A daily gridded precipitation dataset covering a period of more than 57 yr was created by collecting and analyzing rain gauge observation data across Asia through the activities of the Asian Precipitation—Highly Resolved Observational Data Integration Towards Evaluation of Water Resources (APHRODITE) project. APHRODITE's daily gridded precipitation is presently the only long-term, continental-scale, high-resolution daily product. The product is based on data collected at 5,000–12,000 stations, which represent 2.3–4.5 times the data made available through the Global Telecommunication System network and is used for most daily gridded precipitation products. Hence, the APHRODITE project has substantially improved the depiction of the areal distribution and variability of precipitation around the Himalayas, Southeast Asia, and mountainous regions of the Middle East. The APHRODITE project now contributes to studies such as the determination of Asian monsoon precipitation change, evaluation of water resources, verification of high-resolution model simulations and satellite precipitation estimates, and improvement of precipitation forecasts. The APHRODITE project carries out outreach activities with Asian countries, and communicates with national institutions and world data centers. We have released open-access APHRO_V1101 datasets for monsoon Asia, the Middle East, and northern Eurasia (at 0.5° × 0.5° and 0.25° × 0.25° resolution) and the APHRO_JP_V1005 dataset for Japan (at 0.05° × 0.05° resolution; see www.chikyu.ac.jp/precip/ and http://aphrodite.suiri.tsukuba.ac.jp/). We welcome cooperation and feedback from users.

An international cooperative program collected and analyzed rain gauge observations from thousands of Asian stations in addition to those reporting to the WMO Global Telecommunications System, creating a 57-year daily precipitation dataset.

Precipitation is one of the most basic meteorological elements and it directly and indirectly affects human life. It is thus crucial to be able to forecast local precipitation variability according to our changing climate. Recent computer resources and techniques have been used in simulating hydrological changes, such as changes in river runoff and extreme events resulting from changes in precipitation. For such purposes, a gridded precipitation dataset is necessary. In addition, water resources in mountains are important, especially for arid regions in the Middle East, the western part of the Himalayas, and central Asia. In such areas, representation of orographic precipitation on a submonthly time scale is important for use in hydrological models and for verification/modification of simulated precipitation by high-resolution climate models.

National hydrological and meteorological services (NHMs) exchange real-time data, including precipitation under a World Meteorological Organization (WMO) contract. Real-time data, such as radiosonde records, reported via the Global Telecommunication System (GTS) are important in making forecasts.

Recently, satellite mapping patterns have been used to forecast weather and issue disaster warnings. The combined products of the GTS and satellites are used for research and as near-real-time data. Popular monthly and pentad precipitation products—the Climate Prediction Center (CPC) Merged Analysis of Precipitation (Xie and Arkin 1997) and Global Precipitation Climatology Project (GPCP)—are based on precipitation data recorded at 6,000 stations worldwide and transmitted via the GTS and satellite-derived estimates.

The GPCP has led to the creation and release of global precipitation data products, and, recently, the spatial and temporal resolution of satellite products has improved. However, principally, estimation over land areas is difficult. Hence, to estimate precipitation quantitatively and assess changes in the occurrences and magnitudes of extreme events, rain gauge observation records over land areas are required. In Asia, despite there being highly populated regions and a great need for precipitation monitoring and forecasts, large differences in estimates among datasets have been reported (e.g., Yatagai et al. 2005). There are several monthly precipitation products (Chen et al. 2002; Mitchell and Jones 2005; Matsuura and Willmott 2012; Schneider et al. 2008). However, at the beginning of this century, few daily precipitation data for Asia were available. Daily precipitation in East Asia (Xie et al. 2007) was analyzed to determine the precipitation change over the Yellow River in China. Rajeevan and Bhate (2009) developed daily grid precipitation data over India using data recorded by more than 2,500 rain gauge stations.

In terms of reproducing Asian monsoon precipitation with climate models, daily to pentad data are needed because intraseasonal oscillation is one of the most dominant precipitation variations in the tropics and regions under monsoon circulations. Thus, it was clear that one of the most important tasks in monitoring and predicting the Asian hydrometeorological environment is to assemble historical daily observation data and develop reliable datasets through the international cooperation of NHMs and international organizations/projects. Hence, we began the APHRODITE project in 2006 to develop state-of-the-art daily precipitation datasets on high-resolution grids covering the whole of Asia.

DATA COLLECTION AND LOCAL POLICIES.

Most countries and/or projects have their own data-sharing policies. Hence, in most cases, access to data had to be negotiated by the APHRODITE project. The data used in APHRODITE analysis were 1) GTS-based data (the global summary of the day), 2) data precompiled by other projects or organizations, and 3) APHRODITE's own collection. We list the names of the databases of the institutes/projects and NHMs who shared their data in Table 1. Figure 1 shows the distribution of our collected data. We provide more detailed information on the data sources on our website (see www.chikyu.ac.jp/precip/) and in an earlier paper on the APHRODITE version 0902 (V0902) product (Yatagai et al. 2009). Because most NHMs prohibit us from releasing their raw data, including information on their stations, we only release gridded (0.25° and 0.5° resolution) products to third parties.

Fig. 1.

The domains and rain gauge distributions used in APHRODITE V1101 for monsoon Asia (MA), the Middle East (ME), and northern Eurasia [i.e., Russia (RU )], and in V1005 for Japan (JP). Stations derived from the GTS network (blue dots), those from the precompiled dataset (black dots), and those in APHRODITE's individual data collection (red dots).

Fig. 1.

The domains and rain gauge distributions used in APHRODITE V1101 for monsoon Asia (MA), the Middle East (ME), and northern Eurasia [i.e., Russia (RU )], and in V1005 for Japan (JP). Stations derived from the GTS network (blue dots), those from the precompiled dataset (black dots), and those in APHRODITE's individual data collection (red dots).

