**5. TC precipitation**

One of the typical phenomena of TC activities is heavy rainfall, which is also one of the most significant impacts of TCs. The tremendous precipitation from TCs often leads to loss of lives and properties. Flooding from landfall TCs over United States is the leading cause of death related to severe storms [70]. However, TC precipitation can also bring in major economic benefits to areas surrounding its path. Based on the long‐term TRMM rainfall measurements, over 84% of continental convective rainfall is contributed from rain intensity > 5 mm h‐1 [71]. Analysis of the numerical weather prediction (NWP) model rainfall forecasts indicates TC precipitation could contribute 15–17% of the total annual rainfall over broad latitude zones [72]. TC rainfall can contribute up to 15% of total precipitation over a hurricane season in Carolinas of United States [73]. Therefore, even precipitation from one TC activities could ease the stress of drought over some areas. A good review of TC rainfall's structure, intensity, and forecasts was recently reported [21].

such as 37 GHz principally showing liquid hydrometeor emissions near and below the freezing level. Both of these previously mentioned channels are measured at horizontal (H) and vertical (V) polarizations. Near the interface of the outer TC and the environment, interpretations at either polarization become muddled due to multiple competing influences (e.g., water vapor, cloud water, and sea surface). The polarization correction temperature (PCT) can improve the representation of atmospheric features, allowing them to stand out from surface background

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

Despite its relatively new arrival, some progresses are apparent in using microwave imagers to examine tropical cyclones. For example, the NHC extensively uses microwave imagery to better locate a TC center and subjectively diagnose changes in structure [32]. The Morphed Integrated Microwave Imagery at CIMSS (MIMIC), a technique to create "morphed" anima‐ tions of passive microwave imagery using an advection function between satellite passes, was introduced to allow a visually appealing depiction of TC structure changes [61]. Other studies also revealed relationship between microwave imager data to TC intensity [62, 63]. The microwave data have been used to improve TC intensity estimates through early detection of a forming eyewall [64], while a color composite of the H‐pol, V‐pol, and PCT data at 37 GHz developed at NRL‐MRY has shown particular promise in diagnosing TC inner core formation [39]. A symmetrical and closed TB threshold ("cyan ring") was applied to predict the TC onset

Some efforts focused on cataloging TCs through an extended climatology of microwave imagers. The microwave data interpolated onto an 8‐km grid in the hurricane satellite (HURSAT) archive was created in 2008 [65]. The HURSAT‐microwave consists of data from the SSM/I platforms between 1987 and 2009, using global best track data from the International Best Track Archive for Climate Stewardship (IBTrACS) to search for TCs [24]. Based on this dataset, the TCs composited by their intensification rate and environmental wind shear were analyzed to compare microwave signatures during different intensity regimes [66]. Recently, a new study on eyewall size estimates using the HURSAT‐microwave data compared to the aircraft reconnaissance measurements demonstrates the similarity of in‐situ and satellite‐ derived structural profiles [67]. A more advanced TC PMW TB database at 1 km spatial resolution has been developed at NRL‐MRY from all available PMW sensors in 1987–2012 with more consistent inter‐calibrated TBs at 89 GHz, better eye fixing, and high quality interpolation

One of the typical phenomena of TC activities is heavy rainfall, which is also one of the most significant impacts of TCs. The tremendous precipitation from TCs often leads to loss of lives and properties. Flooding from landfall TCs over United States is the leading cause of death related to severe storms [70]. However, TC precipitation can also bring in major economic benefits to areas surrounding its path. Based on the long‐term TRMM rainfall measurements, over 84% of continental convective rainfall is contributed from rain intensity > 5 mm h‐1 [71].

[60].

rapid intensification [44].

scheme [68, 69].

**5. TC precipitation**

Rain retrievals from advanced algorithms based on PMW measurements over ocean have been proven accurate and reliable [74–77]. The satellite‐derived instantaneous rain patterns over TCs clearly show the heavy rainfall is normally located in TC eyewall and spiral convective areas. In general, the intensity and pattern of TC precipitation are strongly associated with the TC intensity and radial distance to eyewall [78]. The maximum rainfall appears in the TC eyewall around less than 50 km radii and the rainfall intensity rapidly reduces with increase of its radii. Rain intensity is about 13 mm h‐1 for major TCs (category 3–5), 7 mm h‐1 for minor TCs (category 1–2), and 3 mm h‐1 for tropical storms. By the radii of 300–350 km, rain intensity for all kinds of TCs is almost same.

