1. Introduction and background
The measurement of precipitation from space dates back to the advent of geostationary satellites (Barrett 1970) and their infrared sensors. Techniques for retrieving the surface rain at that time were based on information of cloud-top height, whose relationship to surface rainfall was rather ambiguous. The advantage of using microwave frequencies to penetrate clouds was recognized with the introduction of passive microwave (PMW) sensors. If sampled across the microwave spectrum (1–300 GHz), surface-originated upwelling radiation detected at the top of the atmosphere (TOA) can offer a valuable insight to the entire atmospheric column. The Special Sensor Microwave Imager (SSM/I; Hollinger et al. 1990) was the first widely used multichannel sensor that allowed rainfall to be detected in a more physical sense. The retrievals used lower frequencies (e.g., 19 GHz), where radiation is absorbed and reemitted by liquid hydrometeors, to derive information of column-integrated liquid water, while the upwelling radiation at higher frequencies (e.g., 85 GHz), strongly affected by ice scattering, offered insight into the upper layers of convective clouds. Thus, PMW retrievals employed both absorption and scattering properties of hydrometeors to relate the observed radiances at TOA to surface rainfall. Despite the improvements, precipitation measurements still suffered from serious discrepancies when compared with ground-based products (Ferraro 1997). Their inability to fully capture the variability of scattering and absorptive elements within the cloud was seen as the main obstacle for further improvements.
The first spaceborne Precipitation Radar (PR) was launched in 1997 aboard the Tropical Rainfall Measuring Mission (TRMM; Kummerow et al. 2000), along with the state-of-the-art microwave radiometer [TRMM Microwave Imager (TMI)]. With high spatial resolution, PR, upon calibration, directly measured vertical profile of reflectivity, which is proportional to the sixth moment of raindrop diameters in the measurement layers. For the first time, PMW measurements could be directly compared with radar’s vertical samplings of the atmospheric column, in both time and space. Profiling capabilities of PR allowed for better understanding of hydrometeor absorption and scattering signatures relative to the brightness temperature (Tb) vector by TMI (i.e., PMW sensor). Ground validation sites, such as Kwajalein, Marshall Islands, offered valuable ground-based radar rainfall measurements as a reference for both PR and TMI estimates (Kim et al. 2004; Houze et al. 2004; Schumacher and Houze 2000). This greatly improved PMW retrieval performance over the ocean where a low surface emissivity ensures a strong contrast between a radiometrically cold background and a warm, precipitation-related atmospheric signature (Ferraro and Marks 1995; Kummerow et al. 2001). Consequently, the PMW retrieval output over the ocean quickly came into reasonable agreement with independent estimates (e.g., TRMM ground radar validation sites at Kwajalein; Melbourne, Australia; Houston, Texas; and Darwin, Australia; Wolff et al. 2005). However, comparable agreement was much harder to achieve over land backgrounds (Wolff and Fisher 2009). Land surfaces are all highly emissive, which leads to Tb emission signatures similar to rain itself. With no obvious contrast between rain and surface-background emission signals, rainfall detection over land is based primarily on ice-induced scattering signatures (Wang et al. 2009) and its relationship to the surface rain rate (Ferraro 1997; Ferraro et al. 2013). As will be shown here, this limitation remains one of the greatest challenges in the global rainfall observations over land. TRMM, however, played a key role in revealing many of the PMW retrieval shortcomings, including the sensitivity to different geophysical parameters related to rainfall such as variability of the surface emissivity (Ferraro et al. 2013; Petty and Li 2013) as well as the assumptions inherent in the algorithms used to retrieve rainfall from Tb or reflectivity measurements (Kummerow et al. 2011) such as the drop size distribution (DSD) and hydrometeor vertical profile. While many improvements have been made over time (e.g., Grody 1991; Ferraro et al. 1994; Adler et al. 1994; Conner and Petty 1998; Seto et al. 2005; Kummerow et al. 2015), PMW retrievals essentially still rely on ice-scattering signal to retrieve surface rainfall and this relationship, it turns out, is quite sensitive to storm system dynamics.
The modern era of satellite observations emphasizes the importance of understanding this relationship even more thoroughly. The launch of the GPM (Hou et al. 2014) core satellite, with a dual-frequency precipitation radar (DPR) and the most accurate microwave imager to date (GMI), affords the opportunity to intercalibrate a multitude of PMW radiometers to the same reference. Blended products of global rainfall measurements, such as Integrated Multisatellite Retrievals for GPM (IMERG; Huffman et al. 2014, 2015), are becoming available at 30-min temporal resolution across the globe. Consistent and reliable retrievals over land are thus more critical than ever. Facing the limitation of having to infer surface precipitation from just the ice-scattering signal, this study seeks to better understand such relationships in nature and explores synoptic-scale structural and environmental parameters that may be used to characterize this relationship between the ice scattering signal and the surface rainfall.
Accuracy and bases of the PMW rainfall retrieval over land
Validating satellite precipitation retrievals on a global scale is a complex and difficult task (Turk et al. 2002). Recognizing the qualitative performance, fortunately, is much easier. A comparison of TRMM rainfall estimates (PR and TMI) identifies the presence of large-scale systematic differences in the retrievals (Berg et al. 2006, 2008; Yamamoto et al. 2008; Adler et al. 2012; Wang and Wolff 2012; Maggioni et al. 2016; Liu et al. 2017). An example is given in Fig. 1, where one year of TRMM data is used to present the difference in mean daily rain rate of the PR and Goddard profiling algorithm (GPROF) 2010, version 2 (version 7 for TMI; hereinafter referred to as PMW retrieval) algorithms over land. The comparison is made using the TRMM 3G68 product, which is an hourly gridded product containing TRMM 2A12 (Kummerow et al. 2001), 2A25 (Iguchi et al. 2000), and 2B31 (Haddad et al. 1997a,b) precipitation estimates (note this is not a standard TRMM product). Inspection of Fig. 1 reveals two types of problems: 1) surface contamination (e.g., Himalaya region) and 2) mean biases with a significant random component over large regions. The most pronounced error corresponds to regions in the Himalayas. This, however, is a surface-screening problem where the upwelling microwave signal, depressed by accumulated snow and ice on the ground, is erroneously related to the rainfall by the PMW retrieval. This is easily verified using the PR profiles, which show that the majority of PMW “precipitation” events are not associated with any echo in the atmosphere over this region. In the past, PMW algorithms have employed a number of screening steps taken prior to the retrieval process to avoid this misinterpretation (Ferraro et al. 1998; Kummerow et al. 2001; Gopalan et al. 2010; Meyers et al. 2015). In its most recent version, GPROF (Kummerow et al. 2015) uses daily snow-cover updates to ameliorate this problem. This has improved precipitation screening over regions with snow on the ground, resulting in systematic differences that are significantly lower than those seen in Fig. 1 (not shown here). Surface screening errors, however, are not the topic of this study. To avoid any contamination by this type of scattering signature, the focus of this paper is on tropical regions marked in Fig. 1 that show opposite PMW-to-PR systematic differences but have very similar surface backgrounds. Tropical Africa and South America stand out (highlighted in Fig. 1) although a similar dipole is seen in Australia, as well as over Southeast Asia, and the central and eastern United States. While both PR and PMW retrievals may be contributing to this disagreement, the PMW algorithm must rely on ice-scattering signatures only and is thus less reliable than the radar retrieval. The focus of this study is therefore on PMW retrieval and the potential elements that contribute to these systematic differences.
Mean daily rainfall differences between PMW (TMI) and PR sensors (TMI-PR) over land for 2008 on a 0.5° grid of the 3G68 product. The figure reveals regions where PMW retrieval (GPROF2010, version 2) underestimates and overestimates PR observations.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Mean daily rainfall differences between PMW (TMI) and PR sensors (TMI-PR) over land for 2008 on a 0.5° grid of the 3G68 product. The figure reveals regions where PMW retrieval (GPROF2010, version 2) underestimates and overestimates PR observations.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Mean daily rainfall differences between PMW (TMI) and PR sensors (TMI-PR) over land for 2008 on a 0.5° grid of the 3G68 product. The figure reveals regions where PMW retrieval (GPROF2010, version 2) underestimates and overestimates PR observations.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
It has been suggested that systematic differences over these areas may be a result of the differences between the instruments’ capability to detect light rainfall, errors in assumptions about the DSD in the radar algorithm, or presence of supercooled water. A considerable amount of research has been done on these (e.g., Olson et al. 2006; Yang et al. 2006; Shige et al. 2006; Seo et al. 2007; Wang et al. 2009). Nevertheless, the regional systematic differences over land remain significant. The possibility that different regions have systematically different ice contents and related Tb depressions for the same surface rainfall has not been sufficiently explored. Therefore, this study will focus on understanding mean differences over regions of generally similar atmospheric and surface background conditions but opposite PMW-to-PR differences.
