1. Introduction
Finescale rainfall information over undulating landscapes is important for improving hydrological modeling and evaluating spatial representativeness of various observational data (e.g., Yamamoto et al. 2011). Differences were observed among satellite and ground observation datasets with respect to regional heterogeneity. This could be due to measurement difficulties and the paucity of samples. Typically, microwave radiometer algorithms exhibit geographical dependence because of differences in the background surface emission (McCollum and Ferraro 2005; Mega and Shige 2016; Yamamoto et al. 2017) and a priori profile model (Shige et al. 2013, 2014; Shige and Kummerow 2016). It is therefore difficult to determine finescale rainfall features over land because of the aforementioned retrieval uncertainties and relatively low resolution of observations (Dinku and Anagnostou 2005; Liu and Zipser 2014; Seo et al. 2015; Wang et al. 2014). Thus, there is an urgent need for studies that focus on high-quality finescale global rainfall data.
The Precipitation Radar (PR) on board the Tropical Rainfall Measuring Mission (TRMM) over a 16-yr life-span has measured various precipitation types over the tropics globally. However, TRMM PR data also exhibited local retrieval uncertainties because of complex clutter interferences (e.g., Hamada and Takayabu 2014, hereinafter HT14; Hirose et al. 2012; Takahashi and Oki 2010). Nevertheless, data from the active sensor have an advantage of detecting finescale features around mountains and coastlines because retrievals are less sensitive to the background environment than those of passive microwave or infrared radiometers (Dinku and Anagnostou 2005; Ogino et al. 2016; Shige et al. 2013). Thus, it is expected that overall data based on homogeneous observations over land and oceans provide refined benchmark climatologies of rainfall (Biasutti et al. 2012; Adler et al. 2007). Therefore, it is important to determine the sufficiency of the full TRMM PR product. Additionally, it is necessary to perform a quantitative analysis of sampling properties of short-range subsets to better understand long-term variation and reproducibility of detected regional characteristics. The TRMM project was originally designed for 3 yr. A major objective of the TRMM project included estimating monthly rainfall at 5° horizontal resolution with an accuracy of 10% (Oki and Sumi 1994; Simpson et al. 1988). A number of studies utilized short-range data to infer climatological fields (e.g., Adler et al. 2003; Kidd 2001; Kummerow et al. 2000; Rasmusson et al. 2006). However, there is a paucity of research focusing on robustness of regional characteristics based on data during a limited period and mean feature differences based on data accumulation. Long-term operation of the satellite enables the examination of sampling properties of subsets with respect to finescale precipitation system climatology.
An ongoing debate persists on the sampling sufficiency of finescale gridded products by a low-orbital satellite for climatic use. This is because highly fluctuating precipitation properties affect local uncertainties (Fisher 2007; Fisher and Wolff 2011; Nesbitt and Anders 2009; Yamamoto et al. 2011). In addition to regional validation studies, sampling characteristics of simulated precipitation for orbital geometry were examined to evaluate the effect of orbital sampling at arbitrary scales and the differences between satellite-derived products and model outputs (Guan et al. 2013; Kotsuki et al. 2014; Lin et al. 2002). Related studies offer an adequate comparison of model outputs with satellite data collected over a limited and specific time period. However, understanding of precipitation regimes based on observations and evaluation of model performance continue to be important challenges. Furthermore, satellite observation is required for monitoring climate change as well as for integrated data mapping. For example, interannual climate variability was examined using global mean precipitation data (Liu and Allan 2012; Wang et al. 2008). Dividing data into finer spatiotemporal intervals and precipitation types is beneficial as it reduces uncertainty in interpreting variability, but it increases uncertainty of sampling errors related to local precipitation characteristics. Hirose et al. (2008) demonstrated that detection capability of a local diurnal feature improved with the data accumulation period. They also indicated that relatively short-term observations included significant sampling errors in hourly statistics with respect to spatiotemporal incoherent diurnal features at finescale resolution and suggested a noise screening method based on a type of texture method. However, it was difficult to extract finescale climatic features of rare but significant rainfall events, such as those related to tropical cyclones, despite the availability of an entire dataset (e.g., Hirose et al. 2009). Specifically, with respect to finescale regional characteristics, it is necessary to evaluate data sufficiency of the long-term TRMM PR record based at least on the number of major storms to differentiate signals from occasionally sampled noise. It is also necessary to understand the detectability of precipitation systems observed by the TRMM PR to discuss current and future satellite precipitation measurements. Further research is required to develop precipitation climatology based on unified products of the TRMM PR and the dual-frequency precipitation radar (DPR) onboard the Global Precipitation Measurement (GPM; Hou et al. 2014) core observatory. Additionally, impact evaluation of additional high-resolution precipitation data or frequent but low-resolution storm sampling provides useful information for future space-based precipitation observation strategies.
This study traced the history of the TRMM PR’s climatological database. A component of this study included a follow-up to Hirose et al. (2008, 2009) and Hirose (2015) with a focus on the impact of resolution enhancement to a 0.1° scale and usage of an almost complete period of TRMM PR data. Thus, the present study aimed to quantify the degree to which detection of climatological features of rainfall improved in accordance with an increase in sampling size.
