Abstract

Utilizing the cloud parameters derived from the Tropical Rainfall Measuring Mission (TRMM) Visible and Infrared Scanner and the near-surface rainfall detected by the TRMM Precipitation Radar, the differences of cloud parameters for precipitating clouds (PCs) and nonprecipitating clouds (NPCs) are examined in tropical cyclones (TCs) during daytime from June to September 1998–2010. A precipitation delineation scheme that is based on cloud parameter thresholds is proposed and validated using the independent TC datasets in 2011 and observational datasets from Terra/MODIS. Statistical analysis of these results shows that the differences in the effective radius of cloud particles Re are small for PCs and NPCs, while thick clouds with large cloud optical thickness (COT) and liquid water path (LWP) can be considered as candidates for PCs. The probability of precipitation increases rapidly as the LWP and COT increase, reaching ~90%, whereas the probability of precipitation reaches a peak value of only 30% as Re increases. The combined threshold of a brightness temperature at 10.8 μm (BT4) of 270 K and an LWP of 750 g m−2 shows the best performance for precipitation discrimination at the pixel levels, with the probability of detection (POD) reaching 68.2% and false-alarm ratio (FAR) reaching 31.54%. From MODIS observations, the composite scheme utilizing BT4 and LWP also proves to be a good index, with POD reaching 77.39% and FAR reaching 24.2%. The results from this study demonstrate a potential application of real-time precipitation monitoring in TCs utilizing cloud parameters from visible and infrared measurements on board geostationary weather satellites.

1. Introduction

The accurate and quantitative documentation of global precipitation provided by satellites is crucial in determining the general circulation of Earth’s atmosphere and the global climate. However, mistakes in precipitating cloud determination can lead to considerable errors in satellite-based rainfall retrievals (i.e., radiometer or visible- or infrared-based rainfall retrievals). Thus, discrimination between precipitating (PCs) and nonprecipitating clouds (NPCs) is the most fundamental step (Stephens and Kummerow 2007) in numerous satellite-based rainfall retrieval techniques (Adler and Negri 1988; Arkin and Ardanuy 1989; Lensky and Rosenfeld 2003; Joyce et al. 2004; Huffman et al. 2007).

The traditional visible/infrared (VIS/IR) sensors on board satellites have been widely applied in global precipitation monitoring, because VIS/IR instruments can be loaded aboard geostationary satellites with the advantage of high sampling frequency and fine spatial resolution. There have been numerous efforts over the past few decades to improve the algorithms for the delineation of precipitation from VIS/IR instruments on board weather satellites (Huffman et al. 2001; Hong et al. 2004; Casella et al. 2013; Xu et al. 2014). Generally, clouds with colder tops have a higher probability of raining (Arkin and Ardanuy 1989). The Geostationary Operational Environmental Satellite precipitation index used a threshold of tops of clouds colder than 235 K. However, this cloud-top temperature threshold (235 K) varies with geographical location and the type of precipitation (Huffman et al. 2001; Hong et al. 2004). The major limitation of this infrared threshold scheme is the uncertainty in discriminating between nonprecipitating anvil clouds (with a lower cloud-top temperature, but no precipitation) and precipitating warm clouds (with a higher cloud-top temperature and precipitation), which may cause the overestimation or underestimation of surface rainfall. To overcome this drawback, researchers improved the algorithms by taking advantage of signals from multiple channels, for example, the ratio of the reflectance at 0.6 μm to the reflectance at 1.6 μm or the difference in brightness temperature between 8.7 and 10.8 μm (Inoue and Aonashi 2000; Behrangi et al. 2009; Liu and Fu 2010; Feidas and Giannakos 2011; Giannakos and Feidas 2011), from which additional information regarding cloud properties can be inferred. Generally, the dual signals improve the false-alarm situation and the accuracy of detecting low-level thick precipitating clouds.

As compared with the VIS/IR signals, the cloud parameters retrieved from VIS/IR radiances are direct indicators of cloud microphysical properties and have been applied to identify PCs and NPCs. The conceptual model is that precipitating clouds are mostly characterized by relatively high cold tops, sufficient water, and mostly large ice particles in the upper clouds (Rosenfeld and Gutman 1994; Lensky and Rosenfeld 1997; Rosenfeld and Lensky 1998; Nauss and Kokhanovsky 2006, 2007; Thies et al. 2008). Cloud droplets with a large effective radius near cloud top were shown to be conducive to the precipitation formation process using Advanced Very High Resolution Radiometer (AVHRR) data (Rosenfeld and Gutman 1994; Lensky and Rosenfeld 1997; Rosenfeld and Lensky 1998). Nauss and Kokhanovsky (2006) proposed that clouds with a combination of a large effective droplet radius and a large optical thickness have a high probability of forming precipitation. A few studies have also revealed the ability of the liquid water path to discriminate precipitation in warm clouds based on the joint observations of the CloudSat Cloud Profiling Radar and the Moderate Resolution Imaging Spectroradiometer (MODIS) (Chen et al. 2011; Suzuki et al. 2011). These studies confirmed the feasibility of delineating precipitation based on the microphysical properties of clouds. However, these results may not be statistically robust because of the nonsynchronous data for precipitation and cloud parameters obtained from observations on two different platforms (i.e., ground-based weather radar or rain gauge data and spectral observations from satellites), which are relatively poorly matched in both space and time.

The launch of the Tropical Rainfall Measuring Mission (TRMM) in 1997 (Kummerow et al. 1998), with nearly simultaneous measurements from the Precipitation Radar (PR) and Visible and Infrared Scanner (VIRS) instruments, provides a unique opportunity to integrate the measurements of clouds and precipitation. The TRMM VIRS provides the multichannel cloud-top radiance from the visible to infrared bands, which can be used to retrieve cloud parameters, including the effective radius of cloud particles Re, the cloud optical thickness (COT), and liquid water path (LWP) (Twomey and Seton 1980; Nakajima and King 1990; (see section 2 for details). The TRMM PR, which is the first spaceborne instrument capable of measuring three-dimensional radar reflectivity, provides a more direct validation of ground precipitation than passive spaceborne instruments. By merging the datasets from the TRMM PR and VIRS instruments (Liu et al. 2008; Liu and Fu 2010; Fu 2014; Liu et al. 2014), researchers have examined the differences in cloud parameters between PCs and NPCs over the tropics and subtropics. In addition, as one of the most devastating natural disasters, tropical cyclone (TC) precipitation has been studied in much previous research (Rodgers et al. 2001; Jiang and Zipser 2010; Jiang et al. 2011; Chen and Fu 2015). However, the feasibility of precipitation delineation based on cloud properties in weather systems such as TCs over the northwest Pacific (NWP) has not been examined, although this is important in forecasting extreme weather events.

The work reported here aims to develop the precipitation clouds detection scheme in TCs over the NWP using cloud parameter datasets from joint observations of TRMM PR and VIRS. To achieve this goal, utilizing the cloud parameters (Re, COT, and LWP) retrieved from TRMM VIRS observations (Fu 2014), and the precipitation clouds determined by PR, differences between cloud parameters for PCs and NPCs in TCs are first investigated over the NWP from June to September 1998–2010. The thresholds of cloud parameters that are able to discriminate between PCs and NPCs in TCs are then examined. Based on performance metrics [i.e., probability of detection (POD) and false-alarm ratio (FAR)], the performance of our threshold-based scheme of cloud parameters is quantitatively evaluated using independent observational datasets from 2011 and an example of a precipitation event over the NWP observed by Terra MODIS.