Deeper insight into this problem is offered in Fig. 2 using the aforementioned regions of Africa and South America as a test bed. The rainfall ratio (PR/PMW) between raining scenes detected by the two sensors is plotted as a function of rain rate. The black line reflects a mean ratio of the PR and PMW rainfall over the two regions after the overall systematic difference is removed. Focusing to the mean ratio only, one can easily note that PMW retrieval tends to overestimate PR at low rain rates (0–5 mm h−1), while underestimating higher ones (>5 mm h−1). This is a general property of GPROF retrieval caused by the fact that algorithm has less information content than the PR and thus tends to drive individual pixels toward the mean solution. However, when the regions are analyzed individually, the ratio is seen to be consistently above (Amazon) and below (Africa) the mean value, suggesting a clear regional dependency of the relationship on the observed vector and precipitation. Thus, according to Eqs. (1) and (2) and Fig. 2, it can be hypothesized that cloud microphysical properties must be substantially different in storms over the Amazon River region than in storms over the African region.
PR-to-PMW rainfall estimate ratio (PR/TMI) as a function of rain rate for Amazon (blue), African (red), and both regions combined (black). Overestimations by PMW retrieval are shaded in light red and underestimations in light blue.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
PR-to-PMW rainfall estimate ratio (PR/TMI) as a function of rain rate for Amazon (blue), African (red), and both regions combined (black). Overestimations by PMW retrieval are shaded in light red and underestimations in light blue.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
PR-to-PMW rainfall estimate ratio (PR/TMI) as a function of rain rate for Amazon (blue), African (red), and both regions combined (black). Overestimations by PMW retrieval are shaded in light red and underestimations in light blue.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
In support of this hypothesis is a global map of annualized distribution of total lightning activity shown in Fig. 3 (Stolz et al. 2015). It can be seen that higher lightning densities coincide with strong PMW overestimations over Africa, while lower flash rates coincide with PMW underestimations over the Amazon region of Fig. 1. It is known that the majority of rainfall over these regions originates from a structurally different type of system: mesoscale convective systems (MCSs; e.g., Toracinta and Zipser 2001) in Africa and warmer and shallower storms over the Amazon region. Therefore, it is expected that the microphysical processes and the cloud structure of storms in these regions are likely different as well. Other areas with similar surface types but opposite systematic differences (e.g., east/west United States) may have different lightning density ratios than does the Amazon–Africa example. However, this can be explained by the fact that cloud microphysics and structure are influenced by a number of factors such as topography, mesoscale air masses, and the large-scale environment. Combined, these factors play a critical role in the life cycle of storms (e.g., Rasmussen et al. 2016) by suppressing and enhancing the ice phase and the associated scattering signal upon which the passive microwave rainfall is predicted. From this perspective, it is worth noting that, in addition to earlier studies (e.g., Williams et al. 1992, 2005; Ba et al. 1998; Gilmore and Wicker 2002; Qie et al. 2003) that showed positive correlations between lightning, microphysics, and the amount of passive microwave scattering, the most recent study by Barth et al. (2015), based on data from Deep Convective Clouds and Chemistry (DC3) field campaign, links lightning flash rates to thermodynamical drivers of precipitation regimes over the U.S. Great Plains. They found substantial differences in flash rates of storms occurring during periods of high and low values of shear and CAPE. McCollum et al. (2000) compared Global Precipitation Climatology Project (GPCP) data with satellite infrared and microwave rain estimates to find essentially the same discrepancies as those seen in Fig. 1. As one possible cause, they note that distinct environmental conditions exist over these regions, using the observed difference in lightning density as a supporting argument.
The annualized distribution of total (cloud-to-ground plus cloud-to-cloud) lightning activity (flashes per kilometer squared per year) detected by the Lightning Imaging Sensor (LIS) Very High Resolution Full Climatology (VHRFC) in 1998–2013. Source: NASA Global Hydrological Resource Center (GHRC).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
The annualized distribution of total (cloud-to-ground plus cloud-to-cloud) lightning activity (flashes per kilometer squared per year) detected by the Lightning Imaging Sensor (LIS) Very High Resolution Full Climatology (VHRFC) in 1998–2013. Source: NASA Global Hydrological Resource Center (GHRC).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
The annualized distribution of total (cloud-to-ground plus cloud-to-cloud) lightning activity (flashes per kilometer squared per year) detected by the Lightning Imaging Sensor (LIS) Very High Resolution Full Climatology (VHRFC) in 1998–2013. Source: NASA Global Hydrological Resource Center (GHRC).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
To assess our comprehension of cloud dynamics and microphysics in the regions described above, this study seeks to understand and identify causes and potential predictors of observed systematic differences of passive microwave precipitation relative to PR estimates. The study will focus on the Amazon and African regions to minimize variability due to surface type, proximity to the ocean, and advected air masses.
2. Data and precipitation climatology
This study employs 1 yr (2010) of TRMM PR and TMI data to detect regional systematic differences of PMW retrieval (shown in Figs. 1 and 2) and to provide insight into the vertical structure and character of precipitation regimes. The standard TRMM PR 2A25 product (Iguchi et al. 2000, 2009) provides the attenuation-corrected radar reflectivity (ZE) profile (every 250 m from the surface to 20 km), freezing-level estimate, and near-surface rainfall rate at the PR native spatial resolution (approximately 5 km). Precipitation type at the pixel level (convective–stratiform) is given by the 2A23 product (Awaka et al. 2009). PMW retrieval, surface rain-rate estimates, and corresponding TMI brightness temperatures are obtained from the GPROF (see section 1a) standard output and 1B11 product, respectively. An additional year (2008) of the same datasets is used as an independent sample to test the robustness of this study’s findings.
Environmental parameters—namely, CAPE, wind profile, temperature, dewpoint, and specific humidity—are taken from the European Centre for Medium-Range Weather Forecasts (ECMWF) interim reanalysis (ERA-Interim) model data (Dee et al. 2011), at 0.75° horizontal and 6-h temporal resolution, at four pressure levels (850, 700, 500, and 200 hPa) for the same two years as the TRMM data. While model-induced uncertainties exist, this dataset is still seen as the best resource based on its consistency, coverage, and use in potential applications. Thus, vertical wind shear is defined as the difference in wind magnitude at 500- and 850-hPa levels. Low-level dewpoint depression is defined as the difference between 2-m temperature and dewpoint. A vertical humidity deviation is defined as the ratio between specific humidity at low- and midtropospheric levels. To ensure that the height of the planetary boundary layer (PBL) does not affect these results, midlevel humidity is taken as a mean value of 450 and 500 hPa, while low-level humidity is required to be within the PBL (e.g., 850 hPa). Three different humidity ratios are defined so that the environment is separated into three equally probable states: i) dry aloft, ii) mean, and iii) moist aloft. The “dry aloft” profile is defined as a state with a ratio of low- to midlevel atmospheric moisture greater than that of the domain mean profile, while the opposite is the case for “moist aloft” profile.
This study focuses on land regions in South America (10°S–10°N; 50°–80°W) and Africa (10°S–10°N; 17°W–30°E) of approximately equal area and similar surface type and elevations. While a detailed description of datasets used for the analysis can be found in the cited literature, a brief overview of TRMM instrument characteristics is provided below in addition to the climatology of the Amazon and African regions.
a. TRMM instruments
TRMM’s precipitation radar is a Ku-band radar operating at frequency of 13.8 GHz with a minimum detectable reflectivity of 17 dBZ. This provides reliable rain-rate detection down to 0.5 mm h−1 with limited sensitivity to frozen hydrometeors (e.g., fairly large ice particles such as hail and graupel). With a range resolution of 250 m, PR offers vertical sampling up to 20 km above the mean sea level through its attenuation-based retrieval (version 7; Iguchi et al. 2000, 2009). Each scan contains forty-nine 5-km field-of-view (FOV) pixels forming a swath of approximately 250 km.