This paper is organized as follows: Section 2 introduces the features of a storm database. Section 3 provides a description of sampling adequacy of the finescale database. Concluding remarks are presented in section 4.
2. Data and methodology
This study used TRMM PR 2A25 version-7 data obtained between 1998 and 2013. The main parameter corresponded to the estimated surface rainfall rate (Iguchi et al. 2009). In the 2A25 version-7 algorithm, the rainfall rate was estimated based on attenuation-corrected effective radar reflectivity factor Ze, with Ze–R relationships depending on rain type and height that accounted for the nonuniform beamfilling (NUBF) effect and uncertainty related to drop size distribution by an adjustment parameter ε [details can be found in Iguchi et al. (2009) and TRMM PR Team (2011)]. The estimated surface rainfall rate was calculated based on Ze at the lowest level in the clutter-free region accounting for pressure correction on terminal velocity and evaporation near the surface. The sampling properties of the TRMM PR were poor because of the narrow swath. Additionally, there were retrieval uncertainties related to the attenuation correction of severe storms, drop size distribution variety, adequacy of NUBF correction, orbit boost effect, and incidence angle dependency (e.g., Iguchi et al. 2009; Hirose et al. 2012; Short and Nakamura 2010). Nevertheless, rainfall retrievals were in principle superior to those from passive sensors, particularly over land, as noted earlier. In this study, the aforementioned retrieval errors (with the exception of artificial echoes on ground clutter interferences) were not discussed. This section describes sampling properties, a storm database, and the filtering technique.
a. Number of overpasses
The number of satellite overpasses Npass and footprints Nfov per day for each 0.1° grid box [for a TRMM PR swath width of 215 (245) km before (after) the orbit boost in August 2001] were used as a reference relative to the sampling properties of the TRMM PR data for climatic use. Specifically, Npass was defined as the number of instantaneous snapshots for each grid with at least one centroid of the TRMM PR footprint. The horizontal resolutions of the PR footprint at the nadir were approximately 4.3 and 5 km before and after the orbit boost, respectively. The vertical resolution was 250 m. The TRMM PR observations ranged from 36.3°S to 36.3°N. Figure 1 shows the latitude dependence of their numbers. The zonally averaged Npass corresponded to 0.3 in the tropics (from 13.2°S to 13.7°N; 37% of the observable domain). The area fraction in which Npass exceeded 1 was approximately 8%. The most frequent zones (1.8–1.9 times per day) were located over the latitude belts at 34.3°–33.9°S and 34.0°–34.2°N and corresponded to approximately 1% of the overall domain. Approximately one overpass every three days was observed in 55% of the domain (19.9°S–20.3°N), and one overpass every two days was observed in 28% of the domain (30.1°–19.9°S and 20.3°–30.3°N). For example, observations over the “Maritime Continent” in which the TRMM overpasses averaged 0.3 day−1 archived to approximately 1800 snapshots over 16 yr. On average, a 0.1° grid included approximately 5.5 footprints. The number of footprints for a certain spatiotemporal scale could be derived from the number of overpasses multiplied by average number of footprints (Nfov per 0.1) for each grid box. For example, approximately one million (2500) samples were obtained for a 1° (0.05°) grid box for the Maritime Continent over 16 yr. It should be noted that the number of overpasses at 20°N–20°S was approximately equivalent to only one-third of the number of observation days, although the number of samples was high for a coarse grid. Regional samples, specifically at hourly and monthly scales, were not always sufficient as a result of the rare occurrences of high-impact storms. Thus, sampling sufficiency evaluation requires in-depth information such as the number of precipitation events, statistical contribution of significant storms to total rainfall, and the signal coherency. These will be described in detail.
Latitude dependence of the number of TRMM overpasses and samples per day for each 0.1° grid. Thick solid, dashed, and thin solid lines indicate Npass, Nfov, and the number of footprints in a 0.1° grid per overpass (Nfov/Npass), respectively.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0115.1
b. TRMM PR–captured precipitation systems
In this study, storm information was investigated to examine sampling sufficiency and clarify the regional characteristics of signals. The PR-captured precipitation system (PR-PS) was determined as contiguous rain pixels where rain pixels were detected within a distance of 10 km (Hirose et al. 2009). The total number of all pixels, rain pixels, and PR-PSs approximately corresponded to 41 billion, 1.5 billion, and 79 million, respectively, over 16 yr. The database was designed to understand the composition of scale-based systems of TRMM PR–derived rainfall. Because of the swath-width restriction, some PR-PSs were equivalent to a truncated component of an entire precipitation system. The PR-PSs were grouped into three categories based on the horizontal scale defined with a diameter of an area-equivalent circle. The categories included small PR-PSs (with an area-equivalent diameter < 10 km), large PR-PSs (with an area-equivalent diameter > 100 km), and medium PR-PSs (the remaining areas). The smallest PR-PS was composed of one rain-certain pixel. The rain cell scale with solitary rainy fields of view was equivalent to footprint size. The size of each event was individually recognized, and data pertaining to consecutive rainy areas were compiled for each 0.1° grid with information on scale category. Long-term monthly and hourly PR-PS databases were then generated. It should be noted that the scan edge truncated approximately 88% of large PR-PSs with a scale of 100–200 km. In the study, version 2.2 of the PR-PS database was used (available at the study website online at https://www.rain-clim.com). The website provides visual tools with options for zoom, transparency, temporal sliders, and a tree for research products such as the time of maximum rainfall and the retrieval error (addressed in the following subsection). The results indicated that finescale geographic rainfall features were represented in a considerably clearer manner than those from a 10-yr storm database in a previous study. The previous 10-yr storm dataset was constructed using individual storm information of the centroid location and rainfall area. For example, rainfall by a PR-PS was archived into grids near the storm center as determined by area-equivalent diameter (Hirose et al. 2009). In contrast, the current database utilized pixel-based geographic information to resolve finescale rainfall patterns, particularly for widespread systems.