2. Data and methods

a. Data and cloud parameter retrieval methods

The work reported here is mainly based on the merged datasets of cloud parameters retrieved from TRMM VIRS and precipitation observed from TRMM PR (Liu et al. 2008; Liu and Fu 2010; Fu 2014; Liu et al. 2014). The cloud parameters (Re and COT) were retrieved simultaneously using the bispectral reflectance method (BSR) (Twomey and Seton 1980; Nakajima and King 1990). This BSR algorithm had been adapted by numerous satellite missions. The principles of this method are based on the fact that the reflection function of clouds in the nonabsorbing channel in the visible wavelength region is primarily a function of the COT, whereas the reflection function in a water-absorbing channel in the near-infrared wavelength region is primarily a function of Re (Nakajima and King 1990). The cloud parameters Re and COT were retrieved from the lookup table (LUT) of reflectance in the absorptive and nonabsorptive channels for water and ice clouds. The LWP was obtained by a simple empirical equation of COT and Re (Nakajima and Nakajima 1995), that is, , where is liquid water density.

Because TRMM VIRS senses solar radiation in five spectral channels, with one nonabsorptive (0.6 μm) and one absorptive channel (1.6 μm), Re and COT can be retrieved from the TRMM VIRS 1B01 datasets based on the BSR method. The LUT was established using the radiative transfer mode [Santa Barbara DISORT Atmospheric Radiative Transfer (SBDART); Ricchiazzi et al. 1998]. The corresponding Re and COT were then retrieved by searching and interpolating the LUT. The cloud parameters were only retrieved during daytime because this is when the solar radiation datasets are available. The cloud parameter retrieval from the satellite observations also included the cloud cover (clear sky or clouds) and cloud phase test (Liu et al. 2008; Liu et al. 2014).

The TRMM PR 2A25 dataset (Iguchi et al. 2000; Schumacher and Houze 2003; Fu et al. 2011) was also used; this provides the three-dimensional radar reflectivity for each PR profile and the location in longitude and latitude with a horizontal resolution of 4 km. The 2A25 datasets were proved to be reliable and have been widely applied to meteorological and climatological studies (Fu et al. 2003, 2016; Schumacher and Houze 2003). Pixels with a near-surface rainfall larger than 0.4 mm h−1 were recognized as PCs, and the other pixels passing cloud tests (Liu et al. 2008) were defined as NPCs. Near-surface rainfall is the rainfall estimate at the lowest point not affected by the mainlobe clutter. The sensitivity of the PR instrument is ~17 dBZ (decreased to 18 dBZ after the orbital boost in August 2001), which corresponds to a rainfall of ~0.4 mm h−1 in rain rate (Schumacher and Houze 2003), thus the characteristics of cloud parameters for weak precipitating clouds are not included in this paper. As a result of the quasi-synchronous measurement of PR and VIRS, a merged dataset of PR and VIRS was established by collating the VIRS pixels (with a horizontal resolution of 2 km) at the resolution of the PR instrument to obtain the synchronous observations of precipitation and spectral information; the mean biases of reflectance and brightness of VIRS pixels are less than 0.05% and 2.5%, respectively, during the collating process (Fu et al. 2011). The cloud parameters were then retrieved by the BSR method based on the merged dataset of TRMM PR and VIRS. Synchronous datasets of the cloud parameters and precipitation were then established, providing the foundation for our investigation of the characteristics of the cloud parameters of PCs and NPCs. This dataset had been widely used in many previous studies (Liu et al. 2008; Liu and Fu 2010; Fu 2014; Liu et al. 2014).

The locations of TCs over NWP during the months of June–September 1998–2011 were provided by the Japan Aerospace Exploration Agency Earth Observation Research Center (http://sharaku.eorc.jaxa.jp/cgi-bin/typ_db/typ_track.cgi?lang=e&area=WP). Cloud parameters from the MODIS MYD06 datasets (Chu et al. 2002) were also used, which include Re, COT, and LWP at a spatial resolution of 1 km.

b. Performance metrics of the threshold-based precipitation delineation scheme

To measure the skill of the cloud parameter threshold-based scheme in the delineation of PCs, several widely used performance metrics, including bias B, POD, FAR, threat score (TS), and Heidke skill score (HSS), were used in this study. Among the five indices, POD, FAR, and TS range from 0 to 1. As POD and TS approach 1, and FAR approaches 0, the accuracy of the PC delineation using the threshold-based scheme increases. As B becomes greater (smaller) than 1, the predicted PCs are overestimated (underestimated) relative to the TRMM PR observations. The range of the HSS is from to 1. Negative values indicate that the chance forecast is better, 0 means no skill, and a perfect forecast obtains an HSS of 1. These statistics were calculated according to the following equations, and the meanings of the variables are presented in Table 1:

 
formula
 
formula
 
formula
 
formula
 
formula
Table 1.

Contingency table between the PCs or NPCs observed by TRMM PR, and the PCs and NPCs retrieved from the cloud parameter threshold-based scheme.

Contingency table between the PCs or NPCs observed by TRMM PR, and the PCs and NPCs retrieved from the cloud parameter threshold-based scheme.
Contingency table between the PCs or NPCs observed by TRMM PR, and the PCs and NPCs retrieved from the cloud parameter threshold-based scheme.

3. Results

a. Case studies of tropical cyclone cloud parameters

The tropical depression Choi-Wan was observed over the NWP by TRMM on 17 September 2009. The eye of the tropical cyclone was located at 21°N, 140°E, with the maximum sustained wind speed reaching above 130 kt (1 kt = 0.51 m s −1). Figures 1a and 1b show the spatial pattern of the reflectance at 0.63 μm and the brightness temperature at 10.8 μm (BT4). The reflectance at 0.63 μm reaches above 0.7 for the eyewall of the tropical cyclone, and the BT4 was less than 220 K, which is representative of deep clouds with more ice particles in the upper clouds. The eye of the tropical cyclone had a lower reflectance at 0.63 μm and a higher BT4 than the TC eyewall. A high rate of near-surface rainfall greater than 20 mm h−1 occurred in the eyewall and spiral rainbands of the tropical cyclone (Fig. 1c).

Fig. 1.

TRMM satellite observations of TC Choi-Wan on 17 Sep 2009: (a) Spatial pattern of reflectance at 0.63 μm, (b) brightness temperature at 10.8 μm observed by the VIRS, (c) near-surface rain rate observed by PR, (d) retrieved Re, (e) retrieved COT, and (f) retrieved LWP. The blue line A–B in (b) crosses through the eye of the TC.

Fig. 1.

TRMM satellite observations of TC Choi-Wan on 17 Sep 2009: (a) Spatial pattern of reflectance at 0.63 μm, (b) brightness temperature at 10.8 μm observed by the VIRS, (c) near-surface rain rate observed by PR, (d) retrieved Re, (e) retrieved COT, and (f) retrieved LWP. The blue line A–B in (b) crosses through the eye of the TC.

