Removing Interfering Signals in Spaceborne Radar Data for Precipitation Detection at Very High Altitudes
1. Introduction
Local precipitation climatology can be improved through long-term satellite precipitation observations (e.g., Hirose and Okada 2018). For high-accuracy precipitation estimation, it is essential to reduce unwanted signals, such as interference signals, and to gain more insight into region-specific precipitation retrieval. The geographical characteristics and Earth’s surface condition (e.g., soil moisture and land use status) are important for characterizing local climatology. Active sensors are useful for detecting localized precipitation characteristics, especially over land (Roca et al. 2021). The Tropical Rainfall Measuring Mission (TRMM; Kummerow et al. 1998) operated from 1997 to 2015 and carried the first spaceborne Precipitation Radar (PR; Kozu et al. 2001). This was followed by the Global Precipitation Measurement (GPM; Hou et al. 2014) mission, which has been operating since 2014 and has a Dual-Frequency Precipitation Radar (DPR; Kojima et al. 2012; Iguchi 2020). Together, the TRMM PR and GPM DPR have provided large volumes of three-dimensional (3D) precipitation data for more than two decades (Nakamura 2021). Several validation studies have indicated the presence of data retrieval issues such as the nonuniform beam filling effect and differences in drop size distributions (e.g., Petersen et al. 2020; Iguchi et al. 2009; Liao and Meneghini 2022). Attempts to reduce systematic errors are still ongoing, and spatially fixed artificial echoes have been reported. Hamada and Takayabu (2014) found that extraordinary downward-increasing precipitation profiles were interfered by mainlobe clutter in the TRMM PR version 7 product and, hence, developed a filter to remove the contamination. This filter reduces precipitation by several percent over specific mountainous regions (Hirose et al. 2017).
Some contamination of PR data can be attributed to ground radio transmission (Hirose and Okada 2018). Radio interference can be observed as an increase in received power levels at all altitudes, which can be attributed to an increase in apparent system noise. The International Telecommunication Union–Radiocommunication Sector reported that TRMM PR operated at 13.8 GHz on a secondary basis within the band 13.75–14 GHz, and it was not protected from the primary services allocated to this band. In particular, TRMM PR stopped transmitting radio waves when passing over a specific arid area in Australia to avoid radio interference. Thus, GPM DPR uses a slightly different frequency (ITU-R 2021).
Transmissions from fixed-satellite services could be one source of errors in TRMM PR (ITU-R 1994, 2003). For example, the European Telecommunications Satellite Organization (Eutelsat) and Arab Satellite Communications Organization (Arabsat) operate Earth stations at the same Ku-band-frequency range and have large antennas to meet the constraints imposed by the ITU radio regulations (Eutelsat 2022; Alandjani 2012). For Eutelsat, deliberate jamming of broadcast signals in the Middle East has become a major issue (Eutelsat 2012). According to ITU-R (1994), such localized signals are generally weak and should not materially affect the overall statistics. However, as data accumulate, there is an increasing demand for the ability to detect microscale regional features and minute temporal changes. The removal of interferences was originally implemented in the standard algorithm for TRMM PR, and the TRMM Precipitation Radar Team (2011) described some areas in which TRMM PR experienced radio interference. The standard algorithm places warning flags at locations where the apparent system noise level exceeds the limit. However, such warning flags are unlikely to be able to eliminate contamination from radio interference based on their design; thus, the influence remained in TRMM PR version 7 (Hirose and Okada 2018). This issue is being addressed in the latest products, but has not been fully evaluated. Further investigation of unnatural data with interference is required to improve the accuracy of extracting high-resolution temporal and spatial changes in precipitation from long-term observation data.
After TRMM ended its operation, GPM DPR continued with 3D precipitation observations (Hou et al. 2014). GPM DPR has been upgraded to capture precipitation more precisely. The risk of radio interference for the Ku-band precipitation radar (KuPR) data of GPM DPR is reduced by changing the operation frequency of the radar, but it potentially includes the same issue of radio interference. Furthermore, DPR adopted the variable pulse repetition frequency (VPRF) technique to increase the number of samplings and adjust the observation window relative to the satellite (Kojima et al. 2012). This is because the GPM satellite has a greater variation in altitude relative to Earth’s surface with latitude than the TRMM satellite due to the elliptical shape of Earth relative to the circular orbit of the GPM satellite with an inclination of 65°, whereas the inclination angle of the TRMM satellite is 35°. While the VPRF technique has increased the number of independent samplings, the pulse-to-pulse spacing is decreased from 54 km for TRMM PR to 33.3–37.5 km for GPM DPR. However, this reduction in the sampling window may introduce another uncertainty for some data at high altitudes, as discussed later.
Overshooting convection has been investigated using spaceborne radar and radiometers. Liu and Zipser (2005) used TRMM PR data to obtain the climatological characteristics of highly developed convective systems over land in the tropics. Hourngir et al. (2021) investigated the spatial distribution of intense signals at high altitudes using four years of GPM DPR data. Various sensors have been used to observe extremely deep storms over tropical oceans (e.g., Casey et al. 2007; Kelley et al. 2010; Takahashi et al. 2017; Kim et al. 2017; Proud and Bachmeier 2021). TRMM PR was deficient at detecting weak echoes at high altitudes because of a lack of sensitivity and insufficient observation range (Hirose 2009). Weak precipitation echoes around an altitude of 20 km were difficult to capture fully because of increased uncertainty regarding detection and estimation at high altitudes, where signals become less significant relative to subtle noises. Despite such limitations, Hirose (2009) approximated the average occurrence probability of deep profiles with a top height of >18 km (hereafter called very deep storms) using high-altitude precipitation data accumulated by TRMM PR. The result was 10−7, which corresponds to 3 s yr−1. Data were averaged over the entire observation area and were approximately 5 min yr−1 in the vicinity of the northern Bay of Bengal, which are approximately two orders of magnitude more frequent. This preliminary result is highly susceptible to sensor sensitivity, spatial filtering of numerous random noises, sidelobe clutter contamination, and the upper observation limit. To solve these problems and understand precipitation signals at high altitudes, statistical analysis of long-term DPR data is required. TRMM PR version 8 data reprocessed to be consistent with the DPR V06A algorithm are also available. The detection ability of deep storms should be confirmed.
In this study, unwanted data of suspicious interference profiles in TRMM PR and GPM DPR KuPR products were removed, and the effects of their removal on the extraction of meaningful signals at high altitudes were investigated. Although TRMM PR guarantees observations up to 15 km from the surface and GPM DPR up to 19 km, this study focused on using data beyond these limits. Given the need for further high-altitude observations, it is critical to understand the data characteristics under which conditions the currently available high-altitude statistics can be used. Data analyses were performed to clarify the statistical effects of removing artificial interferences, particularly focusing on the detection of deep precipitation profiles. The rest of this paper is organized as follows. Section 2 describes our data and methodology. Section 3 presents the removal of radio interference in the TRMM PR version 7 product and the impact of the removal on statistics. Section 4 discusses the status of the interference for follow-on products and the remaining uncertainties on echoes at extremely high altitudes in the GPM DPR KuPR product. Section 5 concludes the study.
2. Observation coverage and precipitation data
In this study, the status of suspicious interferences in the Ku-band spaceborne radar products was investigated to evaluate the current detectability of extremely deep storms. The data used were from the TRMM PR versions 7 and 8 products from 1998 to 2013 and the GPM DPR KuPR V06A product from May 2014 to April 2019. The TRMM PR version 7 product was used to understand the existence and effects of radio interference noted in previous studies, and the version 8 product was used to confirm that noise processing was being performed properly. The GPM DPR data were used for comparison with the TRMM PR data, and they were used to confirm the statistics of precipitation echoes reaching higher altitudes. Taking these remaining issues into account, the challenges of the latest V07A products are also presented at the end of section 4.