c. Clutter filtering
Sampling error reduction enhances detectability of climatological features and spatially fixed retrieval errors, particularly in mountainous areas. HT14 demonstrated suspicious extreme rainfall signals in the PR 2A25 version-7 product. In the present study, a removal filter developed by HT14 for clutter-contaminated signals was applied to the PR-PS database to mitigate clutter contaminations in a complex terrain. Figure 2 shows the impact of filtering on total rainfall. The contours were generated based on Shuttle Radar Topography Mission global bathymetry and elevation data at 30 arc s resolution (SRTM30; Farr et al. 2007) data. The impact on 1° averaged rainfall statistics was insignificant over all regions except for a few steep orographic or low rainfall areas in which approximately 2% of the 1° averaged rainfall was modified. The filtering effect grew significantly more pronounced for orographic rainfall at higher resolutions. With respect to 0.1°-scale climatology, more than 10% of the rainfall was filtered out over scattered areas at elevations of 1000 m. On average, the filtering effect was approximately 2% for rugged regions with 1000–2000 m of peak-to-average height in a 0.1° grid.
HT14-filtering effect on (top) 1°- and (bottom) 0.1°- resolution rainfall. Numbers in the color scale denote the percent fraction of observed rainfall that was filtered. The bottom panel focuses on the Himalaya area, delineated by the small rectangle in the top panel. The blue line indicates an altitude of 1000 m.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0115.1
3. Sampling sufficiency of rainfall means
a. Data accumulation of high-impact storms
Large PR-PSs occur infrequently but account for more than half of rain samples and rainfall on average (Hirose 2015). The types of storms that have the largest contribution to rain samples are termed high-impact storms. The impact varies from region to region. Major rainy regions tend to have a greater number of large-PR-PSs, particularly in the midlatitudes. In contrast, occurrence frequency is very low over low-rainfall regions. This section deals with the impacts of prevailing systems on the characterization of finescale rainfall patterns.
Figures 3a and 3b show the occurrence frequencies in which each 0.1° grid is covered by small and large PR-PSs, respectively, over 16 yr. The number of gridded large systems was significant because of the substantial spatial extent, although the number of large PR-PSs was considerably low relative to that of small PR-PSs as shown by Hirose et al. (2009). It should be noted that the larger number of PR-PSs in midlatitudes could essentially be attributed to frequent orbital overlap, and the few PR-PSs in the middle of Australia were due to the observation gap to prevent radio wave interference. Small PR-PSs were widely observed over land and oceans (Fig. 3a). The region with south-facing slopes of the Himalayas was the most obvious region for the occurrence of small convection. These isolated storms develop on the slope as nonshallow convection, and therefore overland concentration was excluded from the geographical distribution of shallow isolated cores investigated by Houze et al. (2015). Convection-suppressed areas corresponded to smaller numbers of PR-PSs. Small PR-PSs generally occurred more frequently over open oceans than over coastal oceans or land. Over land, the frequency was relatively high over wet regions such as East and Southeast Asia and the Amazon. There were no large PR-PSs over dry regions such as the western part of the Tibetan Plateau and off Peru. In contrast, several hundred large PR-PSs were observed over significant rainfall regions (Fig. 3b). Specifically, more than 900 PR-PSs accumulated over the northwest Pacific where both the number of TRMM overpasses and the occurrence frequency of extratropical cyclones were high. This sampling information demonstrated a significant benefit of long-term data accumulation of finescale precipitation climatology in the field wherein high-impact rainfall events were more obtainable as the PR dataset now covered a span exceeding 16 yr. The number of large high-impact systems provided simple but useful information to understand PR climatology components.