**Figure 9.** Rainfall asymmetry calculated in 10‐km rings around the storm center, as a function of storm intensity: (a) 2121 TC observations (total distribution), (b) TS, (c) CAT12, (d) CAT35. The storm motion vector is aligned with the positive *y* axis. The color scale indicates the amplitude of the normalized asymmetry. Red corresponds to the maxi‐ mum positive anomaly and blue to the minimum rainfall within the storm (adapted from Lonfat et al. (2004). ©Ameri‐ can Meteorological Society. Used with permission).

The asymmetry of TC precipitation is a prominent feature. It shows different characteristics depending on what matric is applied. **Figure 9** presents the TC rainfall asymmetry patterns relative to its motion direction as a function of storm intensity based on 3 years' TRMM TMI rain retrievals [78]. For all storms and tropical storms, their maximum rain intensity is in the front quadrants of TC movement. The location of maximum rain intensity shifts from the front‐ left for CAT1‐2 to front‐right quadrants for CAT3‐5. Thus, the asymmetry of TC rainfall is linked with the TC intensity, especially for strong TCs. In addition, the asymmetry has a property of strong dependence on TC geographic locations. Maximum rainfall appears in front quadrants over WP while in front‐right quadrants over AT. Maximum rainfall shifts to the front‐left quadrants over SH. Over EP and IO, it is located in the front quadrants with a cyclonical pattern.

A recent study based on TRMM rain datasets for landfall TCs over different parts of China presents various rainfall patterns relative to TC's motion at different times of landfall [79]. Maximum rainfall is located in the left quadrants for TCs landed in Guangdong province and Taiwan, while in the front‐left quadrants for TCs landed in Hainan and Fujian provinces. Maximum rainfall is generally located in the back‐right quadrants for TCs landed in Zhejiang province. However, maximum rainfall is generally positioned in the front quadrants of TCs relative to its vertical wind shear vector (**Figure 10**), although there is still a slight difference in rainfall distribution between different areas. This feature is an important finding because it has a potential application for improving TC rainfall forecasts.

What are the possible causes for the asymmetric distribution of TC precipitation? Several known factors are associated with this asymmetry feature, such as the advection of planetary vorticity, vertical wind shear, and friction‐induced boundary layer convergence [80, 81]. The maximum rainfall in the front‐quadrants of TC motion indicates the role of friction‐induced boundary layer convergence. Higher TC moving speed leads to strong rain intensity in its front quadrants [78], while its dependence on geographic locations shows importance of the TC's ambient wind influence. The improved consistence of rainfall asymmetry relative to its vertical wind shear for the landfall TCs over China further indicates the role of interaction between TC and its environmental forcing. However, this feature needs more verification studies over other TC basins and its connections to amplitudes of the vertical wind shear.

A better prediction of the TC rainfall distribution is the ultimate goal of efforts in mitigating TC's rainfall impacts on society, life, and property. Several methods have been developed for operational TC rainfall forecasts. A popular one is the Tropical Rainfall Potential (TRaP) which is based on the satellite rainfall estimates, persistence of TC intensity, and satellite‐derived wind vector [19, 82].

$$\mathbf{T} \mathbf{R} \mathbf{a} \mathbf{P} = R\_w \cdot D \cdot V^{-1} \tag{1}$$

where *R*av is the mean rainrate along a line in the direction of TC motion, *D* is the distance of that line across the TC rain area, and *V* is the TC's actual speed. Accurate satellite rain retrievals and satellite wind vectors as well as its easy implementation have made this method popular

Satellite Remote Sensing of Tropical Cyclones http://dx.doi.org/10.5772/64114 151