TRMM PR operated with dual-frequency agility at 13.796 and 13.802 GHz. The horizontal and vertical resolutions were approximately 5 km and 250 m, respectively. Cross-track observations were made 49 times with a maximum zenith angle of 18°, or 0.75° per angular bin. Figure 1 depicts the distribution of the upper limits of the TRMM PR and GPM DPR KuPR observation levels versus the angles of incidence and latitude, using data from June 2014 as a representative example. The highest observation level of TRMM PR was between 15.7 and 19.75 km from the Earth ellipsoid (hereafter “geoid” for simplicity). In August 2001, the altitude of the TRMM satellite was increased from 350 to 402.5 km to extend its lifetime. The median of the upper limit ranged from 18.7 km at the swath edge to 19.6 km at the nadir before the 2001 orbital boost. The upper limit around the nadir decreased slightly after the boost. In addition to the slant-path effect, the observation level was affected by the orbital pattern, which was asymmetric between the Northern and Southern Hemispheres. At latitudes north of 15°S, deficits around 19 km altitude were noticeable, as depicted in the right panel of Fig. 1. The lowest upper limit of the observation level was around 10°–12.5°N. Table 1 summarizes the altitude data for the observations made in June 2014 at 10°–12.5°N and 32.5°–35°N, respectively. After the orbital boost, the upper limit decreased near the nadir by more than 1 km. The median and lowest observation levels at the nadir were 16.7 and 15.4 km, respectively, in the tropical latitudinal band. The observation levels decreased significantly in July and August 2004 because orbital maneuvers were stopped on the basis of operational decisions at that time. For these 2 months, the median and lowest observation levels over the Sahel area were 15.2 and 13.2 km, respectively. Otherwise, the normal operating performance was generally maintained. After July 2014, the degradation of high-altitude measurements became pronounced because TRMM could no longer maintain its altitude. The median observation level at the nadir was approximately 16 km in August 2014. As the upper end of the observation dropped 1 km, the number of snapshots with incomplete observation information at the top increased rapidly in the tropical land, as described later.
Upper limits of observation levels for TRMM PR (solid lines) and GPM DPR KuPR (dotted lines) in June 2014 for (left) each incidence-angle bin and (right) each latitude in increments of 2.5°. The orange, red, black, blue, and light-blue lines indicate the minimum (min), first quartile (Q1), median (Q2), third quartile (Q3), and maximum (max) observation levels, respectively.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
The minimum (min), first quartile (Q1), median (Q2), third quartile (Q3), and maximum (max) observation levels in June 2014. Units are in kilometers.
TRMM PR used an algorithm called 2A25 (Iguchi et al. 2000), which was revised three times after launch. The final 2A25 is extensively used for the complete TRMM PR dataset, which is referred to as the version 7 product. GPM DPR uses an algorithm called V06A that is applicable not only to its data but also to the past data of TRMM PR. The TRMM PR product based on DPR V06A is called version 8, and it is an update of version 7 through several processes, such as rigorous calibration factor adjustment (Seto et al. 2021; Iguchi 2020). In this study, the conventional version 7 product was used to identify and separate radio interference signals. The spatiotemporal characteristics and removal methods of interference signals are described in section 3a. The applicability of a filter was also examined for TRMM PR version 8 data over 16 years, i.e., 1998–2013, and GPM DPR KuPR V06A data over 5 years, i.e., May 2014–April 2019.
GPM DPR is the follow-on successor to TRMM PR, with updated hardware and advanced retrieval algorithms based on dual-frequency information. GPM DPR consists of two radars: KuPR operating at the frequencies of 13.597 and 13.603 GHz and KaPR operating at the frequencies of 35.547 and 35.553 GHz. The resolution of KuPR is almost the same as that of TRMM PR, except that the vertical sampling interval is 125 m and is based on the oversampling approach. The observation area is expanded into mid- to high latitudes and is a few kilometers higher (Fig. 1, Table 1) than that of TRMM PR. GPM DPR observation ensures data below 19 km and obtains echoes up to approximately 22 km. Almost all observations obtain data at 20 km at all incidence angles. On average, the observation ranges from 1.2 to 21.3 km from the geoid. GPM DPR uses the VPRF technique to expand the observation range and enable constant height range sampling. This increases the number of independent samples and improves sensitivity, which is suitable for detecting extremely deep storms. GPM DPR has been operating properly at an altitude above 396.5 km, at which VPRF functions properly, since 3 April 2014. Orbit maneuvers are performed every 7–10 days to maintain the altitude at 407 ± 10 km. The minimum detectable threshold of the measured radar reflectivity factor is 20.21 (19.34) dBZ for TRMM PR at 402.5 (350)-km altitude and 15.46 dBZ for GPM DPR KuPR (Masaki et al. 2022; Seto 2022).
Thus, GPM DPR KuPR data using the VPRF technique have improved the ability to extract more accurate and higher-altitude information by minimizing marginal portions of observed data not used in the algorithm. On the other hand, second-trip echoes (the contribution of echoes measured at times outside of a given sampling window to the next averaging process) can be even more problematic. The identification and filtering of nonnatural outliers are discussed in section 4b.
3. Contamination of TRMM PR data by radio interference
a. Radio interference of spaceborne precipitation radar
The power received by radar includes contributions from precipitation and artificial signals, which can be difficult to distinguish and require careful analysis. Interference from ground-transmitted radio appeared in all layers as apparently erroneous echoes. Hereafter, such signals above the noise level are referred to as “echoes.” Figure 2 illustrates two examples of “suspicious” precipitation profiles without and with low-altitude precipitation at a location near Long Beach, Los Angeles, where an artificial long-term mean signal has been previously identified (Hirose and Okada 2018). The image was prepared with a horizontal resolution of 0.05° and a vertical resolution of 250 m. Minimum detectable precipitation was estimated to be 0.40–0.47 mm h−1 for the TRMM PR data (Masaki et al. 2022), and most of the estimated precipitation rates in these profiles were weak. In some of the cases of radio interference, precipitation was observed at low altitudes, as depicted in the right panel of Fig. 2. The occurrence frequency of precipitation at a 0.01° scale over the grid (33.825°N, 118.215°W) and that at the surrounding area over a range of 5–10 km were 10.1% and 1.3%, respectively, from 1998 to 2013. Note that the 0.01° resolution precipitation values were generated using a method proposed by Hirose and Okada (2018), assigning individual observations to finer grids contained in the footprint. Although data in adjacent few kilometer grids are dependent, statistics differ when they are 5 km apart. Although radio interference did not always occur at this location, it had a significant effect on the statistics of precipitation frequency. Generally, the contaminated precipitation profiles could be distinguished because of the rarity of extremely deep storms and the extraordinary spatially biased echo pattern. The interfered profiles had common features in that the attenuation-corrected radar reflectivity (Ze) was almost constant and the precipitation rate calculated using the standard algorithm decreased downward, as depicted in Fig. 3. The lapse rate of the precipitation intensity was approximately 4% km−1 because the 2A25 algorithm assumed an atmospheric pressure correction for the terminal velocity. The solid lines indicate precipitation at low altitudes below 4 km and unnatural echoes above it.