(a),(b) Occurrence frequencies and (c),(d) occurrence fractions of rain samples of small [in (a) and (c)] and large [in (b) and (d)] PR-PSs for each 0.1°.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0115.1
Additionally, the occurrence fractions of rain samples of small and large PR-PSs are shown in Figs. 3c and 3d, respectively. The spatial features resemble Figs. 3a and 3b, respectively, except for the reduction of orbital-induced signal concentration in midlatitudes. The occurrence fraction of 1% corresponded to a rainfall rate of 14.4 min day−1. The occurrence fractions of small-PR-PS rainfall were less than 0.5% over most land areas. It reached approximately 1% over specific low-rainfall oceans. In a manner distinct from the distribution of shallow storms (Short and Nakamura 2000; Schumacher and Houze 2003), nonshallow isolated systems occur frequently at the margin of rainy areas in the western Pacific. Small PR-PSs occurred to the east of continents in tropical oceans (except for the northwest Arabian Sea, which was affected by large-scale subsidence). This offered referential information on precipitation climatology of several-kilometers-scale storms while requiring continuing attention with respect to the TRMM PR deficiency in detecting light drizzles from low clouds in shallow planetary boundary layers (Hamada and Takayabu 2016; Hirose et al. 2012; Wood 2012). The fractions by large PR-PSs exceeded those by small PR-PSs over major rainy areas. For example, the fractions exceeded 1% over western as well as northwestern Pacific given significant zonal differences in the sample numbers as shown in Fig. 3b.
As described above, the impact of large PR-PSs on the total number of samples was high. However, the number was not always high. The spatial pattern of rainfall was examined to show the sampling adequacy of large PR-PSs for long-term data. Figure 4 presents maps of the rain fraction of large PR-PSs generated for periods including 1, 3, and 16 yr. As shown in Fig. 4c, at a finescale of 0.1°, the long-term TRMM PR data provided reasonable accuracy in detecting spatial variations in rainfall based on samples of multiple high-impact events. With respect to midlatitudes, more than 80% of the rainfall resulted from large PR-PSs over most parts of eastern continents and adjacent oceans. For example, fractions exceeded 90% over certain areas such as La Plata basin and East China Sea. Furthermore, the rain fraction by large PR-PSs exceeded 80% in the southeastern part of the United States. Conversely, there was no significant concentration over the ocean to the east of the Australian continent. The organized systems occurred over the western ocean margin related to warm ocean currents or the margins of high orography and within large-scale subtropical convergence zones as demonstrated in previous studies (e.g., Nesbitt et al. 2006; Ulbrich et al. 2009). The region to the east of the Australian continent did not satisfy either of these conditions. In the tropics, the fraction over coastal ocean exceeded that over land (except for central Africa). More than half of the rainfall resulted from large PR-PSs over the majority of rainy regions and indicated that a sufficient accumulation of large PR-PSs was extremely important. The large-PR-PS rain fraction corresponded to 38% over whole areas in which the rainfall rate was <1 mm day−1.
Rain fraction by large PR-PSs during (a) 1998, (b) 1998–2000, and (c) 1998–2013. Gray shading indicates the area where small PR-PSs resulted in more than 90% rainfall.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0115.1
Short-term statistics exhibited extreme fluctuations because of the small number of samples (Fig. 4a). With respect to 3-yr statistics (Fig. 4b), sampling uncertainties were demonstrated in the occasional pattern of substantial variations (of several tens of percentage points) as compared with neighboring grids. Short-term climatology constructed by several instantaneous snapshots exhibited significant uncertainty in relation to a discussion of the spatial coherency.
Small PR-PSs constituted more than half the rainfall off Peru and Angola where rainfall was less than 0.5 mm day−1. The gray shading in Fig. 4 shows that the small-PR-PS fraction exceeded 90% for oceans with rainfall of less than 0.1 mm day−1. The spatial extent of the small-PR-PS prevailing regime over cool waters or areas of extreme aridity was gradually corrected in line with data accumulation. On the other hand, large PR-PSs had significant impact on total rainfall (Fig. 4c) over other low-rainfall regions (<0.5 mm day−1) near the intertropical convergence zone, such as the Sahel and off the west coast of Mexico, despite their low occurrence frequency (Fig. 3b). Prevailing systems appeared to correspond to large PR-PSs over the marginal zone. This implied the nonnegligible impact of rare but significant systems, such as mesoscale convective systems and tropical cyclones, in these regions.
Evidently, the sampling deficiency of high-impact storms for short-term data resulted in the deterioration of spatially coherent climatological features. However, sampling uncertainties of the period average were unclear because of complex precipitation behavior. To understand sampling properties for a given time period, 16-yr statistics were used as a best estimate to assess robustness in the rain fraction of large PR-PSs accumulated over time. The degree of similarity (hereinafter referred to as “concordance ratio”) to the best estimate was used to discuss the stability of stored data statistical characteristics for multiple years (Fig. 5). The concordance ratio was defined as the proportion of grid cells in which the large-PR-PS rain fraction was within ±10% of the corresponding 16-yr value. The concordance ratio illustrated the clarity of finescale precipitation climatology according to the sequential data period, although it was potentially associated with year-to-year variability in rainfall data during early years (e.g., Nakazawa and Rajendran 2009).
Concordance ratio of the large-PR-PS rain fraction of elapsed years with respect to 16-yr statistics. Solid lines indicate the ratio based on rainfall at a 0.1° scale. The dotted (dashed) line corresponds to that for coarse grid data with a resolution of 0.5° (2.5°).