Figures 1d–f show the values of Re, COT, and LWP retrieved from the TRMM dataset. The COT and LWP for the main body of TC clouds (a radius of ~200 km from the center of the TC) were greater than 90 and 1000 g m−2, respectively, which is larger than the values for the clouds near the outer edges of the TC (shown as NPCs in Fig. 1c). It is evident that the COT was smaller than 40 and the LWP is smaller than 200 g m−2 for these NPCs. These discrepancies suggest that the areas of precipitation are in good agreement with the areas with optically thick clouds and sufficient water, which suggests that the COT and LWP are effective measures of the potential for precipitation to form in clouds in this example. The cloud droplet size (Fig. 1d) for the clouds near the outer edges of the TC (NPCs) was roughly less than 12 μm, whereas the PCs in the inner spiral rainbands (approximately 100 km from the TC center) of the TC had a smaller cloud droplet size. It should be noted that Re retrieved from the visible band represents the cloud droplet size near the top of the clouds. A possible reason why the Re of PCs is smaller than the Re of NPCs may be that the large cloud particles near the tops of the clouds fall downward while precipitating. This may explain the relatively poor ability of Re to delineate areas of precipitation in tropical cyclones.

The left-hand panels of Fig. 2 show the near-surface rain rate, VIRS signals, and cloud parameters along the blue line A–B across the eye of the TC in Fig. 1b, with the aim of making a more quantitative assessment of the differences in the VIRS signals and cloud parameters between precipitating and nonprecipitating clouds. There was almost no precipitation in the eye of the TC (Fig. 2a). The maximum rain rate occurred at 100 km from the eye, with the reflectance at 0.63 μm reaching above 0.8 and BT4 less than 200 K. The NPCs were mainly distributed at distances of above 200 km from the eye of the TC and had a low reflectance and a large brightness temperature. Figures 2c and 2e also show differences between PCs and NPCs for the ratio of reflectance at 0.63 μm to reflectance at 1.6 μm and the difference in the brightness temperature between 10.8 and 12.0 μm. Previous studies have shown that these two indices contain information about the cloud droplet phase and the absorption/emission strength for ice and water droplets, demonstrating the advantages of discriminating precipitation using multispectral information (Thies et al. 2008; Feidas and Giannakos 2011; Giannakos and Feidas 2011).

Fig. 2.

(a) Near-surface rain rate observed by PR, (b) the retrieved Re, (c) the reflectance at 0.63 and 1.6 μm, (d) retrieved COT, (e) brightness temperature at 3.75, 10.8, and 12.0 μm observed by the VIRS, and (f) retrieved LWP along the blue line A–B in Fig. 1b. The green star denotes the eye of the TC. The region between the two gray dashed lines denotes the PCs.

Fig. 2.

(a) Near-surface rain rate observed by PR, (b) the retrieved Re, (c) the reflectance at 0.63 and 1.6 μm, (d) retrieved COT, (e) brightness temperature at 3.75, 10.8, and 12.0 μm observed by the VIRS, and (f) retrieved LWP along the blue line A–B in Fig. 1b. The green star denotes the eye of the TC. The region between the two gray dashed lines denotes the PCs.

The right-hand panels of Fig. 2 present the retrieved values of Re, COT, and LWP along the blue line A–B across the eye of the TC shown in Fig. 1b. The cloud droplet size for PCs was generally larger than that of NPCs, although the differences were relatively small—for example, Re ranges from 5 to 25 μm for NPCs and from 5 to 40 μm for PCs. The differences between PCs and NPCs were more distinct for COT and LWP. The COT of NPCs was mainly smaller than 40, whereas that for PCs ranged from 40 to 120; the LWP was usually larger than 500 g m−2 for PCs, but smaller than 100 g m−2 for NPCs. This analysis indicates that pixels with relatively large COT and LWP values could be designated as PCs with a relatively high reliability in TC Choi-Wan, which is in agreement with the conceptual model that precipitation tends to be generated from clouds with sufficient water content. However, the cloud droplet size at the tops of the clouds delineated the precipitation in tropical cyclone Choi-Wan relatively poorly, probably because of the large variations in the movement of cloud droplets caused by the strong vertical motion inside the precipitating clouds of tropical cyclones.

b. Statistical characteristics of TC cloud parameters for PCs and NPCs

We investigated the statistical characteristics of cloud parameters for PCs and NPCs in TCs over the NWP from June to September 1998–2010. Table 2 gives the number of samples and the mean Re, COT, and LWP values for PCs and NPCs in TCs during the study period. There were 1 908 851 and 6 043 681 pixels of PCs and NPCs, respectively, which is sufficient to ensure relatively reliable statistical results. The mean values of Re, COT, and LWP for the PCs were generally larger than those for the NPCs. The Re was 18 μm for PCs and 15 μm for NPCs, a small difference. However, clear differences were found for COT and LWP between the PCs and NPCs. The mean COT (LWP) of PCs was 93 (1087 g m−2), approximately 10 times that of NPCs, which is consistent with the results reported by Liu et al. (2008).

Table 2.

Number of samples and mean cloud parameters for PCs and NPCs in TCs from June to September 1998–2010.

Number of samples and mean cloud parameters for PCs and NPCs in TCs from June to September 1998–2010.
Number of samples and mean cloud parameters for PCs and NPCs in TCs from June to September 1998–2010.

The probability distribution functions (PDFs; %) of Re, COT, and LWP for PCs and NPCs in TCs are shown in Figs. 3a, 3b, and 3c, respectively. The Re of PCs ranges from 5 to 30 μm and reaches a peak frequency at 16 μm; by contrast, the Re for NPCs showed a bimodal pattern, with peak frequencies at 5 and 16 μm. The Re frequency patterns for PCs and NPCs were largely overlapped with similar widths, indicating little difference in the cloud droplet size at the top of TCs. By contrast, the COT and LWP of PCs were systematically higher than those of NPCs. The COT (LWP) was mainly less than 40 (120 g m−2) for NPCs but larger than 40 (120 g m−2) for PCs. The LWP showed a bimodal pattern for both PCs and NPCs, which reached peak frequencies at 11 and 40 g m−2 for NPCs and 130 and 1400 g m−2 for PCs. Thus it can be concluded that precipitation in TCs can be discriminated with a relatively high accuracy using appropriate COT- or LWP-based thresholds. On the physical level, this is easy to understand because the Re values retrieved from the VIRS observations are only a reflection of the particle size near the cloud tops, whereas the retrieved COT and LWP values represent the vertically integrated cloud water content inside the clouds.

Fig. 3.

PDFs of (a) Re, (b) COT, and (c) LWP for PCs and NPCs in TCs from June to September 1998–2010. The horizontal axis of LWP is presented in log scale.

Fig. 3.

PDFs of (a) Re, (b) COT, and (c) LWP for PCs and NPCs in TCs from June to September 1998–2010. The horizontal axis of LWP is presented in log scale.

To establish an empirical relationship between precipitation and cloud parameters, the probability of precipitation (POP) in the tropical cyclones was calculated on a pixel-by-pixel basis as a function of Re, COT, and LWP (Fig. 4). The POP was calculated as the ratio of the number of near-surface precipitating samples to the total number of cloudy samples at certain cloud parameter intervals. The black dots in Fig. 4 were plotted at the center of each cloud parameter interval. As the value of Re increased, the POP increased rapidly and reached a peak of 30% as Re increased to ~18 μm. However, the POP stayed roughly the same and even decreased gradually as Re increased further. By contrast, the POP increased monotonically with the increase of the COT and LWP. At very low values of the COT (<20), the POP was low and reached a maximum of only 9%. As the COT increased, the POP increased dramatically, reaching a maximum value of 70% when the COT was between 80 and 120. Similarly, at values of LWP between ~200 and ~250 g m−2, the POP was less than 15%. The International Satellite Cloud Climatology Project has suggested the clouds with LWP larger than 250 g m−2 can be designated as PCs, which is clearly different from our results (Lin and Rossow 1994, 1997), in which the POP increased from ~30% to 90% as the LWP increased from 750 to 2000 g m−2. These results indicate that there is a relatively high probability of forming precipitation as the single cloud parameter (COT or LWP) reaches a critical value, which provides a potential empirical algorithm for the delineation of the area of precipitation based on appropriate cloud parameter thresholds. It is worth noting that there are still 10% of clouds with LWP ranging from 1500 to 2000 g m−2 that are not precipitating during the study period. One possible reason could be that the clouds with sufficient cloud liquid water content are in the developing process (raindrops are not falling to the ground) during the overpasses of TRMM PR.