Examples of extraordinary precipitation profiles observed by TRMM PR in Los Angeles: (a),(c) horizontal distributions and (b),(d) vertical cross sections of precipitation rate. The cases are on (a),(b) 20 Mar 1999 and (c),(d) 25 Feb 2001.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
Vertical profiles of (left) precipitation rate and (right) attenuation-corrected radar reflectivity observed by TRMM PR at 33.825°N, 118.175°W. The dashed and solid lines indicate the same anomalous profiles observed on 20 Mar 1999 and 25 Feb 2001, respectively, as in Fig. 2.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
b. Removal filter for radio interferences
As depicted in Fig. 3, the basic idea of the removal filter of interferences is to find profiles in which precipitation rate decreases from the top of the observation range. However, such profiles may exist naturally, particularly in the tropics. Thus, vertical and horizontal methods were introduced to define nonrain profiles affected by radio interference. The vertical method specifies the range of the downward-decreasing precipitation-rate profile from the top of the observation range down to certain observation levels (hereafter, DD Prof). Each received radar echo, obtained by averaging 64 independent samples for TRMM PR and typically 100 pulses for GPM DPR KuPR, fluctuates near the truth. The bottom of a DD Prof is extracted from the vertical running-averaged data to reduce sampling uncertainty, which appears as a zigzag pattern, as depicted in Fig. 3. First, precipitation rates of echo-filled profiles are averaged for each of the eight bin ranges (i.e., intervals of 2 km in depth or 250 m multiplied by 8). If the minimum rate is at the highest observation level, this is regarded as a deep storm and a natural phenomenon. In such cases, the mean precipitation rate in the interval from the highest observable altitude down to 2 km (an interval of approximately 18–20 km) is the lowest compared to the mean at each altitude interval. Once a downward-decreasing pattern is detected, the averaged range is changed in the order of eight, four, two, and one bin to determine the observation level of the minimum precipitation rate. DD Profs above the observation level of the minimum signal are preserved as possible interfering signals. In this study, the upper part of a profile is subjected to the filter to avoid removing too many of the precipitation echoes spreading at low altitudes.
To exclude deep storms from potential candidates for suspicious profiles, the horizontal method is used to detect neighboring precipitation signals. The filter for interference is turned off when there is at least one precipitation data with the storm-top height exceeding 15 km in the eight samples regarding the scanning angle and the number of scans of the instantaneous observation. Visual observations have indicated that nonisolated DD Profs are probably a part of highly developed storms. Thus, DD Profs without any well-developed storm in the adjacent area (i.e., isolated DD Profs) are considered artificial signals caused by radio interference. In this study, deep profiles reaching the top edge and filled with echoes across all layers with the minimum precipitation near the top and nonisolated DD Profs were labeled as “overshooting Profs” (Fig. 4).
Flow chart for determining isolated DD Profs and overshooting Profs.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
Figure 5a presents the distribution of overshooting Profs and isolated DD Profs. Most of the overshooting Profs were observed in the northern tropics, especially in central Africa and the Sahel area, where deep convection frequently occurs and the satellite altitude is relatively low. The monthly changes in the number of overshooting Profs and isolated DD Profs (i.e., candidates for interference) are depicted in Fig. 6. The maps created for different periods show the effects of changes in radio frequency usage by region and changes in satellite altitudes. The overshooting Profs and isolated DD Profs were very few during the early operation, as illustrated in Fig. 5b. After the orbit boost in 2001, the number of overshooting Prof detections has increased (Fig. 5c). The monthly average of overshooting Profs was approximately 12 before the orbit boost, while it has become approximately 120 after the orbit boost. Thus, the number of overshooting Profs detected increased even though sensitivity was degraded with the changes in the satellite altitude. An increase in the number of overshooting Profs can be attributed to the deterioration of the upper limit of the observation levels caused by the orbital change. In particular, the number of detections was very high in July and August 2014 when the satellite altitude decreased (Figs. 5d and 6). In July and August 2004, 1249 and 2246 overshooting Profs were observed, respectively; these values were more than 10 times higher than the average. As noted in section 2, this period coincides with the period when orbital maneuvers were stopped. Most of the precipitation signals at very high altitudes during these two months were likely to be related to storms truncated by the relatively low upper limit of the observation levels, such as in the Sahel area (Fig. 5d). These results show that the statistical data obtained from the TRMM on overshooting Profs were sensitive to the orbit characteristics. The deterioration that occurred when the designed operation was discontinued can be observed from the observations at high altitudes in tropical land. This also indicates the need for higher-altitude observations to determine the structure of precipitation systems with undetectable upper limits accurately, which are observed more than 100 times per month.
Distribution of TRMM PR profiles filled with echoes. The black and red points indicate overshooting Profs and isolated DD Profs, respectively. (a) The entire period; (b)–(e) different periods from 120°W to 150°E: (b) 1998–2000, (c) 2002–04, excluding July and August 2004, (d) July and August 2004, and (e) 2011–13.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
Number of overshooting Profs (black lines) and isolated DD Profs (red lines) per month observed by TRMM PR.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
Unlike overshooting Prof statistics, the spatial distribution and temporal variation of isolated DD Profs exhibited little correlation with the satellite altitude. Compared with the widespread distribution of overshooting Profs, isolated DD Profs occurred frequently at specific points. Half of them appeared over the Middle East, and some were observed in Brazil and India. The number of isolated DD Profs gradually increased over time, and the average number in 2011–13 was 11 times more than that in 1998–2000. Figures 5b, 5c, and 5e show that isolated DD Profs are detected in very limited areas in the early stage of satellite operation and that the signals are scattered over a wide area in the late stage of satellite operation, due to factors such as the expansion of radio frequency allocation (ITU-R 2003, 2021). Thus, there was a concern that echoes with unnatural spatial structures might appear depending on the region and period.
Further investigation into the occurrence of isolated DD Profs revealed that it was much less common in the inner angle bins (15–35 out of 49), where apparent system noise is sampled in the four range bins at the bottom of the data sampling window (Takahashi and Iguchi 2008). The difference in the incidence angle of noise processing is most likely visible during the initial stage of algorithm processing for the TRMM PR version 7 product. On the other hand, overshooting Profs were observed mainly near the nadir. This is related to the incidence angle dependence of the upper limit of the observation level in the tropics, as depicted in Fig. 1.
c. Effects of removing radio interference on average precipitation
The removal of radio interference may modify precipitation statistics over dry areas or at high altitudes where isolated DD Profs prevail. Based on TRMM PR data from 1998 to 2013, Fig. 7 depicts the annual precipitation at the surface, 15 and 19 km calculated from the geoid, as well as the effect of the removal filter. It is difficult to discern the effect of isolated DD Profs on the precipitation map at low observation levels, except around the Arabian Peninsula. After the correction, a slight negative impact on surface precipitation was observed in 48% of the 2° grid precipitation averages in the arid Middle East (13°–37°N, 30°–75°E) (Figs. 7a,d). At 15 km, where deep storms were observed over tropical landmasses such as Africa (Fig. 7b), more than half of the precipitation was removed in 13% of the regions for the 2° grid statistics in the Middle East (Fig. 7e). The most significant impact was observed on the precipitation map at 19 km from the geoid (Fig. 7f). More than 90% of precipitation at 2° scales close to the highest altitude was deleted by the filter over 5% of the region, and 50% was deleted over 19% of the region. High-altitude precipitation over parts of Brazil and India was reduced by several tens of percent (Fig. 7f).