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0115.1
The concordance ratio for all regions (36°N–36°S; denoted by the thick gray line in Fig. 5) was only 28% and 48% for years 1 and 3, respectively. The value increased almost linearly year by year, and the 12-yr statistics were in close agreement with the 16-yr statistics over more than 90% of the observable areas. On average, a year of observation added five large PR-PSs to each grid, although the number varied with latitude. The additional samples slightly affected the 0.1° average even over the entire period in view of the increase in the coincidence rate. The impact was less significant with respect to global 0.5° and 2.5° statistics. The 1- and 3-yr averages at the 0.5° (2.5°) scale were consistent with long-term averages over 40% (57%) and 64% (80%) of the area, respectively. More than 12 (7) yr of data at a 0.5° (2.5°) scale provided an accurate rainfall fraction by large PR-PSs with a small spatial variation corresponding to <1% yr−1. Given a variation of 2% yr−1, more than 13, 9, and 6 yr of data were sufficient for statistics at scales of 0.1°, 0.5°, and 2.5°, respectively.
With respect to East Asia (the eastern China/Japan area), wherein both the number of overpasses and occurrence frequency of large PR-PSs are high, the 5-yr statistics showed a concordance of 90%. Specifically, even 1 yr and 3 yr of observations involved 59% and 81% statistics, respectively, that were in agreement with the best estimate. The 11-yr statistics were mostly in agreement with the 16-yr statistics. A similar tendency applied to the Indian Ocean and other tropical regions. In contrast, the impact of additional samples still appeared significant over subtropical high pressure belt regions such as the Sahel. Additionally, as shown in Fig. 4c, this was attributed to a chronic shortage of few but significant large-PR-PS samples in these regions. Another example included a high concordance ratio that was observed over the ocean off Peru in which an extremely low number of large PR-PSs were observed, and the number of large PR-PSs per 0.1° only corresponded to 0.2 over 16 yr. In this area characterized by the largest subtropical stratocumulus deck (Wood et al. 2011), under the dominant small-PR-PS regime, 95% of the 1-yr statistics were in agreement with long-term statistics in terms of the large-storm rain fraction.
b. Consistency of rainfall climatology averaged over time
Robust high-resolution climatology is composed of a large volume of data. The sampling uncertainty of the short-term averaged rainfall was described by the distribution condition of rainfall anomaly over the 16-yr statistics. Figure 6 shows the 25th, 50th, and 75th percentiles of the rainfall anomaly for each period and grid resolution. The box length showing 50% of the grids shrank with increased sampling. With respect to first-year data, the central 50% of rainfall anomaly at a 0.1° (2.5°) scale corresponded to between −51.6% (−23.2%) and 39.9% (20.0%). Sampling error reduction for coarse grids was not directly proportional to the number of increased rainfall samples (resulting from grid expansion) because of spatial coherency (e.g., Moron et al. 2007). With respect to the 10-yr mean, the middle 50% of the anomaly for a 0.1° scale ranged from −10% to 10%, which was approximately twice that for a 2.5° scale.
Interquartile range of the rainfall anomaly of the 16-yr statistics at 0.1°, 0.5°, and 2.5° spatial resolutions. The first and third quartiles are represented by the lower and upper boundaries of the box, respectively. The white line in the box indicates the median.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0115.1
The median of the 0.1°-scale rainfall anomaly observed in 1998 corresponded to −14.7%. The preliminary investigation for elapsed years beginning with an arbitrary year indicated that the medians did not vary significantly but that the third quartiles fluctuated by several percent from year to year. The average and standard deviations of the 25th, 50th, and 75th percentiles for each year from 1998 to 2013 were −49.7% ± 1.3%, −15.0% ± 1.6%, and 32.0% ± 3.6%, respectively. The study highlighted time series statistics observed by the TRMM PR. With respect to a 0.1° scale, the median was negative for shorter time records (≤7 yr) and slightly positive for longer periods. Underestimated (overestimated) regions occurred widely (sporadically) for short-term statistics, and this could be attributed to the detectability of rare and high-impact systems. The median value at 0.5° and 2.5° scales was positive in the third and second year, respectively. The coarse rainfall averages for more than 2 yr were equally divided into underestimates and overestimates. The data periods into which the first and third quartiles fell within ±10% (±5%) corresponded to 10 (14), 7 (12), and 5 (8) yr for 0.1°, 0.5°, and 2.5° average scales, respectively.
Figure 7 shows the areal fraction of robust rainfall climatology at 0.1° and 2.5° scales based on elapsed years. The robustness of climatological values was determined by the difference between the average for a given time and for all 16 yr, with uncertainties of 5%, 10%, and 50%. The target area corresponded to 36.3°N–36.3°S, and color was used to denote the range after the removal of low-rainfall areas. With respect to a 0.1° scale with 10% accuracy, 20% (50%) of the entire domain showed 3-yr (10 yr) averaged rainfall that was consistent with that of the 16-yr average. The decelerating trend in percentage change was observed for the 2.5° scale with 10% accuracy and for the 0.1° and 2.5° scales with 50% accuracy. Conversely, consistent probability for a 0.1° scale with 5% or 10% accuracy increased at an accelerated rate with the last few years of data, and this implied a degree of sampling error in long-term mean values. The difference for an additional year was evident over low-rainfall areas (not shown in the present study). The fraction of areas with rainfall >1 (2) mm day−1 corresponded to 69% (51%). Excluding the areas with rainfall <1 mm day−1 increased the areal fraction of accurate climatology by 7% with respect to 10-yr statistics of 0.1°-scale rainfall with 10% accuracy. The fraction was increased by an additional 3% over areas with rainfall >2 mm day−1, and year-by-year variations indicated a monotonic increase. Thus, finescale rainfall climatology was not fully converged but asymptotically approached actual conditions.