Table 1.

NHMs that provided daily precipitation data (red dots in Fig. 1), precompiled data that we used (black dots in Fig. 1), and GTS data (blue dots in Fig. 1) that we used in APHRO_V1003R1/V1101.

NHMs that provided daily precipitation data (red dots in Fig. 1), precompiled data that we used (black dots in Fig. 1), and GTS data (blue dots in Fig. 1) that we used in APHRO_V1003R1/V1101.
NHMs that provided daily precipitation data (red dots in Fig. 1), precompiled data that we used (black dots in Fig. 1), and GTS data (blue dots in Fig. 1) that we used in APHRO_V1003R1/V1101.
Table 1.

Continued.

Continued.
Continued.

ALGORITHM.

An important scientific aspect to building a rain gauge–based gridded precipitation dataset is the selection and improvement of an interpolation scheme. Figure 2a shows the algorithm flow for our latest version of products, APHRO_PR (precipitation; MA, ME, RU) V1101 and APHRO_JP (V1005). The framework (Fig. 2a) is similar to that of Xie et al. (2007), but we developed different climatology and used a different interpolation scheme.

Fig. 2.

Flowchart of our general algorithm and objective QC. QC steps in which single- (light gray rectangles) and multiple-station (black rectangles) data are used, and internal processing (hatched rectangles) is indicated.

Fig. 2.

Flowchart of our general algorithm and objective QC. QC steps in which single- (light gray rectangles) and multiple-station (black rectangles) data are used, and internal processing (hatched rectangles) is indicated.

QUALITY CONTROL.

The initial conversion of all data to a common digital data format was the most labor-intensive task in producing APHRODITE products, but it allowed us to manually (or subjectively) find data-handling errors, for example, shifted columns and invalid dates (e.g., 31 April). Metadata were incorrect on some occasions, and we needed to check locations on Google Earth. However, since 2007, we have modified the quality control (QC) system so that it is as automated (or objective) as possible, because we handle too many data to check them manually and because we prefer to apply one scheme throughout the domain and the entire period as we describe below. When we produced the V1101 product, the only manual QC procedures that remained were 1) to check initial errors (e.g., invalid dates and shifted columns), 2) to check locations (where there was no location information) and 3) to draw up a black list to eliminate erroneous station data.

Figure 2b is a flowchart of the objective QC system employed when we created V1101. We employed 14 procedures, including two procedures for the QC of metadata [i.e., the location was compared with national boundaries and the elevation was compared with Global 30 Arc-Second Elevation Data Set (GTOPO30) data]. For serial data, QC procedures can be divided roughly into two groups—tests of single-station records (light gray rectangles in Fig. 3b) and tests of multiple-station records (black rectangles). There can be discrepancies between two or more databases containing the same measurements, and it is important to check duplicate data for consistency. However, it is sometimes not possible to judge which data are correct. Additionally, we can also find unit conversion errors (e.g., errors with factors of 10 or 2.54, with the latter relating to conversion between units of millimeters and inches). Details of the 12 QC procedures and their performances were outlined by Hamada et al. (2011). Figure 3 presents a time series of the number of gauges that were passed by our QC system in the creation of the APHRO_V1101 product.

Fig. 3.

Number of gauges that passed our QC. (b)–(d) Three areas correspond to those in Fig. 1. If there is more than one station in a 0.05° cell, then we averaged the precipitation values, and in such cases, the count is 1 in this diagram. The line GTS indicates the station number for the GTS network.

Fig. 3.

Number of gauges that passed our QC. (b)–(d) Three areas correspond to those in Fig. 1. If there is more than one station in a 0.05° cell, then we averaged the precipitation values, and in such cases, the count is 1 in this diagram. The line GTS indicates the station number for the GTS network.

CLIMATOLOGY.

To improve the quantitative estimation of a monthly rain gauge–based precipitation product, the precipitation anomaly or ratio, instead of the total precipitation, is interpolated (Chen et al. 2002; Schaake 2004; Matsuura and Willmott 2012). When employing the Mountain Mapper (MM) method (Schaake 2004), the climatology constructed using a dense network of rain gauges aids interpolation when only relatively sparse data are available. In the case of daily precipitation analysis, Xie et al. (2007) and Yatagai et al. (2008) adopted MM by first defining daily climatology and then interpolating the ratio of daily precipitation to daily climatology. Furthermore, Xie et al. (2007) used the Parameter-elevation Regressions on Independent Slopes Model (PRISM; Daly et al. 1994) for China/Mongolia in addition to rain gauge–based climatology, because MM does not work well for areas where no rain gauge data are available.

The APHRODITE project used the same climatology as Xie et al. (2007) for its earlier product [East Asia version V0804 (EA_V0804)]. From APHRO_PR_V0902 (V0902; Yatagai et al. 2009) onward, we adopted WorldClim (Hijmans et al. 2005), because it provides data throughout the domain.

The steps in producing our daily climatology are as follows:

  1. Daily (except for GTS oriented) data that have undergone QC are summed as monthly values.

  2. The monthly data, including those obtained in step 1, are gathered and the averaged value is calculated if the station has recorded data for more than 5 yr.

  3. The world climatology is prepared at 0.05° resolution [or in the case of APHRO_JP, the Japan climate is prepared at 0.05° resolution (Kamiguchi et al. 2010)].

  4. The ratio of the data obtained in step 2 (station climatology) to the data obtained in step 3 is taken for each month.

