Comparing high-resolution gridded precipitation data with satellite rainfall estimates of TRMM 3 B 42 over Iran

To evaluate satellite rainfall estimates of Tropical Rain Measurement Mission (TRMM) level 3 output (3B42) (TRMM 3B42) over Iran (20 –45 N, 40–65 E), we compared these data with high-resolution gridded precipitation datasets (0.25 ×0.25 latitude/longitude) based on rain gauges (Iran Synoptic gauges Version 0902 (IS0902)). Spatial distribution of mean annual and mean seasonal rainfall in both IS0902 and TRMM3B42 from 1998 to 2006 shows two main rainfall patterns along the Caspian Sea and over the Zagros Mountains. Scatter plots of annual average rainfall from IS0902 versus TRMM3B42 for each 0.25×0.25 grid cell over the entire country (25 –40 N, 45–60 E), along the Caspian Sea (35 –40 N, 48–56 E), and over the Zagros Mountains (28 –37 N, 46–55 E) were derived. For the entire country, the Caspian Sea region, and the Zagros Mountains, TRMM3B42 underestimates mean annual precipitation by 0.17, 0.39, and 0.15 mm day −1, respectively, and the mean annual rainfall spatial correlation coefficients are 0.77, 0.57, and 0.75, respectively. The mean annual precipitation temporal correlation coefficient for IS0902 and TRMM3B42 is∼0.8 in the area along the Zagros Mountains, and ∼0.6 in the Caspian Sea and desert regions.


Introduction
Global precipitation is a major component of the global water and energy cycle that influences significantly Earth's climate system, and in is turn affected by climate system variability and change.The Radiation Panel of the Global En-Correspondence to: S. Javanmard (sohailajavanmard@gmail.com) ergy and Water Cycle experiment (GEWEX) of the World Climate Research Programme (WCRP) was formed to improve understanding of seasonal and inter-annual and longerterm variability of the global hydrological cycle, to determine the atmospheric latent heat rates needed for weather and climate prediction models, and to provide an observational precipitation data set for model validation, initialization and other hydrological applications (Gruber and Levizzani, 2008).Many attempts have been conducted to develop and improve global and regional gridded precipitation gauge data sets in recent years (e.g.New et al., 2000;Mitchell and Jones, 2005;Schneider, 2008).However, due to the lack of observational precipitation datasets over land areas for most of Asia, Yatagai et al. (2008Yatagai et al. ( , 2009) ) have created a highresolution rain gauge-based daily precipitation grid dataset for East, Middle East, and Russia.To achieve more accurate gridded precipitation data over Iran, Javanmard et al. (2008) presented the improvement of gridded precipitation data over Iran using highly quality-controlled data from ∼200 synoptic stations of the Islamic Republic of Iran Meteorological Organization (IRIMO).
Although rain gauge observations provide longer records (Yatagai et al., 2009) they do not provide a reliable spatial representation of precipitation (Gruber and Levizzani, 2008).For example, the distribution of gauge networks is sparse in desert and mountainous areas in Iran.Remote sensing techniques, such as those using radar or satellites, are useful for monitoring rainfall over large mountainous and oceanic regions, and their time series accuracy and precision being potentially locally be lower than existing and corresponding data sets derived from ground-based measurements, they provide more homogeneous data quality compared ground observations (Schulz et al., 2009)     the performance of five satellite-based high-resolution precipitation outputs by comparison against gauge analysis: TRMM (3B42, 3B42RT ) (Huffman et al., 2004;Huffman et al., 2007), National Weather Service, USA (NWS), Climate Prediction Center (CPC), Morphing Techniques (CMORPH; Joyce et al., 2004), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PER-SIANN; Sorooshian et al. 2000), Naval Research Laboratory (NRL; Turk et al., 2004).They found that all satellite data performed better in depicting precipitation for wet climate regions and wet seasons, but were limited skills in estimating precipitation over central Asian arid and semi-arid regions.The performance of satellite-based precipitation has not been verified over arid and semi-arid areas in Iran.To address this, in this paper we compare rainfall estimates of TRMM 3B42 with gauge-based high-resolution gridded precipitation data over Iran from 1998 to 2006 to evaluate and verify TRMM-based precipitation outputs.from northwest to southeast and reach a maximum altitude of ∼3500 m (Fig. 1).These two ranges play an influential role in determining the amount and distribution of rainfall over the country.The dominant flow directions of air masses are from the west and from the south.Several air masses influence Iran throughout the year.The major ones are the Mediterranean low-pressure system which flows from the west throughout the year, the Siberian high-pressure system (Siberian High) which flows into the country from the north mostly during the winter, and the Sudan tropical low pressure (Sudan Low) which flows from the southwest (Golestani, 2000).Iran is classified generally as arid or semiarid with mean annual rainfall ∼250 mm for the entire country, ranging from 50 mm in the deserts to 1600 mm on the Caspian coast (Fig. 2).
The rainy period for most of the country is from November to May, during which the average rainfall is ∼240 mm.Maximum precipitation falls on the Alborz and Zagros slopes, facing north and west, respectively, where the mean annual rainfall is >1200 mm.Farther inland, the precipitation decreases to <100 or <50 mm annually.Precipitation varies considerably with topography.In the northern and western mountains the annual mean precipitation is >480 mm.The plateau receives most of the rainfall in spring, while on the western and southern coasts most rain falls in winter.On the Caspian coast where the rain falls earlier, maximum rainfall is in autumn.In the dry period between May and October, rain is rare in most of the country.Ninety percent of total precipitation occurs during the cold and humid seasons in the northern and western parts of the country, and only 10% occurs during the warm and dry seasons and in the central, southern and eastern regions.Fifty-two percent of precipitation falls on 25% of the country's land area; resulting in a lack of water resources and potential water crises in the near future.No. of Synoptic Gauges