TMI measures the microwave radiances with horizontal (H) and vertical (V) polarization at nine channels (10V/H, 19 V/H, 24, 37 V/H, and 85 V/H GHz) with footprint sizes ranging from 63 km × 37 km to 5 km × 7 km (Kummerow et al. 1998).
b. Climatology of precipitation over the Amazon and central and West African regions
Central Africa is known for deep, intense, well-organized storms that produce as much as 70% of total rainfall in that region. In the form of squall lines and mesoscale convective complexes (MCCs), these storms typically initiate near the Ethiopian highlands, the Darfur mountains, and Jos Plateau. While propagating toward the Atlantic, they are strongly affected by African easterly waves, the African easterly jet and moisture convergence in the lower troposphere. The average annual precipitation, ranging from 600 to 2000 mm, is unevenly distributed between the dry (October–March) and wet (April–September) seasons (Conway et al. 2009). Detailed descriptions of precipitation systems over central Africa can be found in studies by Payne and McGarry (1977), Bolton (1984), Tetzlaff and Peters (1988), Machado et al. (1993), Rowell and Milford (1993), Laing and Fritsch (1993), and Novella and Thiaw (2013). Two subregions, namely, the Sahel (northern Africa) and the Congo (equatorial Africa), are often characterized by somewhat different precipitation patterns. A comprehensive analysis of those differences is given by Laing et al. (2011), who emphasize a role that equatorial coupled waves play in development and life cycle of the MCSs in central Africa. They point out that, in comparison with their Sahel counterparts, equatorial MCSs are often exposed to slowing equatorial Kelvin waves, the Madden–Julian oscillation (MJO) signal, less continental mass, and weaker contribution of local shear. Therefore, while generally very similar, it may be expected that typical precipitation systems in these regions have different structures under the same local atmospheric conditions.
Amazon precipitation preferentially comes in the form of shallow, less intense but persistent systems with more oceanlike characteristics. A dominant feature of the rainfall variability is a diurnal cycle with seasonal contrasts less pronounced than over the African region. Most of the rainfall occurs between November and May. Mean annual accumulation over the majority of the area considered in this study ranges between 1000 and 3000 mm. Detailed analysis of Amazon rainfall systems, including occasional deeper, continental-like systems, is given by Petersen et al. (2002) and citations therein. A literature review of the most important precipitation properties in this region can be found in Silva Dias et al. (2002).
3. Understanding the origins of systematic difference in PMW retrieval
Because the PMW algorithm [Eqs. (1) and (2)] guarantees similar rain rates for similar Tb depressions at 85 GHz, an error is introduced any time the scattering in a cloud differs from the average assumed relation. Since the 85-GHz brightness temperature decreases primarily because of ice scattering (Vivekanandan et al. 1991), it is hypothesized that ice-scattering variability is responsible for the systematic differences in rain rates seen between the two sensors in Fig. 1. To test this, PR reflectivity above the freezing level and rain rate over the two regions of Fig. 1 are first compared and then linked to a proxy for PMW retrieval rain rate (i.e., 85-GHz Tb depression). Only for this purpose, to ensure that 85-GHz Tb depression is not caused by any surface-related sources, such as standing water, a polarization-corrected temperature (PCT) is used. Before the result is discussed, PCT and PR’s total reflectivity above the freezing level are defined.
The total reflectivity above freezing level (TRFL) is simply the sum total of all reflectivity values in range gates above the freezing level observed by PR. Because of a wide range of hydrometeor properties (e.g., type, phase, density, and size distribution) it provides only a qualitative estimate of the cloud content in the freezing portion of atmospheric column, which, because of PR’s sensitivity threshold (17 dBZ), mainly relates to large frozen hydrometeors. To estimate this quantity, the freezing level is obtained from the 2A25 PR product and used to locate PR range bins with frozen hydrometeors.
These two diagnostic variables allow for easier verification of the hypothesis that the variability in ice scattering is the dominant error source. Figure 4 shows the relationship between the TRFL and rain given by PR. While noisy, the TRFL in the cloud is clearly depressed over the Amazon and enhanced over the African region for a given rainfall rate.
Total reflectivity above the freezing level as a function of rain rate. A comparison of the means from the Amazon (blue), African (red), and the overall (black) regions (as defined in Fig. 1). Pixel data are given by diamonds; mean values for each rain-rate bin are given by the times signs. Note that only 5% of the pixels, randomly chosen, are plotted (the means account for full dataset) to avoid clutter in the figure.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Total reflectivity above the freezing level as a function of rain rate. A comparison of the means from the Amazon (blue), African (red), and the overall (black) regions (as defined in Fig. 1). Pixel data are given by diamonds; mean values for each rain-rate bin are given by the times signs. Note that only 5% of the pixels, randomly chosen, are plotted (the means account for full dataset) to avoid clutter in the figure.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Total reflectivity above the freezing level as a function of rain rate. A comparison of the means from the Amazon (blue), African (red), and the overall (black) regions (as defined in Fig. 1). Pixel data are given by diamonds; mean values for each rain-rate bin are given by the times signs. Note that only 5% of the pixels, randomly chosen, are plotted (the means account for full dataset) to avoid clutter in the figure.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Alternatively, one can also use the 250-K 85-GHz PCT depression from the PMW instrument as a proxy for ice scattering. Figure 5 depicts the relationship between this PCT depression and TRFL for collocated PMW and PR measurements. The comparison takes 1 yr of observations over the two regions marked in Fig. 1. Although not perfectly aligned, the two quantities are seen to be related. An increase in PCT depression corresponds to an increase in total reflectivity above the freezing level. Stronger scatter at the bottom of the plot is a reflection of the variability in ice particle size (Bennartz and Petty 2001) as well as in Tb caused by liquid-phase particles seen by the radar above the freezing level but not contributing to the PCT depression. As the PCT depression increases, the scatter reduces.
Relationship between the total reflectivity above the freezing level (PR) and 85-GHz Tb PCT 250-K depression (TMI).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Relationship between the total reflectivity above the freezing level (PR) and 85-GHz Tb PCT 250-K depression (TMI).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Relationship between the total reflectivity above the freezing level (PR) and 85-GHz Tb PCT 250-K depression (TMI).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Clearly, Eqs. (1) and (2) show that PMW rainfall depends on the 85-GHz temperature depression, which in Fig. 5 is seen to be proportional to the reflectivity observed above the freezing level (i.e., to the reflectivity related to frozen hydrometeors). At the same time, Fig. 3 suggests that the amplitude in the signal related to the presence of the ice in clouds, typical for Amazon and African regions, differs. This is consistent with the hypothesis that the ice aloft is indeed related to the variability in the PR-to-PMW rainfall difference over these two regions.
4. Addressing the variability of the ice aloft to rain-rate relation
The most striking property in Fig. 4 is a large pixel-to-pixel variability in the relationship between the total reflectivity above the freezing level and the surface rainfall. For moderate to high rain rates, the reflectivity can vary more than 15 dBZ (e.g., at 10 mm h−1 TRFL ranges from 45 to 58 dBZ). It is well known that clouds undergo substantial microphysical and thermodynamical changes through their life cycle (e.g., chapter 14 in Stull 2015; Cotton et al. 2011). The ice phase responds to these changes (Imaoka and Nakamura 2012). Therefore, the pixel-level variability in Fig. 4 is expected, as different life cycle stages are captured by TRMM’s random sampling. The goal of this study, however, is not to interpret this pixel-level variability. Instead, the focus is on understanding the separation of the two mean relations in Fig. 4. Considering the sample size (over 50 000 pixels) the variability of the relationship between the mean total reflectivity above the freezing level and the surface rainfall over the two regions is significant.
a. Stratiform–convective classification
The PR standard product 2A23 separates raining pixels into a number of categories, the majority of which (95%) fall into stratiform and convective classes. Categorization is made based on criteria such as presence of a bright band, precipitation depth, reflectivity value, and type of neighboring pixels (Awaka et al. 2009). The outcome is shown to be in good agreement with similar schemes applied to ground-based classifications. Thus, this pixel categorization is expected to efficiently recognize contrasts in vertical structures of stratiform and convective cloud types. Because of their different ice structures, systematic differences of these cloud types may be able to explain the observed regional differences in Fig. 4.
The PR reflectivity profiles of stratiform and convective pixels over the central African and Amazon regions are compared in Fig. 6. As expected, the two classes separate well. However, using the freezing-level height (4–5 km according to the bright band) as a reference, it is easy to notice that the reflectivity above the 0°C level is significantly larger over the African than the Amazon region for both convective as well as stratiform cloud structures. This result thus adds little to our existing result from Fig. 4 that Africa has systematically higher reflectivities in the ice region than South America.