(a)–(c) Annual precipitation and (d)–(f) the effect of removal filter on precipitation from TRMM PR data from 1998 to 2013 at a grid resolution of 2°: (a),(d) at the surface, (b),(e) 15 km from geoid, and (c),(f) 19 km from geoid.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
In Fig. 7c, the median (interquartile range) of annual precipitation at 19 km was 0.15 mm (0.13–0.16 mm) over the ocean and 0.24 mm (0.20–0.27 mm) over land. The values were slightly smaller around 10°–12.5°N with fairly small sample sizes. The almost homogeneous distribution of precipitation at high altitudes than at low altitudes over land and ocean might be due to misjudgment of precipitation. The signal is determined as the received power minus the noise defined for each ray, but the sampled noise fluctuates due to the wide fading range of the system noise (Takahashi and Iguchi 2008). The dynamic range of the noise depends not only on the fading noise, but also on the receiver temperature and radiation from the ground. This interference appears sporadically in all layers and throughout the entire area. The annual precipitation of approximately 0.2 mm is probably susceptible to noise. Near the ground, the effect of noise is usually negligible as it is overwhelmed by the precipitation signal, but caution should be exercised when statistics are taken at high altitudes, where the precipitation is extremely low.
The performance of the filter was remarkable for deep storms that reached the upper limit of the observation levels. Figure 8 shows the average precipitation-rate profiles for each storm-top height. The storm-top height was calculated from the geoid considering the slant-path effect. For example, the range of 19 km from the geoid at an incidence angle of 10° was converted into a vertical height of 18.7 km. The precipitation rate at low observation levels interfered with by ground clutter was interpolated using the near-surface precipitation and estimated surface precipitation. This is consistent with the assumption that the effective radar reflectivity factor at the surface is identical to that at the lowest point free from mainlobe clutter. The average profile of precipitation is often divided according to whether it is linked to surface precipitation or filled with echoes. The conditional average of precipitation at a given altitude (i.e., average 1) is depicted in the left panels of Fig. 8. The conditional average per-level precipitation rates >0 at all altitudes below the storm-top height (i.e., average 2) is depicted in the right panels of Fig. 8. The difference between average 1 and average 2 can be attributed to suspended or intermittent structures caused by evaporation at the dissipating stage, as a part of slant convective cores at the mature stage, or convective echoes at the initial stage. Furthermore, numerous random noises could be a major factor causing the difference. Precipitation samples are abundant at low altitudes, and random noise interference has only a minor effect. However, in areas with very low precipitation, such as off the coast of Peru, altitude-independent random noise interference becomes obvious, as described in the previous paragraph, and the storm-top height tends to be higher. Thus, profiles with higher storm-top heights are more likely to include more random noise effects. The difference between average 1 and average 2 became significant as the storm-top height increased and the number of precipitation samples decreased (Figs. 8a,d). The average precipitation-rate intensity of average 2 was approximately twice that of average 1. The middle panels (Figs. 8b,e) show the results after filtering. Figure 8e shows that the average 2–based precipitation intensity at the surface and above was almost proportional to the storm-top height at observation levels <18 km. The vertical precipitation-rate characteristics are consistent with the results of Takayabu (2002) for a storm-top height <16 km. The higher profiles indicate a bimodal structure. The surface precipitation rates of storms with top heights between 18 and 19 km are lower than those of storms with top heights between 17 and 18 km; this implies a weak link between extreme rainfall and the tallest convection, as reported by Hamada et al. (2015). The bottom panels show the effect of the removal filter of isolated DD Profs. The most significant difference in average 2 with the filter was observed for storm-top heights > 18 km (Figs. 8c,f). Hence, the filter is critical for assessing very deep storms. More precisely, the filtering effect was the highest at storm-top altitudes of 18–19 km because the detected radio interference was concentrated in the outer angle bins, where the highest upper limit of the observation level was approximately 500 m lower than in the inner angle bins. The filter was confirmed to have moderate and significant impacts on the statistics for deep profiles at heights > 16 and 18 km, respectively. The precipitation intensity of very deep profiles with a storm-top height > 18 km increased by a factor of 2 or 3 at each observation level when interference was removed. Not only extremely high profiles at >19 km but also globally averaged profiles with a storm-top height of approximately 17 km were several percent weaker in intensity. The storm-top height of interfered profiles always corresponded to each upper limit. As noted previously, the top height varied with the incidence angle, region, and observation period. The filter decreased the rain intensity for average 1 and increased it for average 2. The former indicates that the interference had a significant effect relative to random noises. The latter shows that the interfering signals were weak compared with the precipitation signals of deep storms.
Averaged TRMM PR precipitation-rate profiles sorted by storm-top heights (a),(d) before filtering, (b),(e) after filtering, and (c),(f) the difference. (a)–(c) Average 1 is for all cases, and (d)–(f) average 2 is for profiles filled with echoes between the storm-top height and near-surface level. The storm-top height statistics were obtained in increments of 250 m.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
The precipitation-rate profile in Fig. 8e is continuous with respect to storm-top height below 19 km, but it varies at higher storm-top height, suggesting the presence of residual echoes that are not present in nature. As shown in Fig. 1, the number of stored data decreases with altitude, and the remaining artificial signal is probably due to sampling issues and less stringent filtering requirements. Radio interference with echoes at high altitudes in the vicinity could occur. We removed a spatially discontinuous profile of isolated all-layer echoes, but more careful removal may modify the statistics. Furthermore, there are differences in sampling due to the angle of incidence, which varies above 19 km altitude, as depicted in Fig. 1. Because a very intense precipitation rate exhibits pronounced incidence-angle dependency (Hirose et al. 2012), the statistics for precipitation reaching altitudes higher than 19 km may reflect biases in estimates of intense precipitation rates at different incidence angles, which is not considered in depth in this study.
4. Uncertain signals at high altitudes in GPM DPR data
a. Updated status on radio interference
Here, the results for the extraordinary profiles with updated algorithms are described: TRMM PR version 8 and GPM DPR KuPR V06A. First, the differences in processing possible contamination exceeding the noise-level limit between TRMM PR versions 7 and 8 should be discussed. As described in section 1, the warning flags of apparently erroneous system noise were prepared for the TRMM PR 1B21 product, but the removal procedure was not fully performed for the version 7 product. These flags help identify radio interference for the noise level obtained within a certain sampling window at a relevant location; however, the flag-based removal process does not function properly for this product (T. Kubota 2022, personal communication). The number of warning flags was almost identical to that of isolated DD Profs. From 2001 to 2013, the average and standard deviation of the ratio of the annual number of flags to the number of the isolated DD Profs were 1.06 and 0.05, respectively. Whole data in a few orbits were flagged in 1998 and 1999, which had a significant effect in those years. Excluding these issues, the flags worked well at detecting suspicious profiles and steadily improved every year.
The processing glitch was corrected in the version 8 product. The updated algorithm treats the 49 data points in the scan direction containing the target point as missing, which eliminated most of the isolated DD Profs. For example, the two profiles of radio interference shown in Figs. 2 and 3 were removed, along with all their cross-track directional data. In the Middle East (13°–37°N, 30°–75°E), the number of profiles filled with full-layer echoes in the PR version 7 data for the period 1998–2013 was 22 548, of which 94% were isolated DD Profs. The number of profiles filled with full-layer echoes under the same conditions using the version 8 data was reduced to 74, all of which were accompanied by spatially coherent precipitation. Not a single isolated DD Prof remained in the area.
GPM DPR KuPR V06A is based on the same algorithm even though GPM DPR KuPR differs from TRMM PR in terms of frequency, sensitivity, and upper observation range. The radio interference seen in TRMM PR version 7 was seldom observed in the GPM DPR KuPR V06A product over a 5-yr period, i.e., May 2014–April 2019. In the Middle East, profiles, where the entire layer was filled with echoes, were found in only two cases. One was observed over the Arabian Sea (two profiles, orbit number 7254), and the other was observed over the Red Sea (orbit number 13772). Both appear to be part of deep convective systems, and no isolated DD Profs were found. For regions outside the Middle East, 20 profiles filled with echoes were detected. They were mostly observed near coastlines in the tropics, and 14 profiles were confirmed to be overshooting Profs based on their spatial extents. Isolated DD Profs detected in the five years of KuPR V06A data included only six profiles at three locations: coastal waters north of the Netherlands (two profiles, orbital number 1862), coastal land areas of eastern China (orbital number 2523 and 4706), and the sea north of Guam (two profiles, orbit number 10531). Thus, the analysis of spatial patterns in this study shows that the latest algorithm eliminated most of the interfering signals even though it did not eliminate all.
b. Unnatural echoes near the upper limit of observation
As discussed in the previous subsection, the effect of radio interference was not a major issue for the KuPR product. However, regarding the detection of extremely deep storms and noise reduction at the highest altitudes, caution is required in the upper parts of KuPR precipitation profiles above 19 km, where accuracy is not guaranteed (Iguchi et al. 2021).