The asymmetry of TC precipitation is a prominent feature. It shows different characteristics depending on what matric is applied. **Figure 9** presents the TC rainfall asymmetry patterns relative to its motion direction as a function of storm intensity based on 3 years' TRMM TMI rain retrievals [78]. For all storms and tropical storms, their maximum rain intensity is in the front quadrants of TC movement. The location of maximum rain intensity shifts from the front‐ left for CAT1‐2 to front‐right quadrants for CAT3‐5. Thus, the asymmetry of TC rainfall is linked with the TC intensity, especially for strong TCs. In addition, the asymmetry has a property of strong dependence on TC geographic locations. Maximum rainfall appears in front quadrants over WP while in front‐right quadrants over AT. Maximum rainfall shifts to the front‐left quadrants over SH. Over EP and IO, it is located in the front quadrants with a

A recent study based on TRMM rain datasets for landfall TCs over different parts of China presents various rainfall patterns relative to TC's motion at different times of landfall [79]. Maximum rainfall is located in the left quadrants for TCs landed in Guangdong province and Taiwan, while in the front‐left quadrants for TCs landed in Hainan and Fujian provinces. Maximum rainfall is generally located in the back‐right quadrants for TCs landed in Zhejiang province. However, maximum rainfall is generally positioned in the front quadrants of TCs relative to its vertical wind shear vector (**Figure 10**), although there is still a slight difference in rainfall distribution between different areas. This feature is an important finding because it

What are the possible causes for the asymmetric distribution of TC precipitation? Several known factors are associated with this asymmetry feature, such as the advection of planetary vorticity, vertical wind shear, and friction‐induced boundary layer convergence [80, 81]. The maximum rainfall in the front‐quadrants of TC motion indicates the role of friction‐induced boundary layer convergence. Higher TC moving speed leads to strong rain intensity in its front quadrants [78], while its dependence on geographic locations shows importance of the TC's ambient wind influence. The improved consistence of rainfall asymmetry relative to its vertical wind shear for the landfall TCs over China further indicates the role of interaction between TC and its environmental forcing. However, this feature needs more verification studies over other

A better prediction of the TC rainfall distribution is the ultimate goal of efforts in mitigating TC's rainfall impacts on society, life, and property. Several methods have been developed for operational TC rainfall forecasts. A popular one is the Tropical Rainfall Potential (TRaP) which is based on the satellite rainfall estimates, persistence of TC intensity, and satellite‐derived

<sup>1</sup> TRaP *R DV* av

where *R*av is the mean rainrate along a line in the direction of TC motion, *D* is the distance of that line across the TC rain area, and *V* is the TC's actual speed. Accurate satellite rain retrievals and satellite wind vectors as well as its easy implementation have made this method popular


has a potential application for improving TC rainfall forecasts.

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

TC basins and its connections to amplitudes of the vertical wind shear.

cyclonical pattern.

wind vector [19, 82].

**Figure 10.** The wavenumber 1 rainfall asymmetry (mm) relative to the storm vertical wind shear. The shear vector is aligned with the positive *y* axis (upward). The *x* and *y* axes are distance (°) from the TC center (origins). Stage (I) is 24 hr prior to, stage (II) is at the time of, and stage (III) is 24 hr after landfall. The color scale indicates the amplitude of the asymmetry relative to the storm motion (adapted from Yu et al. (2014). ©American Meteorological Society. Used with permission).

in operations. A variation of this method called the areal TRaP has been introduced to graphic view of TC precipitation horizontal distributions [83, 84], with three correct assumptions of TC track forecast, satellite rain estimates, and persistent spatial pattern of rainrates relative to the TC center. TRaP is normally valid for short forecasts of less than 24 hr. However, there are limitations associated with TRaP because of no considerations on changes of TC intensity and the TC's vertical wind shear conditions.

Another method is the rainfall climatology and persistence (R‐CLIPER) model which is a parametric model utilized with the TC rainfall climatology from satellite measurements [85]. This method assumes a circularly symmetric distribution of rainfall and its rainfall is translated in time. Although it accounts for TC intensity and moving speed, it does not include the TC's unique asymmetry rainfall patterns. An improved method called the parametric hurricane rainfall model (PHRaM) was introduced to incorporate with the TC rainfall asymmetry feature by including the azimuthal Fourier decomposition for shear and a term indicating the topographical uplift [86]. Results show that PHRaM is improved significantly compared with the standard R‐CLIPER. A new parametric model was recently developed for including more factors such as TC motion speed and intensity, vertical wind shear, and typical features of the TC boundary layer [87].