Mean PR reflectivity profiles over 4 mm h−1 0.5° grid boxes within the two regions marked in Fig. 1. Both convective (red) and stratiform (blue) pixels tend to have stronger reflectivity over Africa than over the Amazon.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Mean PR reflectivity profiles over 4 mm h−1 0.5° grid boxes within the two regions marked in Fig. 1. Both convective (red) and stratiform (blue) pixels tend to have stronger reflectivity over Africa than over the Amazon.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Mean PR reflectivity profiles over 4 mm h−1 0.5° grid boxes within the two regions marked in Fig. 1. Both convective (red) and stratiform (blue) pixels tend to have stronger reflectivity over Africa than over the Amazon.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
b. Cloud cluster type classification
An alternative classification of clouds consists of the clustering method developed in studies by Elsaesser et al. (2010) and Elsaesser and Kummerow (2013). Elsaesser et al. (2010) classify precipitation regimes based on a cloud type probability density functions (PDFs), following the notion that regimes are characterized by distinct combinations of shallow, midlevel and deep convective cloud types (as discussed in Johnson et al. 1999). Using their approach, we tested self-similar PDFs of cloud characteristics allowing them to group into three clusters. In the original study, tropical oceanic clouds are clustered into classes defined by 1) shallow, 2) deep-organized, and 3) deep-unorganized convection. To mimic this, we first define a regular 1° × 1° grid along the TRMM track, and then seek cluster centroids, or “regimes,” that minimize the Euclidean distance in an x-dimensional space of standardized variables. The variables, following Elsaesser et al. (2010), are chosen to be 1) echo-top heights (ETHs), 2) mean convective rainfall rate, and 3) ratio of convective rainfall rate to overall rainfall rate, all given by PR over the 1° grids. The first are given by the altitude of the highest non-isolated range bins with reflectivity of at least 17 dBZ [more details are given in Short and Nakamura (2000)], with shallow systems dominated by clouds with ETHs less than 5 km, deep-organized systems with ETHs from 5 to 9 km, and deep-organized ones with ETHs greater than 9 km. The other two variables are simply based on the total number of pixels within each category. While the echo-top heights represent a proxy for the amount of ice in the cloud column, they also relate relatively well to the level of cloud system organization [e.g., deeper clouds yield more organization, as seen in Johnson et al. (1999)]. Application of the clustering algorithm results in the same cloud regimes described by Elsaesser et al. (2010) and Elsaesser and Kummerow (2013). The corresponding change of reflectivity with height and Tb values at 85 GHz of each regime are shown in Fig. 7.
As in Fig. 6, but separated by 1) precipitation regime: (a) shallow, (b) deep unorganized, and (c) deep organized; 2) precipitation type: convective (red) and stratiform (blue); and 3) region: Amazon (solid lines) and African (dashed lines).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
As in Fig. 6, but separated by 1) precipitation regime: (a) shallow, (b) deep unorganized, and (c) deep organized; 2) precipitation type: convective (red) and stratiform (blue); and 3) region: Amazon (solid lines) and African (dashed lines).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
As in Fig. 6, but separated by 1) precipitation regime: (a) shallow, (b) deep unorganized, and (c) deep organized; 2) precipitation type: convective (red) and stratiform (blue); and 3) region: Amazon (solid lines) and African (dashed lines).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Profile comparison suggests that, unlike convective–stratiform profiles, regime vertical structures are extremely consistent between Africa and South America. For example, if a vertical profile (convective or stratiform) in Fig. 7 is chosen to compare its extent and shape over the Amazon and African regions, very little variability will be found between the two, regardless of the chosen regime. Regional differences must therefore be related more to a change in regime frequency than in regime properties. High-frequency (i.e., 85 GHz) Tbs further show significantly lower values in the case of deep-organized cloud systems, implying enhanced ice content of this regime. This is consistent with higher reflectivities detected throughout the column of dBZ profile above the freezing level (i.e., brightband) in deep-organized systems (Fig. 7c) than in shallow ones (Fig. 7a). These results are generally consistent with clouds and cloud system properties described in the existing literature (e.g., Steiner et al. 1995; Houze et al. 1990).
c. Explaining the variability of scattering-signal-to-rain-rate relationship using cloud clusters
Consistency of the regimes with respect to ice aloft and Tb depressions between the two regions suggests that using the clustering approach has more potential to address the systematic discrepancies over the two regions than the simple stratiform–convective classification. Figure 8 shows the probability of occurrence of each regime along with the total rain fraction of that regime. Additionally, Fig. 9 shows the relative difference in total rainfall contribution separated by regime as well as the corresponding scatterplots showing the differences within each regime. The results suggest a strong correlation between the three cloud system types and rainfall differences seen in Fig. 1. While no obvious correlation is noticeable between the systematic differences and deep-unorganized regime (Fig. 9, middle), PMW negative deviations (purple areas in Fig. 1) coincide well with the relative frequency of occurrence (RFO) of the shallow systems (Fig. 8, top row) and PMW positive deviations (warm colors in Fig. 1) coincide with RFOs of deep-organized systems (Fig. 8, bottom row). The results are in agreement with those from studies that use a similar approach to describe tropical convection (e.g., Mohr et al. 1999; Zipser et al. 2006; Wall et al. 2013; Houze et al. 2015). For example, Mohr et al. (1999) found that well-organized storms, MCSs, in the African region constituted 10%–20% of the regional populations of convective systems but contributed 70%–80% of the rainfall. Zipser et al. (2006) show the distribution of intense thunderstorms that coincide well with the relative frequency of occurrence of deep-organized systems in Fig. 8. The same study found that precipitation distributions over the Amazon have relatively few intense storms, which relates to properties of the shallow systems seen here.
(left) RFO and (right) contribution to the total PR rain by (top) shallow, (middle) deep-unorganized, and (bottom) deep-organized precipitation regimes over the Amazon and African regions in 2010.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
(left) RFO and (right) contribution to the total PR rain by (top) shallow, (middle) deep-unorganized, and (bottom) deep-organized precipitation regimes over the Amazon and African regions in 2010.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
(left) RFO and (right) contribution to the total PR rain by (top) shallow, (middle) deep-unorganized, and (bottom) deep-organized precipitation regimes over the Amazon and African regions in 2010.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
(left) The relative difference between PR and PMW total rainfall contribution separated by precipitation regime. (right) Density scatterplots of the PMW and PR rainfall estimates (log axes) for each regime: the PWM sensor (top) underestimates PR by 33% in the shallow regime, (middle) underestimates PR by 10% in the deep-unorganized regime, and (bottom) overestimates PR by 41% in the deep-organized regime.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
(left) The relative difference between PR and PMW total rainfall contribution separated by precipitation regime. (right) Density scatterplots of the PMW and PR rainfall estimates (log axes) for each regime: the PWM sensor (top) underestimates PR by 33% in the shallow regime, (middle) underestimates PR by 10% in the deep-unorganized regime, and (bottom) overestimates PR by 41% in the deep-organized regime.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
(left) The relative difference between PR and PMW total rainfall contribution separated by precipitation regime. (right) Density scatterplots of the PMW and PR rainfall estimates (log axes) for each regime: the PWM sensor (top) underestimates PR by 33% in the shallow regime, (middle) underestimates PR by 10% in the deep-unorganized regime, and (bottom) overestimates PR by 41% in the deep-organized regime.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
By showing the relative difference between PMW and PR total rainfall estimate with corresponding rain-rate plots, Fig. 9 quantifies the contributions to the total systematic difference by each of the three regimes. Shallow and deep-unorganized regimes tend to be underestimated by the PMW sensor relative to the PR (by 33% and 10%, respectively), while the deep-organized regime rainfall is overestimated (by 41%). Most of the areas with positive differences in the top two panels of Fig. 9 correspond to regions where shallow and deep-unorganized regimes contribute less than 50% to the total rain. Combined information from Figs. 8 and 9 suggests that cloud systems RFO explains up to 50% of the systematic differences over the Amazon and African regions. The overall conclusion is that PMW sensor indeed tends to overestimate ice-rich deep-organized convection and underestimate the other two, relative to the PR. This, coupled the changes in the relative frequencies of occurrences of these systems, generally explains the systematic difference seen in Fig. 1.
The above conclusion is supported by Fig. 10, which uses the integrated reflectivity above the freezing level versus surface rainfall, as done for Fig. 4, to repeat the analysis. The figure clearly shows that deep-organized systems relate to positive deviations, while shallow systems relate to negative deviations from the mean PR’s ice-column-estimate-to-rain ratio of Fig. 4. This suggests that if one has knowledge on a type of a cloud system, then PMW-to-PR deviation of that system may be predicted, at least in the mean sense. Unfortunately, PR measurements are not always available to provide this information to PMW algorithms. Therefore, an alternative approach to link the observed storms to the systematic differences is desirable. Based on the current understanding of the interactions of storms with the environment, a potential solution exists if the environment can be used to predict cloud organization level (i.e., raining regime).