First, sidelobe clutter contamination echoes appeared at specific observation levels and incidence angles, and they exhibited a nonnegligible effect as an error factor for some high-altitude precipitation data. The clutter interference was caused by the backscatter from the surface in antenna sidelobes, which can be predicted to some extent based on the geometry of the antenna pattern. A larger scan angle is more likely to lead to interference at higher altitudes. The received power of the sidelobe clutter was significantly attenuated, but some remained (Kubota et al. 2016). Almost all echoes detected at observation levels > 10 km in the outer half of the scan and those exceeding 18 km in two outermost angle bins over the ocean in 60°–67°N could be attributed to the sidelobe contamination.
Second, oversampled data were not obtained at high altitudes for the KuPR product. DPR quality assurance was initially assumed to be 18 km, but after discussions in 2006, it was decided to guarantee quality up to 19 km. At that time, the number of packets could not be increased because of data rate limitations: thus, normal sampling was used instead of oversampling only at high altitudes. Echo power measurements were performed every 250 m as normal sampling at altitudes above approximately 15 km and every 125 m as oversampling at lower altitudes. At high altitudes, data with a 125 m interval were obtained by the interpolation of data with a coarser range resolution due to a lack of oversampled data (Iguchi et al. 2021; Kanemaru et al. 2020). The vertical interpolation may lead to noise smearing and false signal detection (K. Kanemaru 2022, personal communication). Factors such as noise misjudgment, pressure corrections related to the terminal velocity, and interference effects (described in the following paragraph) caused a slight increase in precipitation at observation levels above 18 km at all incidence angles. Furthermore, the precipitation routinely reached altitudes of 15–18 km at low latitudes, resulting in the observation of the precipitation minima at an altitude of approximately 18 km.
The third problem is related to a “mirror” image of precipitation below the ocean surface (Battaglia 2021; Li and Nakamura 2002). Most Ze profiles with echoes at the upper end of the observation range have an unnatural feature that decreases downward at observation levels between approximately 18 and 22 km (Fig. 9). Signals above the noise in the highest range bin were observed in 1768 profiles from 2015 to 2017. The upper part of the unique profiles can be attributed to second-trip mirror echoes from significant solid precipitation at observation levels around 15 km based on the sampling window of approximately 35 km (T. Iguchi 2018, personal communication). For TRMM PR, the sampling window was 54 km because the pulse repetition frequency was 2776 Hz. The narrow sampling window of KuPR was determined by variable pulse repetition intervals according to the satellite altitude and incidence angle, resulting in easier second-trip echoes. Battaglia (2021) investigated the mirror image and other second-trip echoes resulting from multiple scattering using the cloud profiling radar on board the CloudSat satellite. An interfered profile typically appeared as an isolated profile characterized by a downward-decreasing precipitation rate with a maximum at the highest observation level. At the end of the previous subsection, we reported that there were 17 overshooting Profs in the KuPR observations from May 2014 to April 2019. The orbit numbers containing the profiles are listed as 4491, 6952, 6984, 7254, 11398, 11764, 13114, 13772, 14016, 14953, 21936, 23790, 25319, 25588, 28671, and 29044. These deep profiles exhibited spatially unnatural features around 20 km above the geoid, including mainly isolated profiles with downward-decreasing precipitation rates above very deep nonisolated storms. Most other mirror image cases appeared as echoes floating above the top of deep profiles, as shown in Fig. 9. In most of these cases, Ka-band radar exhibited no echoes at altitudes corresponding to the mirror image. This is probably due to differences in attenuation caused by rain and atmospheric liquid water along the path and the surface radar cross section (Battaglia et al. 2020; Li and Nakamura 2005).
Upper part of Ze profiles with echoes at the top of the observation range of KuPR from 2015 to 2017.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
c. Elimination of suspicious echoes near the upper limits and its effects
Radio interference was negligible for the current GPM DPR KuPR data, but the upper parts of profiles with echoes at the highest observation level were suspicious. Therefore, we established a simple filter to reduce uncertainties caused by sampling degradation and the mirror image effect. Here, we obtained high-altitude precipitation echoes by excluding profiles that reach the upper edge and decrease downward, as shown in Fig. 9, and have no precipitation in the surrounding pixels at high altitudes, i.e., extraordinary deep profiles and isolated high-altitude echoes with no horizontal extent (Fig. 10). In this section, the results for echoes at high altitudes above 15 km are presented. Outer swath data were excluded because of the sidelobe contamination. If all angle data were used (i.e., gray lines in Fig. 11), significant peaks derived from the sidelobe clutter at the seven outer bins appeared at observation levels around 17–20 km. The black lines in Fig. 11 indicate the statistics based on angle bins between 8 and 42 out of 49. The difference from the gray lines indicates sidelobe clutter interferences. When restricted to data in the inner swath, the number of echoes at 19–20 km was reduced by 96%.
Flow chart for determining precipitation echoes and artificial signals at high altitudes.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
(right to left) Number of total KuPR precipitation samples and those above the thresholds of 17, 20, and 30 dBZ per million samples at each altitude above 15 km, respectively. The gray and black lines indicate occurrence frequency with all angle bins and excluding the seven outer bins, respectively. The blue and red lines represent the same information as the black lines but for profiles without echoes at the highest range bin and those with precipitation in the vicinity, respectively.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
The blue and red lines in Fig. 11 indicate the occurrence frequency of precipitation echoes above thresholds at high altitudes for inner-swath profiles, excluding cases reaching the upper edges and those with precipitation in the surrounding pixels at high altitudes, respectively. As indicated by the blue line, simply excluding profiles that reach the upper edge reduced precipitation at >19 km by more than an order of magnitude. In the left panel, the blue line nearly overlaps the red line, indicating that most of the prominent echoes above 30 dBZ at high altitudes are accompanied by nearby precipitation. Compared with our previous study using noisy TRMM PR data (Hirose 2009), the effect of random noise removal, as shown in the difference between the blue and red lines, appears small in the present data.