  5. The ratio in step 4 is interpolated using Sheremap (Willmott et al. 1985) at a resolution of 0.05°.

  6. The interpolated values in step 5 are multiplied with the world climatology prepared in step 3.

  7. The first six components of the fast Fourier transform of the values obtained in step 6 are taken and the daily climatology is obtained.

SPATIAL INTERPOLATION.

The APHRODITE products interpolate the ratio of daily precipitation to daily climatology. The daily climatology sometimes has a value of 0 mm day−1, such as in desert areas. Additionally, we know that an area that has a local minimum value (spot minimum) in the climatology yields abnormally large precipitation values when there is heavy rainfall near the spot minimum, because we interpolate the ratio. Hence, for these technical reasons, we take the ratio daily, defined as

 
formula

After interpolation, we have

 
formula

We make full use of a Sheremap-type scheme by employing angular distance weighting as follows:

  1. A small weight is given to target cells on the leeward of a high ridge if the ridge is between the target cell and its nearest rain gauge.

  2. A large weight is given to a target cell on a slope that inclines to a rain gauge.

  3. A lookup table is constructed for each month to define the correlation distance, which is obtained from the global 20-km mesh model developed by Meteorological Research Institute (MRI) of the Japan Meteorological Agency (JMA; Mizuta et al. 2006) and used to define the weighting function.

These steps yield better representation of the orographic precipitation pattern as shown in Fig. 4. Without employing MM and the above procedures 1 and 2 (middle panels), heavy rainfall spreads across the Himalayas into the southern part of the Himalayas, and dry areas in the center of Borneo are unclear. Employing the interpolation procedures, we reduce such errors. Additionally, past products only considered the orographic pattern in a climatological sense, whereas we have now applied daily weighting that varies according to the rainfall distribution.

Fig. 4.

Comparison of gridded precipitation for different weighting schemes employed in APHRODITE. (a) Geography (elevation in m) over the Himalayas around Nepal, (b) annual precipitation (mm day−1) over the Himalayas interpolated by the Sheremap (Willmott et al. 1985) without using MM. (c) The same as (b), but shows the result of using a weighting scheme that considers the presence of a ridge between a rain gauge and a target cell and the presence of inclined slopes, and adopts MM. (d) The same as (a), but on Borneo, Indonesia. (e) and (f) are the same as (b) and (c), but show monthly precipitation (mm day−1) in Jun 1998 on Borneo.

Fig. 4.

Comparison of gridded precipitation for different weighting schemes employed in APHRODITE. (a) Geography (elevation in m) over the Himalayas around Nepal, (b) annual precipitation (mm day−1) over the Himalayas interpolated by the Sheremap (Willmott et al. 1985) without using MM. (c) The same as (b), but shows the result of using a weighting scheme that considers the presence of a ridge between a rain gauge and a target cell and the presence of inclined slopes, and adopts MM. (d) The same as (a), but on Borneo, Indonesia. (e) and (f) are the same as (b) and (c), but show monthly precipitation (mm day−1) in Jun 1998 on Borneo.

We carefully interpolated regions with a very dense network of rain gauges where available. When there are data for more than one station in a 0.05° cell, we average the data and set the location of the station as the middle of the cell (e.g., 30.025°N, 120.025°E). This is done because 1) sometimes a network that is too dense relative to the target cell size makes the analysis unstable and results in erroneous values, 2) the analysis of extreme weather events prefers that the station data are preserved to the target cell, and 3) sometimes data in duplicate data sources are not easy to identify (i.e., they are difficult to judge with our current automated QC).

Considering the above processes, we include ancillary information of the ratio of valid 0.05° cells with rain gauge(s) to show the reliability of data for each 0.25° and 0.5° resolution cell. In an earlier product (V0804) and other widely used products, information on the number of gauges is usually given. However, it is difficult and sometimes nonsensical to attribute the data of all gauges in a 0.25° and 0.5° resolution cell when there is more than one gauge in a 0.05° cell. Hence, in this case, we used a factor of 1/25 (= 4%) for the 0.25° product and 1/100 (= 1%) for the 0.5° product. As we show in the ratio of the station grids (RSTN) diagrams in Fig. 5, for example, 4 (%) means there is one 0.05° grid cell with at least one rain gauge among 25 grid cells to define the 0.25° grid value.

Fig. 5.

An example of daily precipitation analysis (mm day−1) for 21 Jul 1998. (left) APHRODITE's latest product (V1101) and (right) GTS data. (bottom) RSTN in the 0.25° analysis. For example, 4% means there is one 0.05° grid cell with at least one rain gauge among 25 grid cells to define the 0.25° grid value. An altitude of 3,000 m is shown.

Fig. 5.

An example of daily precipitation analysis (mm day−1) for 21 Jul 1998. (left) APHRODITE's latest product (V1101) and (right) GTS data. (bottom) RSTN in the 0.25° analysis. For example, 4% means there is one 0.05° grid cell with at least one rain gauge among 25 grid cells to define the 0.25° grid value. An altitude of 3,000 m is shown.

DIFFERENCES IN MONSOON PRECIPITATION ACCORDING TO PRODUCT.

We now look at how our understanding of monsoon precipitation has changed after our efforts in collecting, converting, and checking data. Figure 5 compares the results obtained by V1101 and our test analysis (GTS analysis) that only includes GTS-oriented data and does not employ MM. The heavy rainfall that fell along the Himalayas and the narrow precipitation zone along the Western Ghats in India is well represented by the use of the denser gauge network and improved algorithm. V1101 represents the rainfall pattern over South and Southeast Asia more precisely. The southern foothills of the Himalayas are well known for their heavy monsoon precipitation in summer. Therefore, without input data for Nepal, Bhutan, and northern India, precipitation was underestimated. Next, we show the total difference due to the use of additional rain gauge data.