As in Fig. 4, but separated by the system type (shallow, deep unorganized, and deep organized).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
As in Fig. 4, but separated by the system type (shallow, deep unorganized, and deep organized).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
As in Fig. 4, but separated by the system type (shallow, deep unorganized, and deep organized).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
5. Cloud system types and the large-scale environment
Clusters form a convenient basis for understanding the relationship between systematic differences and the synoptic state of the atmosphere (also referred to as the “large-scale environment”). Shaped and governed by thermodynamical and microphysical processes, precipitation systems change and evolve during their life cycle. By grouping the storms into structurally self-similar systems, clusters form subsets of data that potentially have less pixel-to-pixel variability induced by these life cycle changes. Sampling the atmospheric conditions by criteria that are well known to play a key role in cloud development reveals links between the environment and the level of cloud system organization. Using findings from Mohr and Zipser (1996), Mohr et al. (1999), Petersen et al. (2002), and other above-mentioned studies, a relationship between the regimes and large-scale parameters is tested.
Atmospheric parameters are taken from ERA-Interim and collocated with the existing 1° × 1° raining scenes. To ensure that environmental variables are not affected by precipitation thermodynamics, the ERA-Interim data preceding the time of PR precipitation are used. Thus, the time gap between the environment state and rainfall observations can be as large as 6 h (temporal resolution of the 3D ERA-Interim). A number of parameters and their combinations [e.g., CAPE, total column water vapor (TPW), skin temperature, vertical and horizontal winds at 700 hPa, velocity of the midlevel jet, the magnitude of the low-level wind shear, and the surface equivalent potential temperature] are tested as cloud regime predictors. Several parameters stand out: the midlevel vertical wind shear, vertical humidity deviation, CAPE, and the low-level dewpoint depression. Results are presented in Fig. 11 and Table 1.
RFO of deep-organized (red), shallow (black), and deep-unorganized (blue) systems as a function of the environment: (a) vertical distribution of moisture, (b) low-level dew point depression, (c) CAPE, and (d) wind shear. For the exact values of RFOs and the environments, see Table 1.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
RFO of deep-organized (red), shallow (black), and deep-unorganized (blue) systems as a function of the environment: (a) vertical distribution of moisture, (b) low-level dew point depression, (c) CAPE, and (d) wind shear. For the exact values of RFOs and the environments, see Table 1.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
RFO of deep-organized (red), shallow (black), and deep-unorganized (blue) systems as a function of the environment: (a) vertical distribution of moisture, (b) low-level dew point depression, (c) CAPE, and (d) wind shear. For the exact values of RFOs and the environments, see Table 1.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Environment bin limits with the corresponding PR-to-PMW (TMI) rainfall estimate ratios and RFO of the three regimes for the year 2010. Note: the bin limits for the vertical distribution of humidity are given in the reversed order of how they appear in Fig. 11a (i.e., lower bin limit values correspond to the dry-aloft conditions).
The analyses show that high CAPE values, strong shear, and dry-aloft conditions are favorable precursors of deep-organized convection. In these environments, intense well-organized systems are 2 to 3 times more common than any others. Conversely, low CAPE, week shear, and a moist-aloft setup are conditions that favor the shallow regime. While the two more-organized regimes (shallow and deep-organized) are highly predictable, it is interesting to note that deep-unorganized systems show no, or very little, sensitivity to any of the tested variables. This may not be too detrimental since this is also the regime that showed the least disagreement in rain estimates between the two sensors. The role of the environment parameters favoring specific cloud regimes is examined next.
a. Wind shear and convective available potential energy
A role of CAPE and wind shear in cloud development is often coupled and as such broadly discussed in the literature. As a measure of atmospheric column potential energy, CAPE is widely used to predict convection intensity and longevity. Whether only a part or most of the column energy is exploited in the cloud development process depends on a number of factors where wind shear plays an important role (e.g., Xu 1992; Xu and Moncrieff 1994). Studies and theoretical models of cloud system organization describe a necessary balance between the strength of the updrafts and their vertical tilts to allow for deep-organized features such as MCSs. By defining a displacement of downwelling motions relative to the updrafts, wind shear controls the ability of precipitating system to utilize environment resources (e.g., CAPE). While downward motion of the hydrometeors straight through the updraft limits the storm’s potential to propagate and feed on the unperturbed unstable environment, too much wind shear may force the rainwater into dryer layers away of the storm’s core and cause evaporation that can stabilize the atmospheric column too fast (e.g., Rotunno et al. 1988; LeMone et al. 1998). Whether cold pool, updraft intensity, front and rear inflows, or in-cloud thermodynamical processes are more favorable for one cloud regime over another is a complex question that cannot be simply answered by CAPE and wind shear alone. However, these two variables have a strong influence on all of these factors and therefore serve as good predictors of a storm’s organization level.
b. Low-level dewpoint depression
Using a low-level dewpoint depression as a cloud system type predictor is motivated by the facts that this quantity simultaneously provides information on the low-level relative humidity and state of the soil moisture, both of which have been shown (e.g., Ford et al. 2015) to relate to the cloud system initiation and development. Near the surface relative humidity is a relatively good proxy for the state of the boundary layer. Soil moisture and 2-m temperature are also related to the height of the boundary layer, which plays an important role in defining the cloud-base height. This further influences the depth of the cloud determining thermodynamical properties over the course of the cloud life cycle. At the same time, the relative humidity and cloud-base height strongly affect the amount of rainwater evaporated before reaching the ground. This does not only alter the rain rate but also may play an important role in further development and organization of the cloud system through the downdraft bursts and cold pools. Findings of Ek and Mahrt (1994) offer an example of the complex response of cloud properties to the top of a boundary layer. They found that drier soil leads not only to lower boundary layer specific humidity but to cooler temperatures at the boundary layer top because of greater boundary layer growth. When the latter effect dominates, the relative humidity at the boundary layer top is greater over drier soil. In contrast, they saw drier soil leading to lower relative humidity at the boundary layer top when the air above is strongly stratified or very dry.
Clearly, large-scale environments can serve as predictors of a cloud system type. If relationships between the environment, system types, and PMW systematic differences are consistent and robust, then the environment itself must have a well-defined relationship to these differences as well.
6. Potentials of removing PMW climatological errors
The results, to this point, demonstrate that systematic deviations of PMW rainfall retrieval relative to PR rain estimates over land are caused by regionally dependent differences of the ratio between the ice scattering signature and surface rainfall, characterized by strong pixel-level variability. While this localized variability is hard to capture, grouping the pixels into classes (clusters) of distinct mean PMW-to-PR deviations allowed for their difference reduction at large scales.
It was shown that clusters relate well with PMW-to-PR rainfall ratio and that the large-scale environment is a reasonably good predictor of the cluster types. Therefore, it is expected that both can perform well if used to predict their disagreement. To demonstrate these predictors’ potential in reducing PMW-to-PR differences, a simple experiment is performed. A year of TRMM data is used to quantify the relationship between rain estimate of the two sensors (e.g., that seen in Fig. 1), with respect to 1) environment and 2) clusters. For each given environment or cluster, the PMW-to-PR rainfall ratio is recorded. Once available, this ratio is used to adjust the retrieved PMW rain-rate estimates of an independent time interval when similar environmental or cluster conditions exist. The results are presented below.
a. Large-scale environment as a variability predictor
Before any PMW rainfall adjustments are made, Fig. 12 presents the mean PMW-to-PR ratios as a function of the environment categories used in Fig. 11. Clearly, a robust relationship exists, supporting the original hypothesis that the large-scale environment relates to differences between PMW and PR rainfall estimates. Table 2 lists correlations between the environments, suggesting that none of the four is overly linked to the others.
PR-to-PMW rainfall ratio as a function of large-scale environment during 2010. The environmental bins are as in Fig. 11. The bin limits and exact values of PR-to-PMW rainfall ratios are listed in Table 1.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
PR-to-PMW rainfall ratio as a function of large-scale environment during 2010. The environmental bins are as in Fig. 11. The bin limits and exact values of PR-to-PMW rainfall ratios are listed in Table 1.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
PR-to-PMW rainfall ratio as a function of large-scale environment during 2010. The environmental bins are as in Fig. 11. The bin limits and exact values of PR-to-PMW rainfall ratios are listed in Table 1.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Correlation between the large-scale environments of the four categories seen in Fig. 12. Data source: ERA-Interim data for 2010.