Year-to-year variation in the areal percentage of rainfall close to 16-yr statistics for each resolution and precision. The black lines indicate the variations for the entire domain. The cyan and blue lines indicate the areas with rainfall greater than 1 and 2 mm day−1, respectively.
Citation: Journal of Applied Meteorology and Climatology 56, 3; 10.1175/JAMC-D-16-0115.1
4. Summary and conclusions
Finescale geographical distribution of precipitation climatology was refined based on a large number of snapshots observed by the TRMM PR. This study performed corrections to the clutter-contaminated signals developed by HT14. The effect of the removal filter was negligible over most regions but significant at high spatial resolution over specific geographical areas such as 1000-m altitudes in the southern slopes of the Himalayas. The finescale storm dataset and online visualization tool are available to potential users at our website (http://www.rain-clim.com).
The sampling adequacy for detecting local climatic signals was discussed based on the number of high-impact precipitation systems and geographical coherency. The spatial coherency of the rain fraction by the large PR-PSs (≥100 km) indicated that the short-term finescale statistics included a significant percentage of anomalies associated with inadequate sampling or interannual variations when compared with long-term statistics with causal continuity. For example, only half of the 3-yr statistics were consistent with the 16-yr statistics. The short-term TRMM PR data consistently underestimated a mean value over more than half of the areas in any given year, and this was potentially caused by insufficient sampling of high-impact storms. More than 100 large PR-PSs were generally observed over major rainfall areas. In view of the slight coincidence variation of the large-system rain fraction with a high degree of accuracy (±10%), the 0.1° sampling for the entire-term climatology approached sufficiency, particularly in areas dominated by large-scale storms such as storm-track locations east of Japan.
Based on the data for each elapsed year, the study also described differences in the gridded rainfall climatology. Approximately half of the 0.1°-scale rainfall data based on the 1-yr observations differed from long-term averages at a rate exceeding 50%. The anomaly over half of the areas was reduced to 10% in the 10-yr data. The interquartile range of the difference in the 0.1°-scale period average rates was approximately twice that for 2.5° scales. The median of differences from 16-yr statistics indicated that short-term samples reduced gridded averaged rainfall because of an insufficiency of high-impact storm samples over a wide area. Given the negatively biased median, the terms of insufficient sample size of high-impact storms corresponded to 7, 2, and 1 yr for the 0.1°, 0.5°, and 2.5° average scales, respectively. It is important to note this sampling insufficiency with respect to significant systems to argue robust local characteristics based on limited data from not only satellites but also field experiments for ground validation and model simulations.
Another index of data sufficiency included in this study involved a quadratic differential in areal percentages of gridded rainfall climatology beyond a given accuracy and relative to time. The results indicated that there was less deviation in the rainfall average with an increase in the number of samples. Sampling error reduction was not directly proportional to the grid scale because of a lack of storm sample independence across the resolutions. The rate of percentage change indicated that rainfall at a 0.1° scale with 50% accuracy reached the mean level of the 16-yr data after several years of accumulation, particularly over rainy regions. However, 0.1°-scale rainfall climatology with an accuracy <10% continued to be sensitive to further data accumulation. With respect to higher resolutions of space and time or higher precision, sampling could be deficient based on the location.
Thus, very long-term TRMM PR data highlighted the possibilities and uncertainties for multiscale analyses using finescale global information. Unified products of the TRMM PR and GPM DPR will expand detection capability of extremely occasional events that have an enormous impact. Continued high-quality observations will enable further improvements in the capture capabilities of precipitation regimes. This is clearly beneficial for dynamic climatology and the availability of various applications.
Acknowledgments
The authors express their gratitude to the members of the TRMM and GPM projects. The constructive comments of two anonymous reviewers are greatly appreciated.
REFERENCES
Adler, R. F., C. Kummerow, D. Bolvin, S. Curtis, and C. Kidd, 2003: Status of TRMM monthly estimates of tropical precipitation. Cloud Systems, Hurricanes, and the Tropical Rainfall Measuring Mission (TRMM), Meteor. Monogr., No. 51, Amer. Meteor. Soc., 223–234.
Adler, R. F., and Coauthors, 2007: Tropical Rainfall Measuring Mission (TRMM) senior review proposal. NASA Goddard Space Flight Center Rep., 47 pp. [Available online at https://pmm.nasa.gov/sites/default/files/document_files/TRMMSenRev2007_pub.pdf.]
Biasutti, M., S. E. Yuter, C. D. Burleyson, and A. H. Sobel, 2012: Very high resolution rainfall patterns measured by TRMM Precipitation Radar: Seasonal and diurnal cycles. Climate Dyn., 39, 239–258, doi:10.1007/s00382-011-1146-6.