Figure 6 shows the total, averaged difference between our APHRODITE product (APHRO_V1003R1) and the GTS analysis. We did not apply MM and we did not carry out our QC in the GTS analysis.

Fig. 6.

Mean annual precipitation (mm yr−1) for 1990–2007 (18 yr). (a) 0.25° APHRO_V1003R1, and (b) GTS data analysis employing an interpolation method similar to that used for (a). (c) Difference between (b) and (a).

Fig. 6.

Mean annual precipitation (mm yr−1) for 1990–2007 (18 yr). (a) 0.25° APHRO_V1003R1, and (b) GTS data analysis employing an interpolation method similar to that used for (a). (c) Difference between (b) and (a).

Large differences between products can be seen along the Himalayas, in Southeast Asia, around Iran, and in western Siberia. In particular, GTS analysis underestimates precipitation by 2,000 mm yr−1 over the Himalayas and Indonesia. Compared with the standard estimations of APHRODITE, most other products, which rely on the GTS network, underestimate Himalayan precipitation by about one-third and underestimate precipitation over most of Indonesia and part of the Indochina Peninsula by up to one-half. It is considered that GTS data are used in most available global datasets. Therefore, if a satellite-merged analysis only uses GTS data and adjusts the estimate according to GTS gauge data, then there should be bias similar to that seen in Fig. 6.

Figure 7 compares summer monsoon rainfall (June–September) in 1998 over and around the Himalayas. It is clear that GTS-only analysis (Fig. 7c) underestimates precipitation, and, thus, the area-averaged precipitation time series of GTS analysis (Fig. 7d) does not represent the intraseasonal (20–60 days) precipitation variation that is observed in APHRO_V1101 (Fig. 7d, red line). The APHRODITE product is useful for understanding and validating tropical monsoon precipitation because the rainfall system for this season has either submonthly or intraseasonal oscillations.

Fig. 7.

(a) – (c) Jun–Sep precipitation around the Himalayas [mm (4 months) −1]. (a) APHRO_V1101, (b) GPCC full version 4, and (c) simple interpolation result of GTS data obtained by Shepard (1968). (d) Areal average (red; 26°–28°N, 82°–92°E) of APHRO_V1101 daily precipitation and the interpolation of GTS data (black). The domains for the calculation of the areal mean precipitation in (d) are shown in (a), (c) (rectangles). An altitude of 3,000 m is represented (black line).

Fig. 7.

(a) – (c) Jun–Sep precipitation around the Himalayas [mm (4 months) −1]. (a) APHRO_V1101, (b) GPCC full version 4, and (c) simple interpolation result of GTS data obtained by Shepard (1968). (d) Areal average (red; 26°–28°N, 82°–92°E) of APHRO_V1101 daily precipitation and the interpolation of GTS data (black). The domains for the calculation of the areal mean precipitation in (d) are shown in (a), (c) (rectangles). An altitude of 3,000 m is represented (black line).

Because no published products of gauge-based analyses of daily precipitation over the Himalayas are available, we compare the 4-month total with that of the Global Precipitation Climatology Centre (GPCC) full archive product version 4 (Schneider et al. 2008) in Fig. 7b and hereafter. Because the GPCC dataset includes almost the same number of monthly precipitation data over Nepal, Bhutan, and Bangladesh, the pattern (Fig. 7b) is similar to that of our V1101 product (Fig. 7a). The rainband along the Himalayas is narrower in our analysis. In the GPCC dataset, heavy rainfall along the Himalayas penetrates north of the main ridge, which is reduced in our analysis by the interpolation scheme described above (Fig. 4). In Arunachal Pradesh, India (around 28°N, 94°E), to the north of Myanmar, the GPCC estimate is higher than the estimate of our analysis. Because both GPCC version 4 and the APHRODITE analysis adopt the MM method, and because there are few rain gauge stations, the difference seems to relate to the difference in climatology. Our climatology seems to be based on more station data than that of the GPCC, and we incorporate a pattern of world climate that has a resolution of 0.05°, while the GPCC climate normally has resolution of 0.25°. A composite of radar measurements carried out by the Tropical Rainfall Measuring Mission (TRMM) has a pattern more similar to our climatology than the GPCC climatology (data not shown; Yatagai and Kawamoto 2008).

COMPARISON WITH GPCC AND OTHER DATASETS.

The purpose of this section is to provide potential users with information on the quantitative differences between our analysis and a monthly rain gauge–based product that is constructed without GTS-based real-time data. There are four sets of widely used gauge-based monthly precipitation datasets, namely, the GPCC, the University of East Anglia's Climate Research Unit (CRU; Mitchell and Jones 2005), the University of Delaware (UDE; Willmott and Matsuura 1995), and Precipitation Reconstruction over Land (PREC/L) datasets (Chen et al. 2002), and there are no large differences in estimates among the monthly rain gauge–based products over Asia (Yatagai et al. 2005; Xie et al. 2007). Here we mainly present the differences between and statistics for APHRO_PR with the GPCC product at a resolution of 0.5°, because GPCC, as part of the GPCP, collects the largest number of station data worldwide under a WMO mandate (Schneider et al. 2008).

Figure 8 compares the APHRODITE product with the GPCC product for January and July 1998. The overall magnitudes and patterns derived from our analysis (V1101) are consistent with those of the GPCC analysis. Except for Europe and at least for the year 1998, we used more gauge data (Fig. 8, top middle and center panels). The difference (APHRO– GPCC) for January shows that our estimate is less than that of GPCC over western Russia and the Middle East. Because we incorporate data for Saudi Arabia in V1101, our estimate is more reliable and avoids false penetration of the wet zone's precipitation to arid zones. In both January and July, our estimate is less than that of GPCC in large areas of Indonesia. In July 1998, the difference pattern is complicated.