The effects of the filter on profiles reaching the upper edges and isolated echoes can be visualized by the difference between the black and red lines. Above 18 km, statistical differences can be discerned. The downward-decreasing features above 19 km, as shown in the black lines, have been removed. The precipitation statistics indicated by closely spaced blue and red lines in the second panel from the right in Fig. 11 indicate that most of the profiles with echoes above 17 dBZ and at altitudes of 19–20 km belong to a widespread precipitation system. In contrast, the same statistics for all precipitations (including cases below 17 dBZ) in the rightmost panel show a pronounced effect of the horizontal filter, implying the presence of many isolated profiles consisting of weak echoes. The difference between the black and red lines in the right-hand side of Fig. 11 shows the difference before and after the application of the filter, indicating that the filter removes approximately 10% of the echoes at an altitude of 17–18 km, 50% at 18–19 km, and 80%–90% above 19 km. A slight increment at 19–22 km implies that some weak random noise with spatial extension remained. The number of precipitation samples for a million samples at >19 km after applying the filter was 0.682, 0.075, 0.015, and 0.002 for all samples, with echoes exceeding 17, 20, and 30 dBZ, respectively. Given that GPM DPR collects approximately 7 million footprints per day, precipitation above an altitude of 19 km was observed once every 0.2 days, including weak echoes, and echoes above 20 dBZ were observed once every 10 days within the observation area. When converted to time per year, the occurrence frequencies corresponded to 21.5, 2.4, 0.5, and 0.1 s, respectively. No 30-dBZ echoes were observed above 20 km. The number of meaningful signals decreased by a factor of 7 when the altitude was increased by 1 km for observation levels of 16–19 km. The rate of change implies the presence of a small amount of precipitation even above 20 km, where convection has a negligible effect on the stratosphere air composition (Bucci et al. 2020). The number of echoes at 20 dBZ decreased almost exponentially to an altitude of 21 km. Weaker echoes did not decrease to the same extent as at 16–19 km altitude, but whether this is due to noise or weak upwelling requires further investigation.
This preliminary investigation into contaminated echoes indicates that a removal mask considering spatial patterns and sidelobe contamination is required to analyze signals at extremely high altitudes of >19 km. GPM DPR KuPR has collected precipitation samples of extremely deep storms above 19 km every year that TRMM PR could not detect because of its observation limits. The number of precipitation echoes in the 8–42 angle bins at observation levels of >19 km yr−1 and those at >17, 20, and 30 dBZ was 13 964, 3385, 887, and 12 and 2421, 240, 47, and 7, before and after the correction, respectively.
Figure 12 shows a horizontal distribution of corrected high-altitude echoes obtained at angle bins 8–42 of 49 bins and the number of removed echoes. In previous studies on storms penetrating the tropopause, such as Liu et al. (2020), there were significant peaks in the United States and Argentina, but in terms of absolute altitude, they were characteristic in the tropics. Deep profiles with a top height of up to 15–17 km were frequently observed in lower latitudes. The maximum number of precipitation samples (>30 dBZ) per 2° grid from May 2014 to April 2019, 3642 (582), was observed in northern Colombia (7°–9°N, 76°–74°W). Tens of echoes above 20 and 30 dBZ were widely observed over tropical oceans and land, respectively. Precipitation exceeding 20 dBZ at 17–19 km altitude has been observed over tropical land areas such as the interior of South America, Central Africa, the periphery of Bangladesh, and northern Australia. In addition, over the western Pacific, where tropical disturbances cause overshooting convection (Kelley et al. 2010), scattered precipitation systems with echoes of 20 dBZ reaching very high altitudes can be observed. Tens of precipitation echoes < 20 dBZ at 19–22 km altitude were observed over tropical coastal areas such as the northern Bay of Bengal and Central America. Weak random noises detected everywhere at high altitudes were not completely removed, as shown in the bottom panels of Fig. 12, but mirror images in the tropical ocean detected off the coast of Mexico and Colombia and elsewhere were significantly removed at high altitudes of >19 km. For example, in the two regions enlarged in Fig. 12, i.e., near the northern Bay of Bengal and Central America, the filter reduced the number of artifacts (only mirror images) above 19 km altitude by 75% and 83% (70% and 80%), respectively. Thus, a significant portion of the removal effect in these areas was attributed to spatially isolated profiles with no adjacent precipitation sample. As shown in the upper right panel, the filter removed some precipitation echoes > 30 dBZ around 15–17 km altitude. As discussed later, it would be necessary to set an appropriate lower altitude at which to apply the filter in this regard.
Distribution of precipitation samples at high altitudes (gray) and those above 20 dBZ (cool colors) and 30 dBZ (warm colors) based on 5 years of inner-swath KuPR data: (top) 15–17, (middle) 17–19, and (bottom) 19–22 km. (left) Samples filtered for profiles with echoes at the highest observation level and isolated precipitation profiles, and zoomed in on the northern Bay of Bengal and near Central America. (right) Number of removed echoes. Stronger echo information is shown. The spatial resolution is 2°.
Citation: Journal of Atmospheric and Oceanic Technology 40, 8; 10.1175/JTECH-D-22-0114.1
GPM DPR V07A products are now available and TRMM PR V07A products will be released soon. An initial investigation with GPM DPR KuPR V07A data in 2018 and 2019 detected only two cases of ground-based signal contamination observed in the TRMM PR. They were observed in northeastern China (orbit number 30474) and in Abidjan, Cote d’Ivoire (orbit number 32077). Nine other systems were also detected with profiles with full-layer echoes with mirror images at the upper end of the observation, continuing from the V06A data. Mirror images are characterized by downward-decreasing precipitation intensity from the top edge, with precipitation maxima at high altitudes, as described in the previous section. While three-quarters of cases filled with full-layer echoes were observed at the incidence angles on the inner half, the deep precipitation (orbit number 31619) measured in the western Atlantic on 22 September 2019, was accompanied by a large mirror image that also extended horizontally and appeared near the swath edge. No sidelobe-specific features were observed in this case. Thus, the latest DPR KuPR products also include the effect of mirror images, and a flag is provided to indicate the lowest bins where mirror images may appear (Iguchi et al. 2021). The removal of the remaining artifacts should be done as required by the user. As depicted in Fig. 12, precipitation information should be extracted at high altitudes by removing the upper end of the profiles down to the altitude indicated by the flag. This procedure removes mirror images that extend over the upper 4 km of the above-mentioned profiles with orbit number 31619. Although most of the mirror images were found in tropical or subtropical coastal oceans, they were also detected on land. For example, a pronounced precipitation system with a mirror image filled with full-layer echoes was observed near the Zambezi River in Mozambique on 19 March 2019 (orbit number 28671). This phenomenon has also been detected near Kolkata in eastern India, and the occurrence of mirror images should be investigated further.
5. Conclusions
The detection of precipitation at very high altitudes and the collection of data are critical for a thorough understanding of extreme events and accurate retrieval. This study aimed to separate naturally occurring precipitation signals from artificial echoes at altitudes around 20 km. The first half of this paper focused on the occurrence of radio interference in spaceborne precipitation radar and estimated their effect on the precipitation statistics using 16 years of TRMM PR version 7 data. The number of interfering signals increased markedly throughout the observation period. Their effects were subtle in most of the statistics, but a significant impact was observed in the analysis of specific targets regarding microclimatology or extreme events. For example, the 2° grid mean precipitation at an altitude of 19 km in the Middle East was due to radio interference in more than half of the signals in one-fifth of its area. A procedure for eliminating radio interference is critically required to detect localized phenomena. The removal of radio interference and random noises changed the statistics of precipitation-rate profiles with storm-top heights of >16 km. The results of the filter for arid regions and high-altitude precipitation statistics highlight the importance of rigorous quality checks for understanding high-resolution and comprehensive storm characteristics from around the world.
The upper limit of the measurement height of TRMM PR varies by several kilometers depending on the incidence angle and orbital altitude. The upper limit of the observation height, even though less frequent in occurrence, seems insufficient for detecting the overall structure of extremely deep storms, particularly over Africa. The significant deterioration of the observation range was also addressed by the increased number of truncated deep storms, herein, regarded as overshooting Profs, when TRMM temporarily ceased orbital maneuvers in 2004. Thus, the localized impact of artificial signals was observed in the TRMM PR version 7 product. The results from follow-on products exhibited no noticeable problems from radio interference with adequate noise filtering processing. However, note that some regions and years are prone to radio interference, and the effect of the removal of this interference may not be negligible, especially for samples with low precipitation. In locations with frequent interference occurrences, removing the entire scan as in the filter of the standard algorithm increases the loss in the surrounding area; thus, our method, which removes only the upper part of isolated DD Profs, would increase near-surface precipitation data.