All the above‐mentioned methods have one common assumption of the correct satellite rainfall retrievals. Although satellite‐derived precipitation from PMW sensors is very accurate over ocean, there are still relatively large errors over land [75, 88]. Some discrepancies exist among satellite rain datasets, especially with different satellite sensors and retrieval algorithms. In order to minimize errors from different rain retrieval algorithms for different sensors, the physical‐based inversion rain algorithm (GPROF) used in TRMM and GPM is also applied for other PMW sensors [77]. Thus, precipitation from different PMW sensors will be more consistent. The new NASA integrated multi‐satellite retrievals for GPM (IMERG) is based on these consistent PMW rain retrievals and calibrated IR‐based rainfall so that it will produce a higher quality precipitation data [20]. The high quality satellite rainfall will be used to generate a better TC rain climatology. In addition, precipitation has a strong diurnal cycle property [89– 91] and there is also a clear diurnal feature in TC lifecycle [92]. These diurnal properties could also be utilized in improving TC rain forecasts in the near future. However, the best TC rain forecast should come from future advanced cloud‐revolving models which could predict not only TC intensity and track, but also rain distributions at different spatial scales.

#### **6. Impacts of satellite remote sensing on TC forecasts**

The NWP models are used at major weather operation centers to provide regular 7‐day weather forecasts. The prediction skills have been consistently improving about one day per decade in last several decades with advances in NWP model developments, application of more satellite observations, and high‐performance computing power [93]. Especially after global satellite observations are applied in the NWP data assimilations, the NWP skills are proven very accurate for 3‐day forecast, highly accurate for 5‐day forecasts and very useful for

in operations. A variation of this method called the areal TRaP has been introduced to graphic view of TC precipitation horizontal distributions [83, 84], with three correct assumptions of TC track forecast, satellite rain estimates, and persistent spatial pattern of rainrates relative to the TC center. TRaP is normally valid for short forecasts of less than 24 hr. However, there are limitations associated with TRaP because of no considerations on changes of TC intensity and

Another method is the rainfall climatology and persistence (R‐CLIPER) model which is a parametric model utilized with the TC rainfall climatology from satellite measurements [85]. This method assumes a circularly symmetric distribution of rainfall and its rainfall is translated in time. Although it accounts for TC intensity and moving speed, it does not include the TC's unique asymmetry rainfall patterns. An improved method called the parametric hurricane rainfall model (PHRaM) was introduced to incorporate with the TC rainfall asymmetry feature by including the azimuthal Fourier decomposition for shear and a term indicating the topographical uplift [86]. Results show that PHRaM is improved significantly compared with the standard R‐CLIPER. A new parametric model was recently developed for including more factors such as TC motion speed and intensity, vertical wind shear, and typical features of the

All the above‐mentioned methods have one common assumption of the correct satellite rainfall retrievals. Although satellite‐derived precipitation from PMW sensors is very accurate over ocean, there are still relatively large errors over land [75, 88]. Some discrepancies exist among satellite rain datasets, especially with different satellite sensors and retrieval algorithms. In order to minimize errors from different rain retrieval algorithms for different sensors, the physical‐based inversion rain algorithm (GPROF) used in TRMM and GPM is also applied for other PMW sensors [77]. Thus, precipitation from different PMW sensors will be more consistent. The new NASA integrated multi‐satellite retrievals for GPM (IMERG) is based on these consistent PMW rain retrievals and calibrated IR‐based rainfall so that it will produce a higher quality precipitation data [20]. The high quality satellite rainfall will be used to generate a better TC rain climatology. In addition, precipitation has a strong diurnal cycle property [89– 91] and there is also a clear diurnal feature in TC lifecycle [92]. These diurnal properties could also be utilized in improving TC rain forecasts in the near future. However, the best TC rain forecast should come from future advanced cloud‐revolving models which could predict not

only TC intensity and track, but also rain distributions at different spatial scales.

The NWP models are used at major weather operation centers to provide regular 7‐day weather forecasts. The prediction skills have been consistently improving about one day per decade in last several decades with advances in NWP model developments, application of more satellite observations, and high‐performance computing power [93]. Especially after global satellite observations are applied in the NWP data assimilations, the NWP skills are proven very accurate for 3‐day forecast, highly accurate for 5‐day forecasts and very useful for

**6. Impacts of satellite remote sensing on TC forecasts**

the TC's vertical wind shear conditions.