Fig. 4.

POP as a function of (a) Re, (b) COT, and (c) LWP in TCs from June to September 1998–2010. The black dots are plotted at the mean values of each Re, COT, or LWP interval.

Fig. 4.

POP as a function of (a) Re, (b) COT, and (c) LWP in TCs from June to September 1998–2010. The black dots are plotted at the mean values of each Re, COT, or LWP interval.

Previous researchers have found evidence that multispectral signals are good at discriminating between PCs and NPCs (Thies et al. 2008; Feidas and Giannakos 2011; Giannakos and Feidas 2011). A combined index of cloud parameters may therefore supply more information about cloud properties, which may be beneficial for discrimination of the precipitating and nonprecipitating clouds. Figure 5 presents the frequency patterns of NPCs and PCs in two-dimensional space based on different parameters. The frequency pattern of NPCs (PCs) was defined as the number of NPC (PC) pixels at given values of Re and LWP and COT and LWP divided by the total number of NPC (PC) pixels. The PCs and NPCs can be discriminated by a combination of Re and LWP. Although the single cloud parameter Re shows a relatively poor ability in delineating precipitation, Figs. 5a and 5b show that a combination of the cloud parameters Re and LWP can discriminate the pixels with precipitation. For example, when Re is smaller than 10 μm and LWP is less than 100 g m−2, there is a high probability that these clouds are NPCs; by contrast, as the Re reaches above 10 μm and LWP reaches above 100 g m−2, the clouds have a high probability of precipitating. This is consistent with earlier suggestions that the cloud droplet size and the integrated water content inside clouds have a profound effect on the precipitation formation process (Nauss and Kokhanovsky 2006). Similarly, the PCs and NPCs can be designated by a combination of COT and LWP. The COT of NPCs is mainly less than 20, and the LWP is mostly greater than 100 g m−2. By contrast, the LWP of PCs is mainly larger than 100 g m−2, and the COT is less than 20. It is worth noting that there are still portions of PCs with a COT less than 20 and LWP less than 100 g m−2. By analyzing the near-surface rainfall of these clouds, it is shown that these clouds in TCs have a high probability of generating light precipitation (figure omitted).

Fig. 5.

Frequency pattern in two-dimensional space of (a),(b) Re and LWP and (c),(d) COT and LWP for (left) NPCs and (right) PCs in TCs from June to September 1998–2010. The horizontal axis of LWP and the vertical axis of COT are presented in log scale.

Fig. 5.

Frequency pattern in two-dimensional space of (a),(b) Re and LWP and (c),(d) COT and LWP for (left) NPCs and (right) PCs in TCs from June to September 1998–2010. The horizontal axis of LWP and the vertical axis of COT are presented in log scale.

Figure 6a shows the distribution of the POP in two-dimensional space in terms of Re and LWP, and Fig. 6b shows the distribution of the POP in terms of COT and LWP. It is clear that precipitation frequently occurred with large values of Re, COT, and LWP, although there were some differences. When the LWP was less than 500 g m−2, the POP did not depend on the value of Re; however, the POP was found to be a function of the LWP, which increased dramatically from 0% to above 15% as the LWP went from 0 to 500 g m−2. Thus potentially precipitating clouds need to have sufficient cloud water inside the clouds, rather than a large cloud droplet size near the cloud top. When the LWP was greater than 1000 g m−2 (and hence Re was larger than 12 μm), the POP was larger than 60%. This implies that when the LWP is sufficient, Re near the cloud tops is usually large, and these clouds have a much higher probability of forming precipitation. For the combination of COT and LWP (Fig. 6b), the POP varied greatly with LWP. The POP was relatively small at low values of LWP (the correspondent COT was low), but increased when LWP increased (the correspondent COT was relatively large).

Fig. 6.

POP in two-dimensional space consisting of (a) Re and LWP and (b) COT and LWP in TCs from June to September 1998–2010.

Fig. 6.

POP in two-dimensional space consisting of (a) Re and LWP and (b) COT and LWP in TCs from June to September 1998–2010.

The brightness temperature at 10.8 μm is roughly equal to the cloud-top temperature when clouds are considered as blackbodies and is an important index of cloud properties. We therefore examined the feasibility of combing BT4 with the three cloud parameters to delineate the precipitation area. Figure 7 presents the frequency patterns in two-dimensional space for Re, COT, or LWP and BT4 for NPCs and PCs.

Fig. 7.

Frequency pattern in two-dimensional space of (a),(b) Re and BT4, (c),(d) COT and BT4, and (e),(f) LWP and BT4 for (left) NPCs and (right) PCs in TCs from June to September 1998–2010. The vertical axes of LWP and COT are presented in log scale.

Fig. 7.

Frequency pattern in two-dimensional space of (a),(b) Re and BT4, (c),(d) COT and BT4, and (e),(f) LWP and BT4 for (left) NPCs and (right) PCs in TCs from June to September 1998–2010. The vertical axes of LWP and COT are presented in log scale.

There are two centers of maximum frequency in the two-dimensional space of Re and BT4 for NPCs (Fig. 7a). One is at Re ~ 16 μm and BT4 ~ 220 K, whereas the other is at Re ~ 8 μm and BT4 ~ 290 K. These two centers denote two types of nonprecipitating clouds: ice clouds with relatively large cloud particles, possibly cirrus clouds, and warm clouds with small cloud particles at the top. By contrast, the precipitating clouds (Fig. 7b) are clouds with low values of BT4 and large values of Re. Despite the great majority of PCs having a low BT4 and large Re, there were still some parts of the NPCs with a low BT4 and large Re. Consequently, it is not easy to make a reliable decision about whether individual pixels contain precipitating or nonprecipitating clouds because of the overlap of the frequency center at about the same values of BT4 and precipitating or nonprecipitating clouds, for both PCs and NPCs.

The frequency patterns for NPCs and PCs in the two-dimensional space consisting of COT and BT4 (Figs. 7c,d) were very different from each other. The NPCs were most likely to occur when BT4 was larger than 260 K and COT was less than 8. However, for PCs, the BT4 was mostly less than 240 K, and the COT was larger than 50. Similarly, the combined parameters of LWP and BT4 discriminated between PCs and NPCs (Figs. 7e,f). Clouds with an LWP less than 60 g m−2 and BT4 larger than 270 K can be designated with high reliability as nonprecipitating clouds, whereas the PCs have a low BT4 (<240 K) and large LWP (>400 g m−2). These results indicated that clouds with cold tops and sufficient cloud water inside the clouds are more likely to precipitate, although some parts of PCs show large values of BT4 and small values of COT or LWP; these are probably warm precipitation clouds over the ocean.