Fig. 8.

Comparison of monthly precipitation in APHRODITE and GPCC products. (a) Precipitation (mm month−1) for Jan 1998 according to APHRO_V1101 (0.5°), (b) rain gauge density according to APHRO_V1101 for 1998, (c) the same as (a), but for Jul 1998. (d)–(f) The same as (a)–(c), but for GPCC version 4. (g) The difference between APHRO and GPCC for Jan 1998. (h) Precipitation (mm yr−1) for 1998 and three areas for the calculation of statistics in Fig. 11. (i) The same with (g), but for Jul 1998.

Fig. 8.

Comparison of monthly precipitation in APHRODITE and GPCC products. (a) Precipitation (mm month−1) for Jan 1998 according to APHRO_V1101 (0.5°), (b) rain gauge density according to APHRO_V1101 for 1998, (c) the same as (a), but for Jul 1998. (d)–(f) The same as (a)–(c), but for GPCC version 4. (g) The difference between APHRO and GPCC for Jan 1998. (h) Precipitation (mm yr−1) for 1998 and three areas for the calculation of statistics in Fig. 11. (i) The same with (g), but for Jul 1998.

After taking the difference (APHRO – GPCC) for each month (January 1951–December 2007), the averaged bias is presented in Fig. 9a. In most areas, APHROV1101 estimates less precipitation than the GPCC product. In Indonesia, the number of input stations is almost the same (we had more station data for Java, cf. Fig. 1). Hence, the difference seems to be because of 1) QC and 2) different interpolation methods. The root-mean-square error (Fig. 9b) also shows a large difference between APHRO_V1101 and GPCC products over Indonesia, coastal Indochina and the Western Ghats, and around the Himalayas. The spatial patterns of temporal correlation (Fig. 9c) have generally high correlation (more than ~0.8), but there is less correlation for dry areas (the northern part of the Arabian Peninsula and the western part of the Tibetan Plateau).

Fig. 9.

Difference between the monthly product of APHRO_V1101 and the monthly product of GPCC full version 4. (a) Average of monthly bias (1951–2007, mm month−1), (b) root-mean-square error (mm month−1), and (c) serial (1951–2007) correlation between APHRO_V1101 and GPCC products.

Fig. 9.

Difference between the monthly product of APHRO_V1101 and the monthly product of GPCC full version 4. (a) Average of monthly bias (1951–2007, mm month−1), (b) root-mean-square error (mm month−1), and (c) serial (1951–2007) correlation between APHRO_V1101 and GPCC products.

Figure 10 shows the time series of the statistical comparison between APHRO_V1101 and the GPCC product. Statistics are derived from 0.5° grid pairs where both APHRO and GPCC have valid data (shown in Fig. 10b). We inserted a missing value where no rain gauge data are available nearby, and therefore, the number of valid grid cells in our product in the early 1950s is less than in the later period. Because of the summer monsoon, the average precipitation of the total area is a maximum in summer. The two time series (Fig. 10a) vary coincidently, but our estimate is less than that of the GPCC. According to the ratio of APHRO_V1101 to the GPCC product (Fig. 10c), our estimate is about 75%–90% of that of the GPCC estimate, but after 2003, the ratio decreases, indicating a larger difference. The same signal is observed in the bias graph (Fig. 10d). The ratio is less from late 1998 to 1999 than in the preceding and subsequent periods. The correlation coefficient and root-mean-square error have stable time series, except for the case in September 2004, which was attributed to abnormal values (~10,000 mm month−1) in Indonesia in the GPCC product.

Fig. 10.

Time series of the spatial relationship between APHRO_V1101 and GPCC products over Eurasia (plot area of Figs. 8 and 9). Statistics are derived from 0.5° grid pairs where (b) both APHRO and GPCC have valid data. (a) Time series of the areal averaged precipitation for APHRO (black) and GPCC (gray). (c) The ratio of APHRO/GPCC estimates is given in the second panel. The time series of the (d) special pattern correlation coefficient, (e) spatial average of the difference (APHRO – GPCC ), and (f) root-mean-square error.

Fig. 10.

Time series of the spatial relationship between APHRO_V1101 and GPCC products over Eurasia (plot area of Figs. 8 and 9). Statistics are derived from 0.5° grid pairs where (b) both APHRO and GPCC have valid data. (a) Time series of the areal averaged precipitation for APHRO (black) and GPCC (gray). (c) The ratio of APHRO/GPCC estimates is given in the second panel. The time series of the (d) special pattern correlation coefficient, (e) spatial average of the difference (APHRO – GPCC ), and (f) root-mean-square error.

The cause of the difference between the products differs in region and in time. Therefore, we show the time series and number of input data for the three domains—China, India, and central Asia—in Fig. 11. As shown in Fig. 9, very high correlations (~0.9) are observed in central China and India. In China, the input number of data is almost the same, and the difference in average precipitation is very small. The ratio (APHRO/GPCC) is always between 0.9 and 1.1 (data not shown). In India, except for the year 1998/99, the number of input data are the same and the ratio (APHRO/GPCC) is between 0.8 and 1.0 (data not shown). In spite of the large difference in the number of input data in the 2 yr, the ratio (APHRO/GPCC) does not differ from that for the other years.

Fig. 11.

Time series of the areal average of the monthly precipitation (mm month−1) and areal average of the number of gauges for APHRO_V1101 (0.5°, red) and GPCC version 4 (black). The three domains (a) China, (b) India, and (c) central Asia are illustrated in the middle center panel of Fig. 8. Because each box (20° × 20°) contains 1,600 0.5° cells, values of each bottom panel × 1,600 mean number of gauges are included in each box.