Multiple environment predictors of the PMW-to-PR rainfall ratio are also tested at once. Figure 13 depicts an example where CAPE-defined environment is sampled by low-level dewpoint depression. As the dewpoint depression decreases, an increase in the ratio of the two instruments’ rainfall is seen across each of the CAPE environment bins. Similar results are seen for any combination of the environments (not shown here).
Ratio between PR and PMW rainfall as a function of the environment defined by CAPE and the difference between 2-m temperature and dewpoint. Red bars denote bins in which PR rainfall is underestimated by PMW retrieval, while blue bars denote the opposite.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Ratio between PR and PMW rainfall as a function of the environment defined by CAPE and the difference between 2-m temperature and dewpoint. Red bars denote bins in which PR rainfall is underestimated by PMW retrieval, while blue bars denote the opposite.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Ratio between PR and PMW rainfall as a function of the environment defined by CAPE and the difference between 2-m temperature and dewpoint. Red bars denote bins in which PR rainfall is underestimated by PMW retrieval, while blue bars denote the opposite.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
Finally, to quantify the predictability of the two sensor’s disagreements, Fig. 14 compares the original (black) and adjusted (red) PMW estimates of mean daily rain rates with those of PR, for year 2008, using level-3 products at 10° resolution. The adjustments of PMW rain estimate are based on the 2010 dataset constraining the PMW-to-PR ratio by two environments at a time. Using the observed CAPE and wind shear as criteria (Fig. 14, left), improvements of approximately 30% and 35% are made in RMSE and systematic difference of daily rain rates, respectively, while regression coefficient is improved by 25%. Improvements are somewhat less appealing, but still significant, when CAPE is used in combination with the humidity distribution, removing approximately 20% of the relative bias. When individual grids are compared, with no exceptions, improvement is found across the entire domain (not shown here).
The PMW-to-PR conditional rainfall comparison before (black) and after (red) PMW rainfall estimate adjustment. Scenes observed by both sensors as nonraining are excluded. A combination of (left) CAPE and shear values and (right) CAPE and humidity distribution is used to predict and then reduce the PR-to-PMW retrieval difference in the GPROF algorithm on a 10° × 10° grid for the regions marked in Fig. 1.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
The PMW-to-PR conditional rainfall comparison before (black) and after (red) PMW rainfall estimate adjustment. Scenes observed by both sensors as nonraining are excluded. A combination of (left) CAPE and shear values and (right) CAPE and humidity distribution is used to predict and then reduce the PR-to-PMW retrieval difference in the GPROF algorithm on a 10° × 10° grid for the regions marked in Fig. 1.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
The PMW-to-PR conditional rainfall comparison before (black) and after (red) PMW rainfall estimate adjustment. Scenes observed by both sensors as nonraining are excluded. A combination of (left) CAPE and shear values and (right) CAPE and humidity distribution is used to predict and then reduce the PR-to-PMW retrieval difference in the GPROF algorithm on a 10° × 10° grid for the regions marked in Fig. 1.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
b. Cloud system type as a variability predictor
Section 4c and Fig. 9 depict the PMW-to-PR rain ratio as a function of a cloud system type. The performance of the cloud system type in PMW systematic difference removal is evaluated in Fig. 15 by repeating the same test as above. The RMSE of adjusted rain rates yields a value of 1.9 mm day−1, corresponding to approximately 40% of improvement relative to the original value of 3.4 mm day−1. At same time, systematic differences over the two regions are reduced by almost 50%. Clearly, a significant reduction of both RMSE and bias confirms a strong relation between the bias-to-cloud-structure to PMW systematic differences in rainfall estimates over land.
As in Fig. 14, but using the cloud system type as a predictor of the PMW-to-PR retrieval deviation.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
As in Fig. 14, but using the cloud system type as a predictor of the PMW-to-PR retrieval deviation.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
As in Fig. 14, but using the cloud system type as a predictor of the PMW-to-PR retrieval deviation.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0174.1
When comparing the two predictors, cloud system types explain more variability between the two sensors. However, their definition was based on radar characteristics and thus may not be easily applied to radiometer-only retrievals.
7. Summary
This paper tries to provide a better understanding of systematic differences seen in PMW rainfall retrievals over land. The study explores the links between ice scattering PMW signature and estimate of surface rain intensity, cloud system structure, and large-scale environments. It is shown that observed cloud physics and relationship between Tb depression and surface rain intensity correlate well with regional PMW-to-PR rainfall discrepancies in tropical Africa and South America. Variability of ice-scattering-signal-to-rain-rate relationship across these two opposing regions is captured by grouping the pixel-level data into three self-similar cloud classes of distinct levels of organization. Although these groups showed great potential in removing systematic differences seen between PMW and PR rainfall estimates, their diagnosis is too complex for PMW retrieval applications. As a natural driver of atmospheric processes, the role of a large-scale environment in defining these distinct levels of cloud organization is evaluated. When constrained by CAPE, wind shear, dewpoint depression, and vertical humidity distribution, the environment is found to be in favor to a certain storm types. Thus, high CAPE values, as well as dry-aloft conditions, are most commonly seen prior to deep-organized systems. On the other side, low wind shear and weak dewpoint depression are both in favor of shallower unorganized events. The ability of large-scale environments to reduce climate-scale PMW-to-PR rainfall differences is found to be appealing, lowering the current PMW-to-PR regional rainfall ratios by up to 40%.
The possibility is left open that other predictors, or combination of predictors, could be used to further improve upon these results, overcoming the lack of information that the observed vector currently offers to PMW retrievals over land. It is concluded that addressing the role of the cloud structure variability in PWM observations will be an inevitable step in future versions of the PMW algorithms.
Acknowledgments
This research was supported by NASA Earth and Space Science Fellowship 2015 (NESSF15R) and PMM Grant NNX13AG31G. The authors acknowledge Dr. Gregory Elsaesser (NASA Godard Institute for Space Studies and Department of Applied Physics and Applied Mathematics, Columbia University), David Duncan (CSU), and Dr. Gavin Roy for their helpful comments and suggestions.
REFERENCES
Adler, R. F., G. J. Huffman, and P. R. Keehn, 1994: Global tropical rain estimates from microwave-adjusted geosynchronous IR data. Remote Sens. Rev., 11, 125–152, doi:10.1080/02757259409532262.
Adler, R. F., G. Gu, and G. J. Huffman, 2012: Estimating climatological bias errors for the Global Precipitation Climatology Project (GPCP). J. Appl. Meteor. Climatol., 51, 84–99, doi:10.1175/JAMC-D-11-052.1.
Awaka, J., T. Iguchi, and K. Okamoto, 2009: TRMM PR standard algorithm 2A23 and its performance on bright band detection. J. Meteor. Soc. Japan, 87A, 31–52, doi:10.2151/jmsj.87A.31.
Ba, M. B., D. Rosenfeld, and A. Gruber, 1998: AVHRR multispectral derived cloud parameters: Relationships to microwave scattering signature and to cloud-to-ground lightning. Preprints, Ninth Conf. on Satellite Meteorology and Oceanography, Paris, France, Amer. Meteor. Soc., 408–411.
Barrett, E. C., 1970: The estimation of monthly rainfall from satellite data. Mon. Wea. Rev., 98, 322–327, doi:10.1175/1520-0493(1970)098<0322:TEOMRF>2.3.CO;2.
Barth, M. C., and Coauthors, 2015: The Deep Convective Clouds and Chemistry (DC3) field campaign. Bull. Amer. Meteor. Soc., 96, 1281–1309, doi:10.1175/BAMS-D-13-00290.1.
Bennartz, R., and G. W. Petty, 2001: The sensitivity of microwave remote sensing observations of precipitation to ice particle size distributions. J. Appl. Meteor., 40, 345–364, doi:10.1175/1520-0450(2001)040<0345:TSOMRS>2.0.CO;2.
Berg, W., T. L’Ecuyer, and C. Kummerow, 2006: Rainfall climate regimes: The relationship of regional TRMM rainfall biases to the environment. J. Appl. Meteor. Climatol., 45, 434–454, doi:10.1175/JAM2331.1.
Berg, W., T. L’Ecuyer, and S. C. van den Heever, 2008: Evidence for the impact of aerosols on the onset and microphysical properties of rainfall from a combination of active and passive satellite sensors. J. Geophys. Res., 113, D14S23, doi:10.1029/2007JD009649.
Bolton, D., 1984: Generation and propagation of African squall lines. Quart. J. Roy. Meteor. Soc., 110, 695–721, doi:10.1002/qj.49711046509.