Dinku, T., and E. N. Anagnostou, 2005: Regional differences in overland rainfall estimation from PR-calibrated TMI algorithm. J. Appl. Meteor., 44, 189–205, doi:10.1175/JAM2186.1.
Farr, T. G., and Coauthors, 2007: The Shuttle Radar Topography Mission. Rev. Geophys., 45, RG2004, doi:10.1029/2005RG000183.
Fisher, B. L., 2007: Statistical error decomposition of regional-scale climatological precipitation estimates from the Tropical Rainfall Measuring Mission (TRMM). J. Appl. Meteor. Climatol., 46, 791–813, doi:10.1175/JAM2497.1.
Fisher, B. L., and D. B. Wolff, 2011: Satellite sampling and retrieval errors in regional monthly rain estimates from TMI, AMSR-E, SSM/I, AMSU-B, and the TRMM PR. J. Appl. Meteor. Climatol., 50, 994–1023, doi:10.1175/2010JAMC2487.1.
Guan, B., D. E. Waliser, J.-L. F. Li, and A. da Silva, 2013: Evaluating the impact of orbital sampling on satellite–climate model comparisons. J. Geophys. Res. Atmos., 118, 355–369, doi:10.1029/2012JD018590.
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, doi:10.1175/JAMC-D-13-099.1.
Hamada, A., and Y. N. Takayabu, 2016: Improvements in detection of light precipitation with the Global Precipitation Measurement dual-frequency precipitation radar (GPM DPR). J. Atmos. Oceanic Technol., 33, 653–667, doi:10.1175/JTECH-D-15-0097.1.
Hirose, M., 2015: Finescale climatology of widespread precipitation systems observed by TRMM PR. Proc. Geoscience and Remote Sensing Symp. 2015, Milan, Italy, Institute of Electrical and Electronics Engineers, 5138–5141, doi:10.1109/IGARSS.2015.7326990.
Hirose, M., R. Oki, S. Shimizu, M. Kachi, and T. Higashiuwatoko, 2008: Finescale diurnal rainfall statistics refined from eight years of TRMM PR data. J. Appl. Meteor. Climatol., 47, 544–561, doi:10.1175/2007JAMC1559.1.
Hirose, M., R. Oki, D. A. Short, and K. Nakamura, 2009: Regional characteristics of scale-based precipitation systems from ten years of TRMM PR data. J. Meteor. Soc. Japan, 87A, 353–368, doi:10.2151/jmsj.87A.353.
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, doi:10.1175/JTECH-D-11-00067.1.
Hou, A. Y., 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., Jr., 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.
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.
Kidd, C., 2001: Satellite rainfall climatology: A review. Int. J. Climatol., 21, 1041–1066, doi:10.1002/joc.635.
Kotsuki, S., K. Terasaki, and T. Miyoshi, 2014: GPM/DPR precipitation compared with a 3.5-km-resolution NICAM simulation. SOLA, 10, 204–209, doi:10.2151/sola.2014-043.
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.
Lin, X., L. D. Fowler, and D. A. Randall, 2002: Flying the TRMM satellite in a general circulation model. J. Geophys. Res., 107, 4281, doi:10.1029/2001JD000619.
Liu, C., and R. P. Allan, 2012: Multisatellite observed responses of precipitation and its extremes to interannual climate variability. J. Geophys. Res., 117, D03101, doi:10.1029/2011JD016568.
Liu, C., and E. Zipser, 2014: Differences between the surface precipitation estimates from the TRMM Precipitation Radar and passive microwave radiometer version 7 products. J. Hydrometeor., 15, 2157–2175, doi:10.1175/JHM-D-14-0051.1.
McCollum, J. R., and R. R. Ferraro, 2005: Microwave rainfall estimation over coasts. J. Atmos. Oceanic Technol., 22, 497–512, doi:10.1175/JTECH1732.1.
Mega, T., and S. Shige, 2016: Improvements of rain/no-rain classification methods for microwave radiometer over coasts by dynamic surface-type classification. J. Atmos. Oceanic Technol., 33, 1257–1270, doi:10.1175/JTECH-D-15-0127.1.
Moron, V., A. W. Robertson, M. N. Ward, and P. Camberlin, 2007: Spatial coherence of tropical rainfall at the regional scale. J. Climate, 20, 5244–5263, doi:10.1175/2007JCLI1623.1.
Nakazawa, T., and K. Rajendran, 2009: Interannual variability of tropical rainfall characteristics and the impact of the altitude boost from TRMM PR 3A25 data. J. Meteor. Soc. Japan, 87A, 317–338, doi:10.2151/jmsj.87A.317.
Nesbitt, S. W., and A. M. Anders, 2009: Very high resolution precipitation climatologies from the Tropical Rainfall Measuring Mission Precipitation Radar. Geophys. Res. Lett., 36, L15815, doi:10.1029/2009GL038026.