Fig. 11.

Time series of the areal average of the monthly precipitation (mm month−1) and areal average of the number of gauges for APHRO_V1101 (0.5°, red) and GPCC version 4 (black). The three domains (a) China, (b) India, and (c) central Asia are illustrated in the middle center panel of Fig. 8. Because each box (20° × 20°) contains 1,600 0.5° cells, values of each bottom panel × 1,600 mean number of gauges are included in each box.

In the case of central Asia, we used more data than did the GPCC, except for the year 2007. The ratio is around 0.8–0.9, but it falls to 0.3 and the difference increases after 2005. The distribution of input data must be considered when we use recent data for central Asia.

Figure 12 compares various precipitation products over monsoon Asia. The figure shows the area-weighted average of June–August (JJA) mean terrestrial precipitation over the region of 5°–35°N, 65°–130°E (the area shown in Fig.5). The error bars indicate the standard deviation for interannual variability of JJA mean terrestrial precipitation. As gauge-based products, we compare our product with the CPC unified product (Chen et al. 2008) and four above-mentioned monthly gauge-based products. Additionally, the average value of publically available satellite-based products and merged analyses are shown. The products are TRMM PR3A25 (version 6; Iguchi et al. 2000), Global Satellite Mapping of Precipitation (GSMaP) microwave radiometer (MWR; Aonashi et al. 2009), GSMaP MVK (Ushio et al. 2009), CPC morphing technique (CMORPH; Joyce et al. 2004), Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN; Sorooshian et al. 2000), Naval Research Laboratory (NRL)-blended (Turk and Miller 2005), TRMM 3B42V6 (Huffman et al. 2007), GPCP one-degree-daily (1DD; version 1.1; Huffman and Bolvin 2009), CPC Merged Analysis of Precipitation (CMAP; Xie and Arkin 1997), and GPCP (version 2.1; Huffman et al. 2009). Further descriptions and analyses according to altitude were provided by Arakawa and Kitoh (2011). The area-averaged mean values of JJA terrestrial precipitation in the satellite-based datasets are smaller than those in the rain gauge–based datasets, but there is greater scatter among the former than among the latter. APHRO_V1003 provides values lower than those of the four monthly gauge-based products, but higher than those of the CPC unified analysis, which uses only GTS data outside of China. A possible cause of our underestimation compared with the monthly precipitation dataset is insufficient QC, especially for the GTS-based data. Because the GTS data are real-time data, a missing value is sometimes reported as 0 mm, resulting in underestimation by the gauge analysis. Meanwhile, as we have seen in the comparison with GPCC products, our sophisticated algorithm reduced the pattern of false penetration seen in all four gauge-based analyses (Xie et al. 2007) and gave narrower rainbands along the west coast of India and Indochina (Fig. 12). Hence, the real value seems to be between those given by the APHRO_V1003 (or V1101) and monthly 0.5° products. TRMM 3B42 and CMAP may give reliable quantitative estimates, but further analysis is required to confirm this. Additionally, satellite products have been validated against APHRODITE products. For such a purpose, users are advised to check the “rstn ” and use only cells with rain gauges.

Fig. 12.

Mean precipitation observed by gauge and satellite over the region of 5°–35°N, 65°–130°E (the same area as in Fig. 5) in JJA. Error bars show the standard deviation of the interannual variability (Arakawa and Kitoh 2011).

Fig. 12.

Mean precipitation observed by gauge and satellite over the region of 5°–35°N, 65°–130°E (the same area as in Fig. 5) in JJA. Error bars show the standard deviation of the interannual variability (Arakawa and Kitoh 2011).

DATA RELEASE AND FURTHER DEVELOPMENTS.

We released our gridded (0.25° and 0.5° resolution) precipitation data for free for scientific purposes. There were more than 1,275 registered users as of the end of December 2011. We have already released V0804, V0902, V1003R1, and V1101 for Asia [monsoon Asia, Middle East, and northern Eurasia (Russia)] at 0.25° and 0.5° resolution and V1005 for Japan at 0.05° resolution for a period of 100 yr (Kamiguchi et al. 2010; APHRO_JP). Currently we do not release raw data, such as station information or intermediate products with 0.05° resolution, except for APHRO_JP, because most NHMs prohibit us from passing the original data on to third parties.

The time coverage of the latest version, V1101, is the same as that of V1003R1 (1951–2007). The major differences between the products are 1) improved QC and 2) the new inclusion of data for Saudi Arabia and Belarus and data from the European Climate Assessment and Dataset, and some updated inputs for Bhutan, Indonesia, South Korea, Taiwan, and Thailand. Because the Middle East analysis was improved by adding the data of Saudi Arabia (Table 1), we enlarged the Middle East domain V1003R1 to V1101 to cover the whole of Saudi Arabia. In addition, we released a merged areal product APHRO_PR, which covers the APHRODITE monsoon Asia, Middle East, and northern Eurasia (Russia) domains, because it was expected by users who run hydrological models over midlatitudes and over northern Eurasia.

It is important for runoff studies and climate change assessments at high altitudes and high latitudes to discriminate between snowfall and rainfall. We have prepared daily temperature analysis and added snow/rain judgment to our precipitation product (Yasutomi et al. 2011). We will release these new products in the near future.