The VPRF technique of GPM DPR changes the pulse interval so that precipitation at >19 km altitude can be investigated. GPM DPR was used because of the unavailability of high-altitude samples for TRMM PR in tropical regions such as Africa, but a closer look revealed problems when using data at high altitudes, where accuracy cannot be adequately ensured. When creating a composite diagram of precipitation at high altitudes, it is necessary to better capture the continuity and consistency of features based on the storm structure and other sensor information. The detected echoes at these observation levels must be handled carefully to consider sidelobe contamination, sensitivity degradation, and mirror images. If these artificial profiles are removed, rare cases of extremely deep storms over specific areas can be extracted. Precipitation samples decrease exponentially with increasing altitude, and samples at altitudes >19 km are obtained approximately once every few days. For abundant samples detected with high sensitivity, avoiding mirror image contamination on the water surface is crucial for analyzing the statistics of precipitating echoes at altitudes of 20 km. The collection of GPM DPR data over time will provide opportunities to capture rare phenomena.
The latest algorithm has the same mirror image challenge, which can be appropriately handled by the method presented later in this paper. The V07A algorithm includes several improvements. For example, it reduces sidelobe clutter contamination and improves the identification of precipitation signals by considering 3D features, changes in the precipitation detection threshold, and changes in the measured reflectivity near the noise level (Iguchi et al. 2021; Iguchi 2017; Kanemaru et al. 2020, 2021). It has a significant impact on the detection of weak and small precipitation events, which are common at low altitudes. The improved method for determining precipitation echoes has not yet been fully evaluated based on the latest data accumulated over a long period. The noise smearing effect would remain because of the lack of oversampled data at high altitudes, but the improved precipitation determination method is expected to contribute to precipitation surveys at high altitudes by reducing the number of artificial echoes at outer incident angles and increasing the number of spatially continuous precipitation signals. Efforts to detect weak echoes aloft to understand cloud and precipitation systems by comparing data with various precipitation intensity ranges collected using multiple sensors are worth continuing.
This study has deepened our understanding of the retrieval uncertainty of data from extremely deep storm at altitudes beyond the initial specifications. Increased extraordinary TRMM PR signals are most likely caused by interference from the primary services assigned to the TRMM PR band, such as satellite uplink stations, and noncompliance with ITU regulations. The observation environment and targets have changed over the years. Continued efforts are required to eliminate the minute unnatural echoes that can be attributed to observations, algorithms, and the environment to improve the accuracy of collected climate information.
Acknowledgments.
The authors express their gratitude to the members of the TRMM and GPM projects. The authors are grateful for the constructive and informative comments on mirror images by Dr. Iguchi, Dr. Masaki, and Dr. Kubota. The authors also thank Dr. Kanemaru for providing valuable information about noise contamination. This work was primarily supported by the second and third research announcements on the Earth Observations by the Japan Aerospace and Exploration Agency (JAXA). The constructive comments of the three anonymous reviewers are greatly appreciated.
Data availability statement.
TRMM PR and GPM DPR standard products used in this study are available from JAXA G-Portal website (https://gportal.jaxa.jp/gpr/).
REFERENCES
Alandjani, G., 2012: Interference and compatibility analysis of ARABSAT vis-à-vis Eutelsat systems and networks. Int. J. Sci. Eng. Res., 3, 91–94, https://www.ijser.org/viewPaperDetail.aspx?M0R0133.
Battaglia, A., 2021: Impact of second trip echoes for space-borne high pulse-repetition-frequency nadir-looking W-band cloud radars. Atmos. Meas. Tech., 14, 7809–7820, https://doi.org/10.5194/amt-14-7809-2021.
Battaglia, A., K. Mroz, D. Watters, and F. Ardhuin, 2020: GPM-derived climatology of attenuation due to clouds and precipitation at Ka-band. IEEE Trans. Geosci. Remote Sens., 58, 1812–1820, https://doi.org/10.1109/TGRS.2019.2949052.
Bucci, S., and Coauthors, 2020: Deep-convective influence on the upper troposphere-lower stratosphere composition in the Asian monsoon anticyclone region: 2017 StratoClim campaign results. Atmos. Chem. Phys., 20, 12 193–12 210, https://doi.org/10.5194/acp-20-12193-2020.
Casey, S. P. F., A. E. Dessler, and C. Schumacher, 2007: Frequency of tropical precipitating clouds as observed by the Tropical Rainfall Measuring Mission Precipitation Radar and ICESat/Geoscience Laser Altimeter System. J. Geophys. Res., 112, D14215, https://doi.org/10.1029/2007JD008468.
Eutelsat, 2012: Eutelsat condemns jamming of broadcasts from Iran and renews appeals for decisive action to international regulators. Eutelsat Doc., 2 pp., https://www.eutelsat.com/files/live/sites/eutelsat-internet/files/contributed/news/press/en/2012/PR%206212%20interference%20iran.pdf.
Eutelsat, 2022: Earth station technical and operational minimum requirements—Standard M. Eutelsat Doc. EESS 502, 42 pp., https://www.eutelsat.com/files/PDF/support/EESS-502.pdf.
Hamada, A., and Y. N. Takayabu, 2014: A removal filter for suspicious extreme rainfall profiles in TRMM PR 2A25 version 7 data. J. Appl. Meteor. Climatol., 53, 1252–1271, https://doi.org/10.1175/JAMC-D-13-099.1.
Hamada, A., Y. N. Takayabu, C. Liu, and E. J. Zipser, 2015: Weak linkage between the heaviest rainfall and tallest storms. Nat. Commun., 6, 6213, https://doi.org/10.1038/ncomms7213.
Hirose, M., 2009: Detection of rare precipitation events from the long-term spaceborne radar data (in Japanese). Meijo University Faculty of Science and Technology Research Rep., 5 pp.
Hirose, M., and K. Okada, 2018: A 0.01° resolving TRMM PR precipitation climatology. J. Appl. Meteor. Climatol., 57, 1645–1661, https://doi.org/10.1175/JAMC-D-17-0280.1.
Hirose, M., S. Shimizu, R. Oki, T. Iguchi, D. A. Short, and K. Nakamura, 2012: Incidence-angle dependency of TRMM PR rain estimates. J. Atmos. Oceanic Technol., 29, 192–206, https://doi.org/10.1175/JTECH-D-11-00067.1.
Hirose, M., Y. N. Takayabu, A. Hamada, S. Shige, and M. K. Yamamoto, 2017: Impact of long-term observation on the sampling characteristics of TRMM PR precipitation. J. Appl. Meteor. Climatol., 56, 713–723, https://doi.org/10.1175/JAMC-D-16-0115.1.
Hou, A. Y., and Coauthors, 2014: The Global Precipitation Measurement mission. Bull. Amer. Meteor. Soc., 95, 701–722, https://doi.org/10.1175/BAMS-D-13-00164.1.
Hourngir, D., G. Panegrossi, D. Casella, P. Sanò, L. P. D’Adderio, and C. Liu, 2021: A 4-year climatological analysis based on GPM observations of deep convective events in the Mediterranean region. Remote Sens., 13, 1685, https://doi.org/10.3390/rs13091685.
Iguchi, T., 2017: Improvement of the minimum detectable precipitation echoes with the TRMM Precipitation Radar and the GPM Dual-Frequency Precipitation Radar. 2017 Proc. IEEE Int. Geoscience and Remote Sensing Symp., Fort Worth, TX, IEEE, 2728–2730, https://doi.org/10.1109/IGARSS.2017.8127560.