Figure 8 shows the distribution of the POP in the two-dimensional spaces of Re and BT4, COT and BT4, and LWP and BT4. Figure 8a shows that when the cloud-top temperature was low (<210 K), the POP was larger than 60%, regardless of the value of Re. However, as BT4 increased, the clouds with a larger Re near the top of the cloud tended to have a relatively higher probability of forming precipitation. For example, at the same brightness temperature of 220 K, the POP of clouds with Re of 20 μm was larger than 60%, whereas for clouds with Re less than 20 μm, the POP was only about 30%. This means that the clouds with colder tops, which are correlated with a higher ice content, have a high probability of precipitating, whereas in warmer clouds, the large cloud droplet size at the top accelerates the precipitation-forming process. For the POP in the two-dimensional space of COT and BT4 (Fig. 8b), when the COT was larger than 60 and BT4 was less than 260 K, the POP ranged from 30% to above 60%. The value of BT4 at ~260 K in TCs seemed to be a satisfactory divide with which to identify precipitation. When BT4 was less than 240 K, the clouds with a larger COT were more likely to precipitate. Similarly, precipitating clouds were more likely to precipitate at a large value of LWP and a small value of BT4 (Fig. 8c). Clouds with a large LWP had a higher probability of precipitating than clouds with the same BT4 and a smaller LWP, which is consistent with the physical model.

Fig. 8.

POP in two-dimensional space consisting of (a) Re and BT4, (b) COT and BT4, and (c) LWP and BT4 in TCs from June to September 1998–2010.

Fig. 8.

POP in two-dimensional space consisting of (a) Re and BT4, (b) COT and BT4, and (c) LWP and BT4 in TCs from June to September 1998–2010.

From these statistical results of the relations between precipitation and single or combined parameters, it can be concluded that the pixels with precipitation can be discriminated by using the appropriate single cloud parameter or combined cloud parameters. To select the optimum thresholds for the single or combined cloud parameters in delineating PCs, we therefore calculated the HSSs for multiple combined thresholds of cloud parameters in TCs from June to September 1998–2010. The optimum thresholds were hereafter selected based on the best HSS (presented in Table 3). Hereafter, using the independent TC dataset from June to September in 2011, the performance of these thresholds in delineating PCs and NPCs was further evaluated. The performance metrics, including POD, FAR, B, TS, and HSS (see section 2 for details), were calculated to make a more quantitative evaluation on the capability of these indices in distinguishing precipitation events, displayed in Table 3.

Table 3.

Statistically derived POD, FAR, B, TS, and HSS for precipitation discrimination observed by TRMM in TCs for June–September 2011.

Statistically derived POD, FAR, B, TS, and HSS for precipitation discrimination observed by TRMM in TCs for June–September 2011.
Statistically derived POD, FAR, B, TS, and HSS for precipitation discrimination observed by TRMM in TCs for June–September 2011.

It is shown in Table 3 that for the single cloud parameter indices (indices 1–3), an LWP threshold of 750 g m−2 showed the best performance in identifying PCs, exhibiting the best HSS, reaching 0.596, together with the POD reaching 68.9%, a low FAR (32.4%), and the smallest overestimation of precipitating pixels (B = 1.02). A COT threshold of 60 had the largest POD of 71.6%, but with a relatively high FAR of 37.07%. The threshold of Re showed the worst performance (worse FAR, B, TS, and HSS) in discriminating PCs, which is consistent with the results shown in Fig. 4a.

As for the combined thresholds (indices 4–8), the LWP-based indices (index 4, index 5, and index 8) exhibit a better performance than other combined thresholds, with HSS exceeding approximately 0.595. Among these three LWP-based indices, a combined threshold of BT4 and LWP of 270 K and 750 g m−2 (index 8) had a POD reaching 68.2%, with a FAR reaching 31.5% and a small underestimation of the numbers of pixels that were precipitating (B = 0.996), which is the best index. The cloud-top temperature is a reflection of cloud phase, which may add more information about precipitation than COT and Re. Note that the LWP-based indices (index 4, index 5, and index 8) also show a similar performance in comparison with the LWP-only index (index 2), which suggests the importance of sufficient cloud water content in forming precipitation. Although the single index of Re shows a bad performance in delineating PCs, the combined thresholds of LWP and Re (index 5) and BT4 and Re (index 6) perform much better, with much lower FAR and better B. In general, the LWP-only and LWP-based combined thresholds show a relatively good performance of delineating PCs.

c. Application of threshold-based scheme for the delineation of precipitation

With the development of satellite observations and cloud parameter retrieval methods, numerous products of cloud parameters have been released by Terra/Aqua MODIS and the CloudSat Cloud Profiling Radar. An example of precipitation over the NWP captured by TRMM and Aqua was selected to evaluate the performance of our threshold-based scheme in discriminating PCs using the MYD06 cloud parameter datasets. This example of precipitation took place on 1 July 2010, and the time difference between observations by the two satellites was less than 4 min. Therefore, to gain real information about the precipitation event, the MYD06 cloud parameters were spatially matched to the TRMM PR pixels using a nearest neighbor method. Figure 9 shows the cloud parameters observed by MODIS (Figs. 9a–c) matched to the TRMM PR resolution, the predicted precipitating pixels using index 5 and index 8 (Figs. 9d,e), and the precipitating pixels observed by TRMM PR (Fig. 9f).

Fig. 9.

Aqua/MODIS observations of clouds on 1 Jul 2010: the spatial patterns of (a) Re, (b) COT, and (c) LWP, the threshold-based scheme of predicted precipitating pixels by (d) index 5 and (e) index 8, and (f) the true precipitating pixels observed by TRMM PR.

Fig. 9.

Aqua/MODIS observations of clouds on 1 Jul 2010: the spatial patterns of (a) Re, (b) COT, and (c) LWP, the threshold-based scheme of predicted precipitating pixels by (d) index 5 and (e) index 8, and (f) the true precipitating pixels observed by TRMM PR.

Based on the measurements from TRMM PR (Fig. 9f), the precipitating clouds were located in zonal belts from ~27° to ~29°N. The corresponding COT and LWP (Figs. 9b,c) of the PCs were larger than those of the NPCs, with COT larger than 60 and LWP exceeding 1000 g m−2 for most of the PCs. These differences in the COT (LWP) for the PCs and NPCs are evidence of the relatively good potential of single cloud parameters for LWP (COT) in discriminating PCs. The differences between Re for the PCs and NPCs were not obvious, which confirmed the relatively poor ability of Re in classifying precipitation in TCs.

Table 3 showed that indices 2, 5, and 8 achieved a relatively good performance in discriminating precipitation. We predicted the precipitating pixels using index 5 and index 8 based on the MODIS cloud parameters. By comparing the predicted precipitating pixels with the observed pixels, it can be seen that the main body of PCs is predicted well by these indices. However, some pixels around the edge of the clouds were misclassified as precipitating clouds. Previous studies have indicated that clouds are surrounded by a transition zone in which the cloud microphysical properties change sharply (Redemann et al. 2009; Cai et al. 2017). The changes in cloud properties near the outer edges of clouds may account for this relatively poor performance of the threshold-based scheme for precipitation delineation around the cloud edges. Further studies are needed to improve the accuracy of this precipitation delineation scheme in the cloud transition zone.