There are two other large issues remaining: the ending time and interannual variation. The ending time for 24-h precipitation accumulation differs from country to country. Most countries measure rainfall in the morning and record it on the observed day. For example, India measures rainfall at 0830 LT (i.e., 0300 UTC) as 24-h precipitation from 0300 UTC on the previous day to 0300 UTC on the recording day. China and Mongolia measure precipitation at 0800 and 2000 LT, and daily precipitation is that accumulated from 1200 UTC on the previous day to 1200 UTC on the recording day. In Japan, the officially released daily precipitation is the 24-h accumulation from 1500 UTC on the previous day to 1500 UTC on the recording day. In addition, we know that the ending times of NHM data and GTS-oriented data sometimes differ. We could discard GTS-oriented data if NHM data are available for the same station. However, it is usually difficult to identify the station that recorded the data in an objective manner, as we described in the interpolation section. In such cases, the temporal representative is not preserved, but the total quantity of the daily analysis is preserved.

It is possible to convert the original daily reports into daily values with the same reporting time using satellite information, as we did when using APHRODITE data to improve forecasts (e.g., Krishnamurti et al. 2009). However, because our objective has been to apply the same scheme (interpolation, adjustment, and QC) to the entire target domain and to the whole period, we did not apply a time adjustment for the released products.

The number of gauges and the configuration of the gauge network change over many years, and the number of data that we obtained vary in region and time. This is a crucial consideration when using the APHRODITE data for studies on interannual variations. A possible solution is to make a product with stable (consistent) input data, such as the GPCC Variability Analyses of Surface Climate Observations (VASClimO) and East Asia analysis of Xie et al. (2007). However, since our first goal was to produce a quantitative high-resolution product for the validation of high-resolution models and for studies on hydrological applications, such the downscaling of GCM to be used to drive a river runoff model, it has been more important for us to improve the dataset even over the past several years. Hence, as described above, users are encouraged to check the change in RSTN, as we showed Fig. 11.

Kamiguchi et al. (2010) studie effects of changes to a gauge network on long-term trends, including extreme values, and found a change in gauge density not affect the trend of total rainfall but does affect the trend extreme values in Japan. Similar works can be carried out with the APHRODITE scheme, which we to do in the near future.

SUMMARY.

The activities of the APHRODITE project, including outreach programs, have had many important benefits. Investigations of future changes monsoon and/or extreme events and assessments historical/future changes in hydrological flow rely accurate long-term daily gridded observation data precipitation. The APHRODITE project has made data available and has substantially improved precipitation estimates throughout Asia.

Acknowledgments

APHRODITE's water resources project was supported by the Environment Research & Technology Development Fund (GERF-FS051, B062, and A0601) by the Ministry of the Environment, Japan. Part of this work was supported by Grants-in- for Scientific Research (10540467) from JSPS, and a research project on the “Human Life, Aging, and Disease High-Altitude Environments (D-03)” administered by Research Institute for Humanity and Nature (RIHN). thank Dr. Pingping Xie, Professor Pinhas Alpert, and Professor Phil Arkin for their valuable advice.