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

TC boundary layer [87].

**Figure 11.** This image uses the model output from the ECMWF experiments, showing where Sandy was predicted to be located 5‐days out with the normal satellite data inputs into the model (left) and without any polar‐orbiting satellite data (right). Both position and intensity forecasts were affected—Sandy stays out to sea without the polar‐orbiting sat‐ ellite data, and the closer isobar lines encircling the storm also imply a more organized and stronger system (adapted from NOAA at http://www.noaanews.noaa.gov/stories2012/20121211\_poesandsandy.html).

7‐day forecasts for both northern and southern hemispheres. Modern NWP models even show skill for extended forecast beyond 10 days and up to months. The advanced climate models could provide seasonal and longer time predictions with various confidence levels [94].

Accurate predictions of TC genesis, intensity, and track are crucial for preparation and mitigation of TC's impacts. The forecast skills of global NWP models were always superior in the northern hemisphere than the southern hemisphere until 1999 when global satellite measurements were successfully assimilated so that difference of the prediction skills between northern and southern hemispheres diminished [93, 95]. The role of satellite observations in NWP forecast skills is normally assessed by the observing system experiments (OSEs) in which denying or adding a set of satellite data is applied from or to a baseline observing system in order to show its impacts on the forecast skills [96–98]. The famous example of the impacts of satellite observations on the NWP forecast skills is the accurate prediction of Hurricane Sandy's left (westward) turn to make landfall on the New Jersey coast for 7–8 days in advance by the European Center for Medium‐Range Weather Forecasts (ECMWF) [99]. The OSE analyses show that the storm's landfall would be reasonable without observations from geostationary satellites; however, the prediction would not be very useful for 4–5 days before its landfall without measurements from polar‐orbital satellites assimilated into the system. **Figure 11** presents a comparison of the predicted Hurricane Sandy's positions before its landfall with and without the polar‐orbital satellites. It clearly proves that the storm's intensity and position were accurate with the LEO satellite data, while its intensity would be weak and its position offshore without satellite data assimilations.

**Figure 12.** Error trends of the NHC official TC forecast track and intensity at 24, 48, 72, 96, and 120 h for Atlantic basin. Top‐left pane is for TC track forecast error trend, while bottom‐left panel is for the track forecast skill trend. The right panel is same as the left panels except for TC intensity forecast (adapted from Cangialosi and Franklin (2015)).

More accurate forecasts of TC's intensity and track will have to come from the cloud‐resolving models such as the NOAA hurricane weather research and forecasting (HWRF) system and the NRL‐MRY Coupled Ocean/Atmosphere Mesoscale Prediction System for Tropical Cyclo‐ nes (COAMPS‐TC) because of the TC's strong intensity and small spatial size. Thus, a high spatial resolution is required in making accurate TC simulations and predictions. The error trends of the TC's track forecasts have been consistently and significantly reduced in last few decades, while improvements on error trends of the TC's intensity forecasts are not so great [100]. **Figure 12** is an example of the error trends of the National Hurricane Center (NHC) TC official track and intensity forecasts over the Atlantic basin. It demonstrates that the TC track forecast errors are substantially decreased (>50%) from the 1990s to 2014, especially for 4–5 day forecasts. The track error is only around 35, 60, 80, 140, and 190 nm for 24, 48, 72, 96, and 120 hr forecasts, respectively. The track forecast skill increased from about 10% in 1990 to 70% in 2014. However, a decrease of the TC intensity forecast error is very small at 24 hr, small at 48 and 72 hr, while substantial at 96 and 120 hr. The associated TC intensity forecast skill also increases accordingly. Therefore, the TC intensity forecast skill has improved slightly; however, these improvements are still statistically significant [101, 102].

Although improvement on the TC's intensity forecast is relatively small compared with the TC track forecast, accuracy of the TC intensity forecast is obviously improved since 2001 when more satellite observations have been systematically properly applied in NWP simulations.