Table 4 presents the statistically derived POD, FAR, B, TS, and HSS for the discrimination of precipitation for this example. It can be seen from the values of HSS that the prediction of precipitation by indices 2, 5, and 8 was more reliable than that predicted by the other five indices. The corresponding HSS values were 0.691, 0.692, and 0.702. When compared with the classification of precipitation using the TRMM cloud parameters (Table 3), the statistically derived results for the MODIS cloud parameters performed even better, which indicates the potential application of these threshold-based schemes to the MODIS cloud products.

Table 4.

Statistically derived POD, FAR, B, TS, and HSS for precipitation discrimination for the example of precipitation in Fig. 9.

Statistically derived POD, FAR, B, TS, and HSS for precipitation discrimination for the example of precipitation in Fig. 9.
Statistically derived POD, FAR, B, TS, and HSS for precipitation discrimination for the example of precipitation in Fig. 9.

4. Discussion and conclusions

The joint observations of PR and VIRS on board the TRMM satellite make it possible to obtain simultaneous information on precipitation and cloud radiance. Based on the widely used BSR algorithm, the cloud parameters, including Re, COT, and LWP, are retrieved using VIRS 1B01 datasets; the PR near-surface rain rates are used to identify the real PCs and NPCs. By collating the VIRS pixels to the resolution of PR, merged datasets of precipitation and cloud parameters are established, which is the foundation of our investigation on the precipitation detection scheme in TCs using cloud parameters.

In the case studies, the Re value for precipitating clouds is slightly larger than that of nonprecipitating clouds. However, there are large discrepancies between PCs and NPCs for the COT and LWP. The COT for NPCs is mostly smaller than 40, whereas the COT for PCs ranges from 40 to 120. Similarly, the LWP for PCs is larger than 500 g m−2, but is smaller than 100 g m−2 for NPCs. These results indicate that precipitation in tropical cyclones tends to be generated from clouds with sufficient water content. The relatively poor performance in delineating precipitation using the cloud droplet size at the top of clouds is probably due to the large variations in the movement of cloud droplets caused by the strong vertical motion inside precipitating clouds in tropical cyclones.

The statistical results show that Re for PCs has a single peak, reaching a maximum at 16 μm. The Re for NPCs shows a bimodal pattern with peaks at 5 and 16 μm. The overlap of Re for PCs and NPCs makes the cloud droplet size unsuitable for the delineation of precipitation. However, there is a clear difference between the probability distribution function of COT and LWP for PCs and NPCs. In addition, the probability of precipitation increases from ~30% to 90% rapidly, as the LWP increases from 750 to 2000 g m−2. However, as Re increases above 30 μm, the probability of precipitation reaches a peak value of only 30%.

For clouds with a relatively small LWP (less than 500 g m−2), the probability of precipitation is a function of LWP, regardless the value of Re. When the LWP was greater than 1000 g m−2 (and hence Re was larger than 12 μm), the POP was larger than 60%. This implies that when the LWP is sufficient, Re near the cloud tops is usually large, and these clouds have a much higher probability of forming precipitation. For the combination of COT and LWP, POP was relatively small at low values of LWP (the correspondent COT was low), but increased greatly when LWP increased (the correspondent COT was relatively large), which suggests LWP is an important indicator of precipitation.

These differences of cloud parameters for PCs and NPCs suggest the potential feasibility of delineating PCs using cloud parameter thresholds. We therefore propose a precipitation clouds detection scheme using cloud parameter thresholds. The optimum thresholds are selected by analyzing the HSSs for multiple thresholds of single or combined cloud parameters in TCs from June to September 1998–2010, from which the thresholds with the best HSS are chosen as the optimum thresholds. The statistical results show that the indices of LWP-only and LWP-based combined thresholds show a relatively good performance on delineating PCs. Among eight indices, an LWP threshold of 750 g m−2, a combined threshold of an LWP of 770 g m−2 and Re of 9 μm, and a combined threshold of BT4 of 270 K and LWP of 750 g m−2 have the best performance in discriminating precipitation at the pixel level, with HSSs reaching 0.596, 0.596, and 0.599, respectively. For the combined threshold of BT4 of 270 K and LWP of 750 g m−2, the probability of detection is 68.2%, FAR is 31.54%, with a small underestimation of the numbers of pixels that are precipitating (B = 0.996).

The threshold-based precipitation identification scheme derived from TRMM cloud products is further validated based on the MODIS cloud products. These schemes achieve a relatively good performance in discriminating precipitation on the basis of the MODIS cloud products. The composite scheme utilizing BT4 and LWP proves to be the optimal strategy for identifying PCs in TCs, with POD reaching 77.39% and FAR reaching 24.2%. To be noted, some precipitation pixels are overestimated or underestimated around the edges of clouds in this case. Future studies are needed to improve the accuracy of this precipitation delineation scheme in the cloud transition zone.

In our work, the cloud parameter threshold-based scheme performs well in delineating the PCs, but it is also critical to understand the uncertainties and limitations in this study. The BSR method makes several important assumptions about the cloud pixels, which may introduce bias in our study (Nakajima and King 1990; Liang et al. 2015; Zhang et al. 2016). These assumptions are that the cloud pixels are horizontally homogenous and independent from each other, which are known as plane-parallel situations (Zhang et al. 2016). In addition, the observation biases of VIS/IR upwelling radiance may also cause the uncertainties in the COT and Re retrievals (Nakajima and King 1990). The BSR in our study only uses the daytime reflectance to retrieve the cloud parameters; more work is still needed to retrieve the cloud parameters during nighttime and explore the potential of cloud parameters to delineate the PCs during nighttime. In addition, in this paper, we introduced a simple cloud parameter threshold-based yes/no scheme (precipitating or not) in delineating PCs and NPCs. This scheme can be further promoted by establishing cloud parameter thresholds as a function of other cloud parameters in the future.

Despite all the limitations, the threshold-based precipitation identification scheme achieves a relatively good performance in delineating PCs, and the BSR algorithms are currently the most widely used method to obtain the global cloud parameters dataset. A potential application of this threshold-based precipitation identification scheme is the improved rain area delineation in TCs during daytime, utilizing geostationary satellites equipped with VIS/IR instruments. The recently launched Chinese geostationary weather satellite—Fengyun 4 (FY-4)—and the Japanese Himawari-8 geostationary weather satellite, both equipped with VIS/IR sensors with high temporal and spatial resolution, will provide us with a great opportunity for TC precipitation monitoring from cloud parameter information over the NWP.

Acknowledgments

We are grateful to the Goddard Space Flight Center for providing TRMM PR 2A25 and VIRS 1B01 data. This research has been jointly supported by NSFC (Grants 4123041, 91337213, and 41675041), Special Funds for Public Welfare of China (GYHY201306077), and Master and Doctor Fund of Anhui Meteorological Bureau (Grant RC201701).