Iguchi, T., 2020: Dual-Frequency Precipitation Radar (DPR) on the Global Precipitation Measurement (GPM) mission’s Core Observatory. Satellite Precipitation Measurement, V. Levizzani et al., Eds., Advances in Global Change Research, Vol. 1, Springer, 183–192.
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.
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, https://doi.org/10.2151/jmsj.87A.1.
Iguchi, T., and Coauthors, 2021: GPM/DPR level-2. JAXA Algorithm Theoretical Basis Doc., 238 pp., https://www.eorc.jaxa.jp/GPM/doc/algorithm/ATBD_DPR_V07A.pdf.
ITU-R, 1994: Use of the 13.75 to 14.0 GHz band by the space science services and the fixed-satellite service. ITU-R Recommendation SA.1071, 29 pp., https://www.itu.int/dms_pubrec/itu-r/rec/sa/R-REC-SA.1071-0-199407-W!!PDF-E.pdf
ITU-R, 2003: Final acts of the World Radiocommunication Conference. World Radiocommunication Conf. 2004, Geneva, Switzerland, International Telecommunication Union.
ITU-R, 2021: Current and future use of the band 13.25–13.75 GHz by spaceborne active sensors. ITU-R Rep. RS.2068-2, 14 pp., https://www.itu.int/dms_pub/itu-r/opb/rep/R-REP-RS.2068-2-2021-PDF-E.pdf.
Kanemaru, K., T. Iguchi, T. Masaki, and T. Kubota, 2020: Estimates of spaceborne precipitation radar pulsewidth and beamwidth using sea surface echo data. IEEE Trans. Geosci. Remote Sens., 58, 5291–5303, https://doi.org/10.1109/TGRS.2019.2963090.
Kanemaru, K., H. Hanado, and K. Nakagawa, 2021: Improvement of the clutter removal method for the spaceborne precipitation radars. 2021 IEEE Int. Geoscience and Remote Sensing Symp., Brussels, Belgium, IEEE, 840–843, https://doi.org/10.1109/IGARSS47720.2021.9554974.
Kelley, O. A., J. Stout, M. Summers, and E. J. Zipser, 2010: Do the tallest convective cells over the tropical ocean have slow updrafts? Mon. Wea. Rev., 138, 1651–1672, https://doi.org/10.1175/2009MWR3030.1.
Kim, M., J. Im, H. Park, S. Park, M.-I. Lee, and M.-H. Ahn, 2017: Detection of tropical overshooting cloud tops using Himawari-8 imagery. Remote Sens., 9, 685, https://doi.org/10.3390/rs9070685.
Kojima, M., and Coauthors, 2012: Dual-Frequency Precipitation Radar (DPR) development on the GPM Precipitation Measurement (GPM) Core Observatory. Proc. SPIE, 8528, 85281A, https://doi.org/10.1117/12.976823.
Kozu, T., and Coauthors, 2001: Development of precipitation radar onboard the Tropical Rainfall Measuring Mission (TRMM) satellite. IEEE Trans. Geosci. Remote Sens., 39, 102–116, https://doi.org/10.1109/36.898669.
Kubota, T., T. Iguchi, M. Kojima, L. Liao, T. Masaki, H. Hanado, R. Meneghini, and R. Oki, 2016: A statistical method for reducing sidelobe clutter for the Ku-band Precipitation Radar on board the GPM Core Observatory. J. Atmos. Oceanic Technol., 33, 1413–1428, https://doi.org/10.1175/JTECH-D-15-0202.1.
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.
Li, J., and K. Nakamura, 2002: Characteristics of the mirror image of precipitation observed by the TRMM Precipitation Radar. J. Atmos. Oceanic Technol., 19, 145–158, https://doi.org/10.1175/1520-0426(2002)019<0145:COTMIO>2.0.CO;2.
Li, J., and K. Nakamura, 2005: Vertical distribution of the mirror image returns observed by TRMM PR and estimated for a 35-GHz radar. J. Atmos. Oceanic Technol., 22, 1829–1837, https://doi.org/10.1175/JTECH1819.1.
Liao, L., and R. Meneghini, 2022: GPM DPR retrievals: Algorithm, evaluation, and validation. Remote Sens., 14, 843, https://doi.org/10.3390/rs14040843.
Liu, C., and E. J. Zipser, 2005: Global distribution of convective penetrating the tropical tropopause. J. Geophys. Res., 110, D23104, https://doi.org/10.1029/2005JD006063.
Liu, N., C. Liu, and L. Hayden, 2020: Climatology and detection of overshooting convection from 4 years of GPM Precipitation Radar and passive microwave observations. J. Geophys. Res. Atmos., 125, e2019JD032003, https://doi.org/10.1029/2019JD032003.
Masaki, T., T. Iguchi, K. Kanemaru, K. Furukawa, N. Yoshida, T. Kubota, and R. Oki, 2022: Calibration of the Dual-Frequency Precipitation Radar onboard the Global Precipitation Measurement Core Observatory. IEEE Trans. Geosci. Remote Sens., 60, 5100116, https://doi.org/10.1109/TGRS.2020.3039978.
Nakamura, K., 2021: Progress from TRMM to GPM. J. Meteor. Soc. Japan, 99, 697–729, https://doi.org/10.2151/jmsj.2021-035.
Petersen, W. A., P.-E. Kirstetter, J. Wang, D. B. Wolff, and A. Tokay, 2020: The GPM ground validation program. Satellite Precipitation Measurement, V. Levizzani et al., Eds., Advances in Global Change Research, Vol. 2, Springer, 471–502.
Proud, S. R., and S. Bachmeier, 2021: Record-low cloud temperatures associated with a tropical deep convective event. Geophys. Res. Lett., 48, e2020GL092261, https://doi.org/10.1029/2020GL092261.
Roca, R., and Coauthors, 2021: The joint IPWG/GEWEX precipitation assessment. WCRP Rep. 2/2021, 125 pp., https://www.wcrp-climate.org/WCRP-publications/2021/Joint_IPWG-GEWEX_Precipitation_Assessment_web.pdf.
Seto, S., 2022: Examining the consistency of precipitation rate estimates between the TRMM and GPM Ku-band radars. SOLA, 18, 53–57, https://doi.org/10.2151/sola.2022-009.
Seto, S., T. Iguchi, R. Meneghini, J. Awaka, T. Kubota, T. Masaki, and N. Takahashi, 2021: The precipitation rate retrieval algorithms for the GPM Dual-Frequency Precipitation Radar. J. Meteor. Soc. Japan, 99, 205–237, https://doi.org/10.2151/jmsj.2021-011.
Takahashi, H., Z. J. Luo, and G. L. Stephens, 2017: Level of neutral buoyancy, deep convective outflow, and convective core: New perspectives based on 5 years of CloudSat data. J. Geophys. Res. Atmos., 122, 2958–2969, https://doi.org/10.1002/2016JD025969.
Takahashi, N., and T. Iguchi, 2008: Characteristics of TRMM/PR system noise and their application to the rain detection algorithm. IEEE Trans. Geosci. Remote Sens., 46, 1697–1704, https://doi.org/10.1109/TGRS.2008.916205.
Takayabu, Y. N., 2002: Spectral representation of rain profiles and diurnal variations observed with TRMM PR over the equatorial area. Geophys. Res. Lett., 29, 1584, https://doi.org/10.1029/2001GL014113.
TRMM Precipitation Radar Team, 2011: Tropical Rainfall Measuring Mission (TRMM) Precipitation Radar algorithm: Instruction manual for version 7. JAXA–NASA Doc., 175 pp., https://www.eorc.jaxa.jp/TRMM/documents/PR_algorithm_product_information/pr_manual/PR_Instruction_Manual_V7_L1.pdf.