**Figure 13.** Mean forecast errors and standard deviations as functions of forecast lead time for TC track (a and b), maxi‐ mum wind *V*max (c and d), and minimum center pressure *Pc* (e and f) of CTRL2 (solid) and CTRL2 + ATMS (dashed) of the four landfall storms (adapted from Zou et al. (2013). Reproduced by permission of American Geophysical Union).

More accurate forecasts of TC's intensity and track will have to come from the cloud‐resolving models such as the NOAA hurricane weather research and forecasting (HWRF) system and the NRL‐MRY Coupled Ocean/Atmosphere Mesoscale Prediction System for Tropical Cyclo‐ nes (COAMPS‐TC) because of the TC's strong intensity and small spatial size. Thus, a high spatial resolution is required in making accurate TC simulations and predictions. The error trends of the TC's track forecasts have been consistently and significantly reduced in last few decades, while improvements on error trends of the TC's intensity forecasts are not so great [100]. **Figure 12** is an example of the error trends of the National Hurricane Center (NHC) TC official track and intensity forecasts over the Atlantic basin. It demonstrates that the TC track forecast errors are substantially decreased (>50%) from the 1990s to 2014, especially for 4–5 day forecasts. The track error is only around 35, 60, 80, 140, and 190 nm for 24, 48, 72, 96, and 120 hr forecasts, respectively. The track forecast skill increased from about 10% in 1990 to 70% in 2014. However, a decrease of the TC intensity forecast error is very small at 24 hr, small at 48 and 72 hr, while substantial at 96 and 120 hr. The associated TC intensity forecast skill also increases accordingly. Therefore, the TC intensity forecast skill has improved slightly; however,

**Figure 12.** Error trends of the NHC official TC forecast track and intensity at 24, 48, 72, 96, and 120 h for Atlantic basin. Top‐left pane is for TC track forecast error trend, while bottom‐left panel is for the track forecast skill trend. The right panel is same as the left panels except for TC intensity forecast (adapted from Cangialosi and Franklin (2015)).

Although improvement on the TC's intensity forecast is relatively small compared with the TC track forecast, accuracy of the TC intensity forecast is obviously improved since 2001 when more satellite observations have been systematically properly applied in NWP simulations.

these improvements are still statistically significant [101, 102].

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

One of the important developments is to directly assimilate satellite observations into the cloud‐resolving models. For example, a recent study indicates that introduction of the ATMS data into the HWRF system has significant impacts on forecasts of hurricane intensity and track [103]. Four landfall Atlantic hurricanes (Beryl, Debby, Isaac, and Sandy) in 2012 were investigated using two sets of comparisons of four experiments: CTRL1 is for assimilations of the conventional data, GPS RO data and ASCAT surface winds; CRTL1 + ATMS is for CRTL1 plus additional ATMS data; CRTL2 is for experiment setting of CRTL1 plus additional AMSU‐ A, AIRS and HIRS data; CRTL2+ATMS is for CRTL2 plus additional ATMS data. Results show a reduced bias of TC track for CRTL2 than CRTL1 so that impacts of the polar‐orbital satellite data on TC track are further validated. The assimilation of additional ATMS data further reduced the bias of TC tracks and intensity as well as increased its lead time of forecasting. **Figure 13** presents the combined comparison results on track errors and standard deviations of the four TC forecasts between CRTL2 and CRTL2 + ATMS. The track errors are similar because of the abundant polar‐satellite measurements, while the errors of maximum wind speed (*V*max) and minimum center pressure (*Pc*) are significantly reduced. Although the standard deviation of track forecast is slightly large mainly due to the deteriorated Debby track forecasts, the overall *V*max and *Pc* errors are obviously reduced.

The COAMPS‐TC system developed at NRL‐MRY has been transitioned to operations for real‐ time TC forecasts for several hurricane seasons at a spatial resolution of 5 km and systematically evaluated for large samples of TC forecasts over Atlantic and West Pacific basins [104]. Results demonstrate the accurate predictions of TC track and intensity, as well as the sea surface temperature cooling response to the storm, indicating the capability of the COAMPS‐TC system to realistically capture characteristics of the ocean surface waves and their interactions with boundary layers above and below the ocean surface. There are more satellite measurements than what are actually assimilated into the models. Proper utilization of satellite data with positive impacts on forecast skills still requires more investigations and validations.