REFERENCES

REFERENCES
Adler
,
R. F.
, and
A. J.
Negri
,
1988
:
A satellite infrared technique to estimate tropical convective and stratiform rainfall
.
J. Appl. Meteor.
,
27
,
30
51
, https://doi.org/10.1175/1520-0450(1988)027<0030:ASITTE>2.0.CO;2.
Arkin
,
P. A.
, and
P. E.
Ardanuy
,
1989
:
Estimating climatic-scale precipitation from space: A review
.
J. Climate
,
2
,
1229
1238
, https://doi.org/10.1175/1520-0442(1989)002<1229:ECSPFS>2.0.CO;2.
Behrangi
,
A.
,
K.-L.
Hsu
,
B.
Imam
,
S.
Sorooshian
,
G. J.
Huffman
, and
R. J.
Kuligowski
,
2009
:
PERSIANN-MSA: A precipitation estimation method from satellite-based multispectral analysis
.
J. Hydrometeor.
,
10
,
1414
1429
, https://doi.org/10.1175/2009JHM1139.1.
Cai
,
H.
,
Y.
Fu
,
Q.
Chen
,
X.
Feng
,
X.
Ti
,
R.
Tao
, and
K.
Xu
,
2017
:
Optical properties of cirrus transition zones over China detected by CALIOP
.
J. Meteor. Res.
,
31
,
576
585
, https://doi.org/10.1007/s13351-017-6044-3.
Casella
,
D.
, and Coauthors
,
2013
:
Transitioning from CRD to CDRD in Bayesian retrieval of rainfall from satellite passive microwave measurements: Part 2. Overcoming database profile selection ambiguity by consideration of meteorological control on microphysics
.
IEEE Trans. Geosci. Remote Sens.
,
51
,
4650
4671
, https://doi.org/10.1109/TGRS.2013.2258161.
Chen
,
F.
, and
Y.
Fu
,
2015
:
Contribution of tropical cyclone rainfall at categories to total precipitation over the western North Pacific from 1998 to 2007
.
Sci. China Earth Sci.
,
58
,
2015
2025
, https://doi.org/10.1007/s11430-015-5103-9.
Chen
,
R.
,
Z.
Li
,
R. J.
Kuligowski
,
R.
Ferraro
, and
F.
Weng
,
2011
:
A study of warm rain detection using A-Train satellite data
.
Geophys. Res. Lett.
,
38
,
L04804
, https://doi.org/10.1029/2010GL046217.
Chu
,
D. A.
,
Y. J.
Kaufman
,
C.
Ichoku
,
L. A.
Remer
,
D.
Tanré
, and
B. N.
Holben
,
2002
:
Validation of MODIS aerosol optical depth retrieval over land
.
Geophys. Res. Lett.
,
29
,
8007
, https://doi.org/10.1029/2001GL013205.
Feidas
,
H.
, and
A.
Giannakos
,
2011
:
Identifying precipitating clouds in Greece using multispectral infrared Meteosat Second Generation satellite data
.
Theor. Appl. Climatol.
,
104
,
25
42
, https://doi.org/10.1007/s00704-010-0316-5.
Fu
,
Y.
,
2014
:
Cloud parameters retrieved by the bispectral reflectance algorithm and associated applications
.
J. Meteor. Res.
,
28
,
965
982
, https://doi.org/10.1007/s13351-014-3292-3.
Fu
,
Y.
,
Y.
Lin
,
G.
Liu
, and
Q.
Wang
,
2003
:
Seasonal characteristics of precipitation in 1998 over East Asia as derived from TRMM PR
.
Adv. Atmos. Sci.
,
20
,
511
529
, https://doi.org/10.1007/BF02915495.
Fu
,
Y.
,
P.
Liu
,
Q.
Liu
,
M.
Ma
,
L.
Sun
, and
Y.
Wang
,
2011
:
Climatological characteristics of VIRS channels for precipitating cloud in summer over the tropics and subtropics (in Chinese)
.
Daqi Yu Huanjing Guangxue Xuebao
,
6
,
129
140
.
Fu
,
Y.
, and Coauthors
,
2016
:
Recent trends of summer convective and stratiform precipitation in mid-eastern China
.
Sci. Rep.
,
6
,
33044
, https://doi.org/10.1038/srep33044.
Giannakos
,
A.
, and
H.
Feidas
,
2011
: Detection of rainy clouds based on their spectral and textural features on Meteosat multispectral infrared data. Proc. 2011 EUMETSAT Meteorological Satellite Conf., Oslo, Norway, EUMESTAT, https://www.eumetsat.int/cs/idcplg?IdcService=GET_FILE&dDocName=pdf_conf_p59_s3_09_giannako_v&allowInterrupt=1&noSaveAs=1&RevisionSelectionMethod=LatestReleased.
Hong
,
Y.
,
K.-L.
Hsu
,
S.
Sorooshian
, and
X.
Gao
,
2004
:
Precipitation estimation from remotely sensed imagery using an artificial neural network cloud classification system
.
J. Appl. Meteor.
,
43
,
1834
1853
, https://doi.org/10.1175/JAM2173.1.
Huffman
,
G. J.
,
R. F.
Adler
,
M. M.
Morrissey
,
D. T.
Bolvin
,
S.
Curtis
,
R.
Joyce
,
B.
McGavock
, and
J.
Susskind
,
2001
:
Global precipitation at one-degree daily resolution from multisatellite observations
.
J. Hydrometeor.
,
2
,
36
50
, https://doi.org/10.1175/1525-7541(2001)002<0036:GPAODD>2.0.CO;2.
Huffman
,
G. J.
, and Coauthors
,
2007
:
The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales
.
J. Hydrometeor.
,
8
,
38
55
, https://doi.org/10.1175/JHM560.1.
Iguchi
,
T.
,
T.
Kozu
,
R.
Meneghini
,
J.
Awaka
, and
K.
Okamoto
,
2000
:
Rain-profiling algorithm for the TRMM Precipitation Radar
.
J. Appl. Meteor.
,
39
,
2038
2052
, https://doi.org/10.1175/1520-0450(2001)040<2038:RPAFTT>2.0.CO;2.
Inoue
,
T.
, and
K.
Aonashi
,
2000
:
A comparison of cloud and rainfall information from instantaneous visible and infrared scanner and precipitation radar observations over a frontal zone in East Asia during June 1998
.
J. Appl. Meteor.
,
39
,
2292
2301
, https://doi.org/10.1175/1520-0450(2001)040<2292:ACOCAR>2.0.CO;2.
Jiang
,
H.
, and
E. J.
Zipser
,
2010
:
Contribution of tropical cyclones to the global precipitation from eight seasons of TRMM data: Regional, seasonal, and interannual variations
.
J. Climate
,
23
,
1526
1543
, https://doi.org/10.1175/2009JCLI3303.1.
Jiang
,
H.
,
C.
Liu
, and
E. J.
Zipser
,
2011
:
A TRMM-based tropical cyclone cloud and precipitation feature database
.
J. Appl. Meteor. Climatol.
,
50
,
1255
1274
, https://doi.org/10.1175/2011JAMC2662.1.
Joyce
,
R. J.
,
J. E.
Janowiak
,
P. A.
Arkin
, and
P.
Xie
,
2004
:
CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution
.
J. Hydrometeor.
,
5
,
487
503
, https://doi.org/10.1175/1525-7541(2004)005<0487:CAMTPG>2.0.CO;2.
Kummerow
,
C.
,
W.
Barnes
,
T.
Kozu
,
J.
Shiue
, and
J.
Simpson
,
1998
:
The Tropical Rainfall Measuring Mission (TRMM) sensor package
.
J. Atmos. Oceanic Technol.
,
15
,
809
817
, https://doi.org/10.1175/1520-0426(1998)015<0809:TTRMMT>2.0.CO;2.
Lensky
,
I. M.
, and
D.
Rosenfeld
,
1997
:
Estimation of precipitation area and rain intensity based on the microphysical properties retrieved from NOAA AVHRR data
.