Nesbitt, S. W., R. Cifelli, and S. A. Rutledge, 2006: Storm morphology and rainfall characteristics of TRMM precipitation features. Mon. Wea. Rev., 134, 2702–2721, doi:10.1175/MWR3200.1.
Ogino, S.-Y., M. D. Yamanaka, S. Mori, and J. Matsumoto, 2016: How much is the precipitation amount over the tropical coastal region? J. Climate, 29, 1231–1236, doi:10.1175/JCLI-D-15-0484.1.
Oki, R., and A. Sumi, 1994: Sampling simulation of TRMM rainfall estimation using radar–AMeDAS composites. J. Appl. Meteor., 33, 1597–1608, doi:10.1175/1520-0450(1994)033<1597:SSOTRE>2.0.CO;2.
Rasmusson, E., and Coauthors, 2006: Assessment of the benefits of extending the Tropical Rainfall Measuring Mission: A perspective from the research and operations communities. National Research Council Rep., 104 pp.
Schumacher, C., and R. A. Houze, 2003: The TRMM Precipitation Radar’s view of shallow, isolated rain. J. Appl. Meteor., 42, 1519–1524, doi:10.1175/1520-0450(2003)042<1519:TTPRVO>2.0.CO;2.
Seo, E.-K., S. Hristova-Veleva, G. Liu, M.-L. Ou, and G.-H. Ryu, 2015: Long-term comparison of collocated instantaneous rain retrievals from the TRMM Microwave Imager and Precipitation Radar over the ocean. J. Appl. Meteor. Climatol., 54, 867–879, doi:10.1175/JAMC-D-14-0235.1.
Shige, S., and C. D. Kummerow, 2016: Precipitation-top heights of heavy orographic rainfall in the Asian monsoon region. J. Atmos. Sci., 73, 3009–3024, doi:10.1175/JAS-D-15-0271.1.
Shige, S., S. Kida, H. Ashiwake, T. Kubota, and K. Aonashi, 2013: Improvement of TMI rain retrievals in mountainous areas. J. Appl. Meteor. Climatol., 52, 242–254, doi:10.1175/JAMC-D-12-074.1.
Shige, S., M. K. Yamamoto, and A. Taniguchi, 2014: Improvement of TMI rain retrieval over the Indian subcontinent. Remote Sensing of the Terrestrial Water Cycle. Geophys. Monogr., Vol. 206, Amer. Geophys. Union, 27–42.
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.
Short, D. A., and K. Nakamura, 2010: Effect of TRMM orbit boost on radar reflectivity distributions. J. Atmos. Oceanic Technol., 27, 1247–1254, doi:10.1175/2010JTECHA1426.1.
Simpson, J., R. F. Adler, and G. R. North, 1988: A proposed Tropical Rainfall Measuring Mission (TRMM) satellite. Bull. Amer. Meteor. Soc., 69, 278–295, doi:10.1175/1520-0477(1988)069<0278:APTRMM>2.0.CO;2.
Takahashi, N., and R. Oki, 2010: Development of a digital elevation map using TRMM/PR data. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci., 38, 150–153.
TRMM PR Team 2011: Tropical Rainfall Measuring Mission (TRMM) Precipitation Radar algorithm: Instruction manual for version 7. JAXA–NASA Rep., 170 pp. [Available online at http://www.eorc.jaxa.jp/TRMM/documents/PR_algorithm_product_information/pr_manual/PR_Instruction_Manual_V7_L1.pdf.]
Ulbrich, U., G. C. Leckebusch, and J. G. Pinto, 2009: Extra-tropical cyclones in the present and future climate: A review. Theor. Appl. Climatol., 96, 117–131, doi:10.1007/s00704-008-0083-8.
Wang, J.-J., R. F. Adler, and G. Gu, 2008: Tropical rainfall–surface temperature relations using Tropical Rainfall Measuring Mission precipitation data. J. Geophys. Res., 113, D18115, doi:10.1029/2007JD009540.
Wang, J.-J., R. F. Adler, G. J. Huffman, and D. Bolvin, 2014: An updated TRMM composite climatology of tropical rainfall and its validation. J. Climate, 27, 273–284, doi:10.1175/JCLI-D-13-00331.1.
Wood, R., 2012: Stratocumulus clouds. Mon. Wea. Rev., 140, 2373–2423, doi:10.1175/MWR-D-11-00121.1.
Wood, R., and Coauthors, 2011: The VAMOS Ocean–Cloud–Atmosphere–Land Study Regional Experiment (VOCALS-REx): Goals, platforms, and field operations. Atmos. Chem. Phys., 11, 627–654, doi:10.5194/acp-11-627-2011.
Yamamoto, M. K., K. Ueno, and K. Nakamura, 2011: Comparison of satellite precipitation products with rain gauge data for the Khumb region, Nepal Himalayas. J. Meteor. Soc. Japan, 89, 597–610, doi:10.2151/jmsj.2011-601.
Yamamoto, M. K., I. Tanaka, and S. Shige, 2017: Improvement of the rain/no-rain classification method for microwave radiometers over the Tibetan plateau. IEEE Geosci. Remote Sens. Lett., doi:10.1109/LGRS.2017.2666814, in press.
