall record at Shenzhen for TTY category, which landed on the east of Shenzhen with the landfalling distance of 600–700 km. Similar condition is for STY Usagi. Usagi set a new rainfall record at Shenzhen for TTY category, which landed on the east of Shenzhen with the landfalling distance of 200–300 km. For the other TCs, most of the rainfall observations at SMS are within the interquartile range of the historical records (Q1–Q3), such as TS Doksuri, TY Vicente, TY Kai-Tak, TD1303, STS Rumbia, STS Jebi, and STY Utor, which are shown in An Operational Statistical Scheme for Tropical Cyclone-Induced Rainfall Forecast http://dx.doi.org/10.5772/64859 227

forecasts of TC's track for the coming 12 and 24 h are usually reliable. SZMB has been relying on ECMWF for daily operational forecast since 2009. The finest resolution of ECMWF system used in SZMB is 0.125° × 0.125°. Xu et al. [28] reported that ECMWF model could provide valuable information for rainfall forecast up to the next 10 days; however, the ability for storm rainfall forecast was not reliable. For tropical cyclone forecast, ECMWF can usually accurately predict the TC's landfalling intensity and location by the latest forecast around 12 h before TC's

**(°E)**

Time format for landfalling time: month/day/hour; distance refers to the distance between the landfalling center of TC and SMS. The unit for latitude is degree north (°N) and for longitude is degree east (°E). The unit for distance is

**Table 3.** Information about the 11 TCs landfalling within the distance of 700 km to SMS in 2012 (the first three TCs) and

From **Figure 5**, it can be seen that the observed R24 rainfall records at SZMB for the 11 TCs are most of the time between the historical observed maximum and minimum rainfall for the corresponding category of landfalling TC, except for TY Trami and STY Usagi. TY Trami landed in Fujian province, China on August 22, 2013. It continued to move west to Jiangxi Province, until it finally disappeared in Hunan Province. During its process in mainland, Trami interacted with the strong southwest monsoon and brought immense downpours in Southeast China. It set a new rainfall record at Shenzhen for TTY category, which landed on the east of Shenzhen with the landfalling distance of 600–700 km. Similar condition is for STY Usagi. Usagi set a new rainfall record at Shenzhen for TTY category, which landed on the east of Shenzhen with the landfalling distance of 200–300 km. For the other TCs, most of the rainfall observations at SMS are within the interquartile range of the historical records (Q1–Q3), such as TS Doksuri, TY Vicente, TY Kai-Tak, TD1303, STS Rumbia, STS Jebi, and STY Utor, which are shown in

Doksuri 6/30/03 22 113.2 TTS 102 Vicente 7/24/04 22 113 TTY 119 Kai-Tak 8/17/12 21 110.4 TTY 410 TD1303 6/15/17 19.9 110.9 TD 436 Bebinca 6/22/11 19.2 110.7 TTS 506 Rumbia 7/02/06 21.1 110.2 TTS 424 Cimaron 7/18/22 24.1 117.9 TTS 434 Jebi 8/02/20 19.7 110.9 TTS 451 Utor 8/14/15 21.6 111.9 TTY 241 Trami 8/22/03 25.7 119.5 TTY 659 Usagi 9/22/20 22.7 115.4 TTY 263

**Landfalling longitude**

**Landfalling intensity**

**Distance (km)**

landfall, but it cannot predict well the storm rainfall induced by TCs.

**Landfalling latitude**

226 Recent Developments in Tropical Cyclone Dynamics, Prediction, and Detection

**(°N)**

**Name Landfalling time**

kilometer (km).

2013 (the last eight TCs).

**Figure 5.** Comparison of the R24 rainfall forecasts for 11 TC cases in 2012 and 2013 by the boxplots and by ECMWF (brown solid horizontal line), as well as the real rainfall observation (black thick solid horizontal line): (a) for TS Doksuri; (b) for TY Vicente; (c) for TY Kai-Tak; (d) for TD1303; (e) for TS Bebinca; (f) for STS Rumbia; (g) for STS Cimaron; (h) for STS Jebi; (i) for STY Utor; (j) for TY Trami; and (k) for STY Usagi.

**Figure 5a**–**d,f,h,i**, respectively. Similarly, more than 50% of R72 rainfall at SMS due to these TCs are within the interquartile range (Q1–Q3) of the historical records. By comparison, it can be seen that ECMWF by the latest model output can sometime accurately forecast the rainfall due to TCs, such as in **Figure 5f** and **g**. However, for the other cases, the discrepancies between the rainfall forecast by ECMWF and real rainfall observation at SMS are large. Therefore, besides the NWP, forecasters can use the historical rainfall observation boxplots (**Figure 4**) as a good reference to predict the potential short-term rainfall due to a landfalling TC.

The statistical boxplots scheme can provide valuable information to operational forecaster to predict the potential rainfall due to a TC. However, it must be known that there are still many uncertainties for the boxplots because of the small sample size for some TC categories due to the natural features, as well as the short observation history. With more TCs landfalling along the Southeast China coast in future, the larger will be the database of TCs and the more accurate will be for the short-term TC rainfall prediction by the boxplots scheme.