J. Appl. Meteor.
,
36
,
234
242
, https://doi.org/10.1175/1520-0450(1997)036<0234:EOPAAR>2.0.CO;2.
Lensky
,
I. M.
, and
D.
Rosenfeld
,
2003
:
A night-rain delineation algorithm for infrared satellite data based on microphysical considerations
.
J. Appl. Meteor.
,
42
,
1218
1226
, https://doi.org/10.1175/1520-0450(2003)042<1218:ANDAFI>2.0.CO;2.
Liang
,
L.
,
L.
Di Girolamo
, and
W.
Sun
,
2015
:
Bias in MODIS cloud drop effective radius for oceanic water clouds as deduced from optical thickness variability across scattering angles
.
J. Geophys. Res. Atmos.
,
120
,
7661
7681
, https://doi.org/10.1002/2015JD023256.
Lin
,
B.
, and
W. B.
Rossow
,
1994
:
Observations of cloud liquid water path over oceans: Optical and microwave remote sensing methods
.
J. Geophys. Res.
,
99
,
20 907
20 927
, https://doi.org/10.1029/94JD01831.
Lin
,
B.
, and
W. B.
Rossow
,
1997
:
Precipitation water path and rainfall rate estimates over oceans using special sensor microwave imager and International Satellite Cloud Climatology Project data
.
J. Geophys. Res.
,
102
,
9359
9374
, https://doi.org/10.1029/96JD03987.
Liu
,
Q.
, and
Y.
Fu
,
2010
:
Comparison of radiative signals between precipitating and non-precipitating clouds in frontal and typhoon domains over East Asia
.
Atmos. Res.
,
96
,
436
446
, https://doi.org/10.1016/j.atmosres.2010.02.003.
Liu
,
Q.
,
Y.
Fu
,
R.
Yu
,
L.
Sun
, and
N.
Lu
,
2008
:
A new satellite-based census of precipitating and nonprecipitating clouds over the tropics and subtropics
.
Geophys. Res. Lett.
,
35
,
L07816
, https://doi.org/10.1029/2008GL033208.
Liu
,
X.
,
Q.
Liu
,
Y.
Fu
, and
R.
Li
,
2014
:
Daytime precipitation identification scheme based on multiple cloud parameters retrieved from visible and infrared measurements
.
Sci. China Earth Sci.
,
57
,
2112
2124
, https://doi.org/10.1007/s11430-014-4870-z.
Nakajima
,
T.
, and
M. D.
King
,
1990
:
Determination of the optical thickness and effective particle radius of clouds from reflected solar radiation measurements. Part I: Theory
.
J. Atmos. Sci.
,
47
,
1878
1893
, https://doi.org/10.1175/1520-0469(1990)047<1878:DOTOTA>2.0.CO;2.
Nakajima
,
T. Y.
, and
T.
Nakajima
,
1995
:
Wide-area determination of cloud microphysical properties from NOAA AVHRR measurements for FIRE and ASTEX regions
.
J. Atmos. Sci.
,
52
,
4043
4059
, https://doi.org/10.1175/1520-0469(1995)052<4043:WADOCM>2.0.CO;2.
Nauss
,
T.
, and
A. A.
Kokhanovsky
,
2006
:
Discriminating raining from non-raining clouds at mid-latitudes using multispectral satellite data
.
Atmos. Chem. Phys.
,
6
,
5031
5036
, https://doi.org/10.5194/acp-6-5031-2006.
Nauss
,
T.
, and
A. A.
Kokhanovsky
,
2007
:
Assignment of rainfall confidence values using multispectral satellite data at mid-latitudes: First results
.
Adv. Geosci.
,
10
,
99
102
, https://doi.org/10.5194/adgeo-10-99-2007.
Redemann
,
J.
,
Q.
Zhang
,
P. B.
Russell
,
J. M.
Livingston
,
L. A.
Remer
,
2009
:
Case studies of aerosol remote sensing in the vicinity of clouds
.
J. Geophys. Res.
,
114
,
D06209
, https://doi.org/10.1029/2008JD010774.
Ricchiazzi
,
P.
,
S.
Yang
,
C.
Gautier
, and
D.
Sowle
,
1998
:
SBDART: A research and teaching software tool for plane-parallel radiative transfer in the earth’s atmosphere
.
Bull. Amer. Meteor. Soc.
,
79
,
2101
2114
, https://doi.org/10.1175/1520-0477(1998)079<2101:SARATS>2.0.CO;2.
Rodgers
,
E. B.
,
R. F.
Adler
, and
H. F.
Pierce
,
2001
:
Contribution of tropical cyclones to the North Atlantic climatological rainfall as observed from satellites
.
J. Appl. Meteor.
,
40
,
1785
1800
, https://doi.org/10.1175/1520-0450(2001)040<1785:COTCTT>2.0.CO;2.
Rosenfeld
,
D.
, and
G.
Gutman
,
1994
:
Retrieving microphysical properties near the tops of potential rain clouds by multispectral analysis of AVHRR data
.
Atmos. Res.
,
34
,
259
283
, https://doi.org/10.1016/0169-8095(94)90096-5.
Rosenfeld
,
D.
, and
I. M.
Lensky
,
1998
:
Satellite–based insights into precipitation formation processes in continental and maritime convective clouds
.
Bull. Amer. Meteor. Soc.
,
79
,
2457
2476
, https://doi.org/10.1175/1520-0477(1998)079<2457:SBIIPF>2.0.CO;2.
Schumacher
,
C.
, and
R. A.
Houze
Jr.,
2003
:
Stratiform rain in the tropics as seen by the TRMM Precipitation Radar
.
J. Climate
,
16
,
1739
1756
, https://doi.org/10.1175/1520-0442(2003)016<1739:SRITTA>2.0.CO;2.
Stephens
,
G. L.
, and
C. D.
Kummerow
,
2007
:
The remote sensing of clouds and precipitation from space: A review
.
J. Atmos. Sci.
,
64
,
3742
3765
, https://doi.org/10.1175/2006JAS2375.1.
Suzuki
,
K.
,
G. L.
Stephens
,
S. C.
van den Heever
, and
T. Y.
Nakajima
,
2011
:
Diagnosis of the warm rain process in cloud-resolving models using joint CloudSat and MODIS observations
.
J. Atmos. Sci.
,
68
,
2655
2670
, https://doi.org/10.1175/JAS-D-10-05026.1.
Thies
,
B.
,
T.
Nauss
, and
J.
Bendix
2008
:
Discriminating raining from non-raining clouds at mid-latitudes using Meteosat Second Generation daytime data
.
Atmos. Chem. Phys.
,
8
,
2341
2349
, https://doi.org/10.5194/acp-8-2341-2008.
Twomey
,
S.
, and
K. J.
Seton
,
1980
:
Inferences of gross microphysical properties of clouds from spectral reflectance measurements
.
J. Atmos. Sci.
,
37
,
1065
1069
, https://doi.org/10.1175/1520-0469(1980)037<1065:IOGMPO>2.0.CO;2.
Xu
,
S.
,
C.
Wu
,
A.
Gonsamo
, and
Y.
Shen
,
2014
:
Delineation of rain areas with TRMM microwave observations based on PNN
.
Remote Sens.
,
6
,
12 118
12 137
, https://doi.org/10.3390/rs61212118.
Zhang
,
Z.
, and Coauthors
,
2016
:
A framework based on 2-D Taylor expansion for quantifying the impacts of subpixel reflectance variance and covariance on cloud optical thickness and effective radius retrievals based on the bispectral method
.
J. Geophys. Res. Atmos.
,
121
,
7007
7025
, https://doi.org/10.1002/2016JD024837.

Footnotes

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).