REFERENCES

REFERENCES
Aonashi
,
K.
,
and Coauthors
,
2009
:
GSMaP passive microwave precipitation retrieval algorithm: Algorithm description and validation
.
J. Meteor. Soc. Japan
,
87A
,
119
136
.
Arakawa
,
O.
, and
A.
Kitoh
,
2011
:
Intercomparison of the relationship between precipitation and elevation among gridded precipitation datasets over the Asian summer monsoon region
.
Global Environ. Res.
,
15
,
109
118
.
Chen
,
M.
,
P.-P.
Xie
,
J. E.
Janowiak
, and
P. A.
Arkin
,
2002
:
Global land precipitation: A 50-yr monthly analysis based on gauge observations
.
J. Hydrometeor.
,
3
,
249
266
.
Chen
,
M.
,
W.
Shi
,
P.
Xie
,
V. B. S.
Silva
,
V. E.
Kousky
,
R. W.
Higgins
, and
J. E.
Janowiak
,
2008
:
Assessing objective techniques for gauge-based analyses of global daily precipitation
.
J. Geophys. Res.
,
113
,
D04110
,
doi:10.1029/2007JD009132
.
Daly
,
C.
,
R. P.
Neilson
, and
D. L.
Phillips
,
1994
:
A statistical-topographic model for mapping climatological precipitation over mountainous terrain
.
J. Appl. Meteor.
,
33
,
140
158
.
Hamada
,
A.
,
O.
Arakawa
, and
A.
Yatagai
,
2011
:
An automated quality control method for daily rain-gauge data
.
Global Environ. Res.
,
15
,
165
172
.
Hijimans
,
R. J.
,
S. E.
Cameron
,
J. L.
Parra
,
P. G.
Jones
, and
A.
Jarvis
,
2005
:
Very high resolution interpolated climate surfaces for global land areas
.
Int. J. Climatol.
,
25
,
1965
1978
.
Huffman
,
G. J.
, and
D. T.
Bolvin
,
2009
:
GPCP one-degree daily precipitation data set documentation
.
NASA GSFC
,
25
Huffman
,
G. J.
,
and Coauthors
,
2007
:
The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scale
.
J. Hydrometeor.
,
8
,
38
55
.
Huffman
,
G. J.
,
Adler
,
D. T.
Bolvin
,
G.
Gu
,
2009
:
Improving the global precipitat ion record: GPCP Version 2.1
.
Geophys. Res. Lett.
,
36
,
L17808
,
doi:10.1029/2009GL040000
.
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
.
Joyce
,
R. J.
,
J. E.
Janowiak
,
P. A.
Arkin
, and
P.
Xie
,
2004
:
CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution
.
J. Hydrometeor.
,
5
,
487
503
.
Kamiguchi
,
K.
,
O.
Arakawa
,
A.
Kitoh
,
A.
Yatagai
,
A.
Hamada
, and
N.
Yasutomi
,
2010
:
Development of APHRO_JP, the first Japanese high-resolution daily precipitation product for more than 100 years
.
Hydrol. Res. Lett.
,
4
,
60
64
.
Krishnamurti
,
T. N.
,
A. K.
Mishra
,
A.
Simon
, and
A.
Yatagai
,
2009
:
Use of a dense gauge network over India for improving blended TRMM products and downscaled weather models
.
J. Meteor. Soc. Japan
,
87
,
395
416
.
Matsuura
,
K.
, and
C. J.
Willmott
,
cited 2011
:
Terrestrial precipitation: 1900–2008 gridded monthly time series (version 2.01)
.
Mitchell
,
T. D.
, and
P. D.
Jones
,
2005
:
An improved method of constructing a database of monthly climate observations and associated high-resolution grids
.
Int. J. Climatol.
,
25
,
693
712
.
Mizuta
,
R.
,
and Coauthors
,
2006
:
20-km-mesh global climate simulations using JMA-GSM model: Mean climate states
.
J. Meteor. Soc. Japan
,
84
,
165
185
.
Rajeevan
,
M.
, and
J.
Bhate
,
2009
:
A high resolution daily gridded rainfall dataset (1971–2005) for meso-scale meteorological studies
.
Curr. Sci.
,
96
,
558
562
.
Schaake
,
J.
,
2004
:
Application of prism climatologies for hydrologic modeling and forecasting in the western U.S
.
Proc. 18th Conf. on Hydrology
,
Seattle, WA
,
Amer. Meteor. Soc.
,
5.3
.
Schneider
,
U.
,
T.
Fuchs
,
A.
Meyer-Christoffer
, and
B.
Rudolf
,
2008
:
Global precipitation analysis products of the GPCC
.
Global Precipitation Climatology Centre, DWD
,
13
Shepard
,
D.
,
1968
:
A two-dimensional interpolation function for irregularly-spaced data
.
Proc. 23rd ACM national conference
,
Association for Computing Machinery
,
517
524
.
Sorooshian
,
S.
,
K.
Hsu
,
X.
Gao
,
H. V.
Gupta
,
B.
Imam
, and
D.
Braithwaite
,
2000
:
Evaluation of PERSIANN system satellite-based estimates of tropical rainfall
.
Bull. Amer. Meteor. Soc.
,
81
,
2035
2046
.
Turk
,
F. J.
, and
S. D.
Miller
,
2005
:
Toward improving estimates of remotely sensed precipitation with MODIS/AMSR-E blended data techniques
.
IEEE Trans. Geosci. Remote Sens.
,
43
,
1059
1069
.
Ushio
,
K.
,
and Coauthors
,
2009
:
A Kalman filter approach to the Global Satellite Mapping of Precipitation (GSMaP) from combined passive microwave and infrared radiometric data
.
J. Meteor. Soc. Japan
,
87A
,
137
151
.
Willmott
,
C. J.
, and
K.
Matsuura
,
1995
:
Smart interpolation of annually averaged air temperature in the United States
.
J. Appl. Meteor.
,
34
,
2577
2586
.
Willmott
,
C. J.
,
C. M.
Rowe
, and
W. D.
Philpot
,
1985
:
Small scale climate maps: A sensitivity analysis of some common assumptions associated with grid-point interpolation and contouring
.
Amer. Cartographer
,
12
,
5
16
.
Xie
,
P.
, and
P. A.
Arkin
,
1997
:
Global precipitation: A 17- year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs
.
Bull. Amer. Meteor. Soc.
,
78
,
2539
2558
.
Xie
,
P.
,
A.
Yatagai
,
M.
Chen
,
T.
Hayasaka
,
Y.
Fukushima
,
C.
Liu
, and
S.
Yang
,
2007
:
A gauge-based analysis of daily precipitation over East Asia
.
J. Hydrometeor.
,
8
,
607
627
.
Yasutomi
,
N.
,
A.
Hamada
, and
A.
Yatagai
,
2011
:
Development of a long-term daily gridded temperature dataset and its application to rain/snow discrimination of daily precipitation
.
Global Environ. Res.
,
15
,
165
172
.
Yatagai
,
A.
, and
H.
Kawamoto
,
2008
:
Quantitative estimation of orographic precipitation over the Himalayas by using TRMM/PR and a dense network of rain gauges
.
Vertical Heating Profiles from Satellites and Models
,
T. N.
Krishnamurti
,
B. N.
Goswami
, and
Akiyo
Yatagai
,
Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 7148, 71480C,), doi:10.1117/12.811943
.
Yatagai
,
A.
,
P.
Xie
, and
A.
Kitoh
,
2005
:
Utilization of a new gauge-based daily precipitation dataset over monsoon Asia for validation of the daily precipitation climatology simulated by the MRI/JMA 20-km-mesh AGCM
.
SOLA
,
1
,
193
196
,
doi:10.2151/sola.2005-050
.
Yatagai
,
A.
,
P.
Xie
, and
P.
Alpert
,
2008
:
Development of a daily gridded precipitation data set for the Middle East
.
Adv. Geosci.
,
12
,
165
170
.
Yatagai
,
A.
,
O.
Arakawa
,
K.
Kamiguchi
,
H.
Kawamoto
,
M. I.
Nodzu
, and
A.
Hamada
,
2009
:
A 44-year daily precipitation dataset for Asia based on dense network of rain gauges
.
SOLA
,
5
,
137
140
,
doi:10.2151/sola.2009-035
.

Footnotes

*CURRENT AFFILIATION: Research Institute for SustainableHumanosphere, Kyoto University, Kyoto, Japan

APHRODITE: Asian Precipitation—Highly Resolved Observational Data Integration Towards Evaluation of Water Resources