Conner, M. D., and G. W. Petty, 1998: Validation and intercomparison of SSM/I rain-rate retrieval methods over the continental United States. J. Appl. Meteor., 37, 679–700, doi:10.1175/1520-0450(1998)037<0679:VAIOSI>2.0.CO;2.
Conway, D., A. Persechino, S. Ardoin-Bardin, H. Hamandawana, C. Dieulin, and G. Mahé, 2009: Rainfall and water resources variability in sub-Saharan Africa during the twentieth century. J. Hydrometeor., 10, 41–59, doi:10.1175/2008JHM1004.1.
Cotton, W. R., G. Bryan, and S. C. van den Heever, 2011: Cumulus clouds. Storm and Cloud Dynamics, 2nd ed. International Geophysics Series, Vol. 99, Academic Press, 243–302.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/qj.828.
Ek, M., and L. Mahrt, 1994: Daytime evolution of relative humidity at the boundary layer top. Mon. Wea. Rev., 122, 2709–2721, doi:10.1175/1520-0493(1994)122<2709:DEORHA>2.0.CO;2.
Elsaesser, G. S., and C. D. Kummerow, 2013: A multisensor observational depiction of the transition from light to heavy rainfall on subdaily time scales. J. Atmos. Sci., 70, 2309–2324, doi:10.1175/JAS-D-12-0210.1.
Elsaesser, G. S., C. D. Kummerow, T. S. L’Ecuyer, Y. N. Takayabu, and S. Shige, 2010: Observed self-similarity of precipitation regimes over the tropical oceans. J. Climate, 23, 2686–2698, doi:10.1175/2010JCLI3330.1.
Ferraro, R. R., 1997: Special sensor microwave imager derived global rainfall estimates for climatological applications. J. Geophys. Res., 102, 16 715–16 735, doi:10.1029/97JD01210.
Ferraro, R. R., and G. F. Marks, 1995: The development of SSM/I rain-rate retrieval algorithms using ground-based radar measurements. J. Atmos. Oceanic Technol., 12, 755–770, doi:10.1175/1520-0426(1995)012<0755:TDOSRR>2.0.CO;2.
Ferraro, R. R., N. C. Grody, and G. F. Marks, 1994: Effects of surface conditions on rain identification using the DMSP-SSM/I. Remote Sens. Rev., 11, 195–209, doi:10.1080/02757259409532265.
Ferraro, R. R., E. A. Smith, W. Berg, and G. J. Huffman, 1998: A screening methodology for passive microwave precipitation retrieval algorithms. J. Atmos. Sci., 55, 1583–1600, doi:10.1175/1520-0469(1998)055<1583:ASMFPM>2.0.CO;2.
Ferraro, R. R., and Coauthors, 2013: An evaluation of microwave land surface emissivities over the continental United States to benefit GPM-era precipitation algorithms. IEEE Trans. Geosci. Remote Sens., 51, 378–398, doi:10.1109/TGRS.2012.2199121.
Ford, T. W., A. D. Rapp, and S. M. Quiring, 2015: Does afternoon precipitation occur preferentially over dry or wet soils in Oklahoma? J. Hydrometeor., 16, 874–888, doi:10.1175/JHM-D-14-0005.1.
Gilmore, M. S., and L. J. Wicker, 2002: Influences of the local environment on supercell cloud-to-ground lightning, radar characteristics, and severe weather on 2 June 1995. Mon. Wea. Rev., 130, 2349–2372, doi:10.1175/1520-0493(2002)130<2349:IOTLEO>2.0.CO;2.
Gopalan, K., N.-Y. Wang, R. Ferraro, and C. Liu, 2010: Status of the TRMM 2A12 land precipitation algorithm. J. Atmos. Oceanic Technol., 27, 1343–1354, doi:10.1175/2010JTECHA1454.1.
Grody, N. C., 1991: Classification of snow cover and precipitation using the special sensor microwave imager. J. Geophys. Res., 96, 7423–7435, doi:10.1029/91JD00045.
Haddad, Z. S., D. A. Short, S. L. Durden, E. Im, S. Hensley, M. B. Grable, and R. A. Black, 1997a: A new parametrization of the rain drop size distribution. IEEE Trans. Geosci. Remote Sens., 35, 532–539, doi:10.1109/36.581961.
Haddad, Z. S., E. A. Smith, C. D. Kummerow, T. Iguchi, M. R. Farrar, S. L. Durden, M. Alves, and W. S. Olson, 1997b: The TRMM ‘day-1’ radar/radiometer combined rain-profiling algorithm. J. Meteor. Soc. Japan, 75, 799–809.
Hollinger, J. P., J. L. Peirce, and G. A. Poe, 1990: SSM/I instrument evaluation. IEEE Trans. Geosci. Remote Sens., 28, 781–790, doi:10.1109/36.58964.
Hou, A., and Coauthors, 2014: The Global Precipitation Measurement Mission. Bull. Amer. Meteor. Soc., 95, 701–722, doi:10.1175/BAMS-D-13-00164.1.
Houze, R. A., B. F. Smull, and P. Dodge, 1990: Mesoscale organization of springtime rainstorms in Oklahoma. Mon. Wea. Rev., 118, 613–654, doi:10.1175/1520-0493(1990)118<0613:MOOSRI>2.0.CO;2.
Houze, R. A., S. Brodzik, C. Schumacher, S. E. Yuter, and C. R. Williams, 2004: Uncertainties in oceanic radar rain maps at Kwajalein and implications for satellite validation. J. Appl. Meteor., 43, 1114–1132, doi:10.1175/1520-0450(2004)043<1114:UIORRM>2.0.CO;2.
Houze, R. A., K. L. Rasmussen, M. D. Zuluaga, and S. R. Brodzik, 2015: The variable nature of convection in the tropics and subtropics: A legacy of 16 years of the Tropical Rainfall Measuring Mission satellite. Rev. Geophys., 53, 994–1021, doi:10.1002/2015RG000488.
Huffman, G. J., D. T. Bolvin, D. Braithwaite, K. Hsu, R. Joyce, and P. Xie, 2014: GPM Integrated Multi-SatellitE Retrievals for GPM (IMERG). Algorithm Theoretical Basis Doc., version 4.4, PPS, NASA/GSFC, 30 pp. [Available online at http://pmm.nasa.gov/sites/default/files/document_files/IMERG_ATBD_V4.4.pdf.]
Huffman, G. J., D. T. Bolvin, and E. J. Nelkin, 2015: Integrated Multi-satellitE Retrievals for GPM (IMERG). Tech. Doc. NASA/GSFC Code 612, 47 pp. [Available online at http://pmm.nasa.gov/sites/default/files/document_files/IMERG_doc.pdf.]
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, doi:10.1175/1520-0450(2001)040<2038:RPAFTT>2.0.CO;2.
Iguchi, T., T. Kozu, J. Kwiatkowski, R. Meneghini, J. Awaka, and K. Okamoto, 2009: Uncertainties in the rain profiling algorithm for the TRMM Precipitation Radar. J. Meteor. Soc. Japan, 87A, 1–30, doi:10.2151/jmsj.87A.1.
Imaoka, K., and K. Nakamura, 2012: Statistical analysis of the life cycle of isolated tropical cold cloud systems using MTSAT-1R and TRMM data. Mon. Wea. Rev., 140, 3552–3572, doi:10.1175/MWR-D-11-00364.1.
Johnson, R. H., T. M. Rickenbach, S. A. Rutledge, P. E. Ciesielski, and W. H. Schubert, 1999: Trimodal characteristics of tropical convection. J. Climate, 12, 2397–2418, doi:10.1175/1520-0442(1999)012<2397:TCOTC>2.0.CO;2.
Kim, M.-J., J. A. Weinman, and R. A. Houze Jr., 2004: Validation of maritime rainfall retrievals from the TRMM-microwave radiometer. J. Appl. Meteor., 43, 847–859, doi:10.1175/1520-0450(2004)043<0847:VOMRRF>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, doi:10.1175/1520-0426(1998)015<0809:TTRMMT>2.0.CO;2.
Kummerow, C., and Coauthors, 2000: The status of the Tropical Rainfall Measuring Mission (TRMM) after two years in orbit. J. Appl. Meteor., 39, 1965–1982, doi:10.1175/1520-0450(2001)040<1965:TSOTTR>2.0.CO;2.
Kummerow, C., and Coauthors, 2001: The evolution of the Goddard profiling algorithm (GPROF) for rainfall estimation from passive microwave sensors. J. Appl. Meteor., 40, 1801–1820, doi:10.1175/1520-0450(2001)040<1801:TEOTGP>2.0.CO;2.
Kummerow, C., S. Ringerud, J. Crook, D. Randel, and W. Berg, 2011: An observationally generated a priori database for microwave rainfall retrievals. J. Atmos. Oceanic Technol., 28, 113–130, doi:10.1175/2010JTECHA1468.1.
Kummerow, C., D. L. Randel, M. Kulie, N.-Y. Wang, R. Ferraro, S. J. Munchak, and V. Petkovic, 2015: The evolution of the Goddard profiling algorithm to a fully parametric scheme. J. Atmos. Oceanic Technol., 32, 2265–2280, doi:10.1175/JTECH-D-15-0039.1.
Laing, A. G., and J. M. Fritsch, 1993: Mesoscale convective complexes in Africa. Mon. Wea. Rev., 121, 2254–2263, doi:10.1175/1520-0493(1993)121<2254:MCCIA>2.0.CO;2.
Laing, A. G., R. E. Carbone, and V. Levizzani, 2011: Cycles and propagation of deep convection over equatorial Africa. Mon. Wea. Rev., 139, 2832–2853, doi:10.1175/2011MWR3500.1.
LeMone, M. A., E. J. Zipser, and S. B. Trier, 1998: The role of environmental shear and thermodynamic conditions in determining the structure and evolution of mesoscale convective systems during TOGA COARE. J. Atmos. Sci., 55, 3493–3518, doi:10.1175/1520-0469(1998)055<3493:TROESA>2.0.CO;2.
Liu, J., C. D. Kummerow, and G. S. Elsaesser, 2017: Identifying and analysing uncertainty structures in the TRMM Microwave Imager precipitation product. Int. J. Remote Sens., 38, 23–42, doi:10.1080/01431161.2016.1259676.
Machado, L. A. T., J.-P. Duvel, and M. Desbois, 1993: Diurnal variations and modulations by easterly waves of the size distribution of convective cloud clusters over West Africa and the Atlantic Ocean. Mon. Wea. Rev., 121, 37–49, doi:10.1175/1520-0493(1993)121<0037:DVAMBE>2.0.CO;2.
Maggioni, V., M. R. P. Sapiano, and R. F. Adler, 2016: Estimating uncertainties in high-resolution satellite precipitation products: Systematic or random error? J. Hydrometeor., 17, 1119–1129, doi:10.1175/JHM-D-15-0094.1.
McCollum, J. R., and R. R. Ferraro, 2003: Next generation of NOAA/NESDIS TMI, SSM/I, and AMSR-E microwave land rainfall algorithms. J. Geophys. Res., 108, 8382, doi:10.1029/2001JD001512.
McCollum, J. R., A. Gruber, and M. B. Ba, 2000: Discrepancy between gauges and satellite estimates of rainfall in equatorial Africa. J. Appl. Meteor., 39, 666–679, doi:10.1175/1520-0450-39.5.666.
Meyers, P. C., R. R. Ferraro, and N.-Y. Wang, 2015: Updated screening procedures for GPROF2010 over land: Utilization for AMSR-E. J. Atmos. Oceanic Technol., 32, 1015–1028, doi:10.1175/JTECH-D-14-00149.1.
Mohr, K. I., and E. J. Zipser, 1996: Mesoscale convective systems defined by their 85-GHz ice scattering signature: Size and intensity comparison over tropical oceans and continents. Mon. Wea. Rev., 124, 2417–2437, doi:10.1175/1520-0493(1996)124<2417:MCSDBT>2.0.CO;2.
Mohr, K. I., J. S. Famiglietti, and E. J. Zipser, 1999: The contribution to tropical rainfall with respect to convective system type, size, and intensity estimated from the 85-GHz ice-scattering signature. J. Appl. Meteor., 38, 596–606, doi:10.1175/1520-0450(1999)038<0596:TCTTRW>2.0.CO;2.
Novella, N. S., and W. M. Thiaw, 2013: African Rainfall Climatology version 2 for famine early warning systems. J. Appl. Meteor. Climatol., 52, 588–606, doi:10.1175/JAMC-D-11-0238.1.
Olson, W. O., and Coauthors, 2006: Precipitation and latent heating distributions from satellite passive microwave radiometry. Part I: Improved method and uncertainties. J. Appl. Meteor. Climatol., 45, 702–720, doi:10.1175/JAM2369.1.
Payne, S. W., and M. M. McGarry, 1977: The relationship of satellite inferred convective activity to easterly waves over West Africa and the adjacent ocean during Phase III of GATE. Mon. Wea. Rev., 105, 413–420, doi:10.1175/1520-0493(1977)105<0413:TROSIC>2.0.CO;2.
Petersen, W. A., S. W. Nesbitt, R. J. Blakeslee, R. Cifelli, P. Hein, and S. A. Rutledge, 2002: TRMM observations of intraseasonal variability in convective regimes over the Amazon. J. Climate, 15, 1278–1294, doi:10.1175/1520-0442(2002)015<1278:TOOIVI>2.0.CO;2.
Petty, G. W., and K. Li, 2013: Improved passive microwave retrievals of rain rate over land and ocean. Part I: Algorithm description. J. Atmos. Oceanic Technol., 30, 2493–2508, doi:10.1175/JTECH-D-12-00144.1.
Qie, X., R. Toumi, and Y. Zhou, 2003: Lightning activity on the central Tibetan Plateau and its response to convective available potential energy. Chin. Sci. Bull., 48, 296–299.
Rasmussen, K. L., M. M. Chaplin, M. D. Zuluaga, and R. A. Houze Jr., 2016: Contribution of extreme convective storms to rainfall in South America. J. Hydrometeor., 17, 353–367, doi:10.1175/JHM-D-15-0067.1.
Rotunno, R., J. B. Klemp, and M. L. Weisman, 1988: A theory for strong, long-lived squall lines. J. Atmos. Sci., 45, 463–485, doi:10.1175/1520-0469(1988)045<0463:ATFSLL>2.0.CO;2.
Rowell, D. P., and J. R. Milford, 1993: On the generation of African squall lines. J. Climate, 6, 1181–1193, doi:10.1175/1520-0442(1993)006<1181:OTGOAS>2.0.CO;2.
Schumacher, C., and R. A. Houze Jr., 2000: Comparison of radar data from the TRMM satellite and Kwajalein oceanic validation site. J. Appl. Meteor., 39, 2151–2164, doi:10.1175/1520-0450(2001)040<2151:CORDFT>2.0.CO;2.
Seo, E.-K., B.-J. Sohn, and G. Liu, 2007: How TRMM precipitation radar and microwave imager retrieved rain rates differ. Geophys. Res. Lett., 34, L24803, doi:10.1029/2007GL032331.
Seto, S., N. Takahashi, and T. Iguchi, 2005: Rain/no-rain classification methods for microwave radiometer observations over land using statistical information for brightness temperatures under no-rain conditions. J. Appl. Meteor., 44, 1243–1259, doi:10.1175/JAM2263.1.
Shige, S., H. Sasaki, K. Okamoto, and T. Iguchi, 2006: Validation of rainfall estimates from the TRMM precipitation radar and microwave imager using a radiative transfer model: 1. Comparison of the version-5 and -6 products. Geophys. Res. Lett., 33, L13803, doi:10.1029/2006GL026350.
Short, D. A., and K. Nakamura, 2000: TRMM radar observations of shallow precipitation over the tropical oceans. J. Climate, 13, 4107–4124, doi:10.1175/1520-0442(2000)013<4107:TROOSP>2.0.CO;2.
Silva Dias, M. A. F., and Coauthors, 2002: Cloud and rain processes in a biosphere-atmosphere interaction context in the Amazon region. J. Geophys. Res., 107, 8072, doi:10.1029/2001JD000335.
Spencer, R. W., H. M. Goodman, and R. E. Hood, 1989: Precipitation retrieval over land and ocean with the SSM/I: Identification and characteristics of the scattering signal. J. Atmos. Oceanic Technol., 6, 254–273, doi:10.1175/1520-0426(1989)006<0254:PROLAO>2.0.CO;2.
Steiner, M., R. A. Houze Jr., and S. E. Yuter, 1995: Climatological characterization of three-dimensional storm structure from operational radar and rain gauge data. J. Appl. Meteor., 34, 1978–2007, doi:10.1175/1520-0450(1995)034<1978:CCOTDS>2.0.CO;2.
Stolz, D. C., S. A. Rutledge, and J. R. Pierce, 2015: Simultaneous influences of thermodynamics and aerosols on deep convection and lightning in the tropics. J. Geophys. Res. Atmos., 120, 6207–