Synoptic gauge-network data of the Islamic Republic of Iran Meteorological Organization
The surface-based data used in this study are synoptic gauge data of the Islamic Republic of Iran Meteorological Organization (IRIMO).Figure 3 shows time series of the number of synoptic gauges over Iran.The number of synoptic stations used in this study varies from a maximum of 188 in 2004 to a minimum of 10 in 1951.There is a yearly increase, and the mean number of gauges for the analysis period is 130 (Fig. 3).The synoptic gauge data have passed IRIMO quality control procedures such as checking location (latitude, longitude and elevation), consistency with other meteorological parameters, tests for data homogeneity, filling data gaps.Figure 3 also shows a significant increase in number of synoptic gauges from 1985 onwards. Figure 4 shows the spatial distribution of the synoptic gauge stations over Iran, with mean station numbers in 0.25-degree-grid cells averaged for nine years from 1998 to 2006.The distribution of synoptic gauges is denser in along the Zagros Mountains, the Caspian coast and in the north-western regions than in the desert regions.

TRMM and other satellite data (TRMM 3B42)
The No. of Synoptic Gauges

Algorithm of daily gauge-based gridded precipitation data over Iran
The procedure for deriving gridded precipitation data from daily gauges is based on Yatagai et al. (2008), and is as follows: 1. Define monthly precipitation for stations where data are available for ≥30 years (selected from 337 IRIMO climatology stations); 2. Interpolate (1) into 0.05

Results of comparison between IS0902 and TRMM 3B42
3.1 Spatial comparison

Spatial distribution of annual average of rainfall
Figure 5 shows mean annual precipitation in a 0.25 • grid for the nine years from 1998 to 2006.There is good accordance in precipitation pattern (Fig. 5a and b).The maximum mean annual rainfall in IS0902 (gridded precipitation from Iran Synoptic gauges Ver.0902) is 3.5 mm day −1 (Fig. 5a) and for TRMM 3B42 is 2.0 mm day −1 (Fig. 5b).TRMM 3B42 underestimates precipitation along the Caspian Sea and the Zagros Mountains area by a maximum of 2.0 mm day −1 (Fig. 5c).

Spatial distribution of mean winter rainfall
Because Iran receives rainfall mostly during winter (December-January-February (DJF)), we selected the winter rainfall distribution to compare to satellite data.

Statistical comparison
Using the 0.25 • ×0.25 • data set of annual rainfall means in IS0902 and TRMM 3B42 (1998-2006), a scatter plot of IS0902 versus TRMM 3B42 for three regions is presented in Fig. 7a, b and c Statistical parameters derived in linear regression and the spatial correlation coefficients between mean annual rainfall in IS0902 and TRMM 3B42 are presented in Table 1.Spatial correlation coefficients for the above-mentioned regions are 0.77, 0.57, 0.75, respectively, with significance level of derived spatial correlations of 99%.The number of samples shown in Table 1 refers to the number of grid cells used for spatial correlation.Figure 8 shows the spatial distribution of temporal correlation coefficient of mean annual rainfall in IS0902 and TRMM 3B42 from 1998 to 2006.TRMM 3B42 has high accordance (correlation coefficient.>0.7) with IS0902 in the northwest region and along the Zagros Mountains, and has low accordance (correlation.coefficient <0.7) with IS0902 in most part of the desert regions along the Caspian Sea.This result accords well with derived spatial correlations for the three regions, as shown in Table 1.

Discussion and conclusion
One of the most important features of this study is the accordance between observation time in TRMM 3B42 and the synoptic gauges, which is 8 times every three-hours in UTC time.The daily value is derived by summation of the three-hourly observations from 00:00 UTC to 21:00 UTC.A key outcome of the study is the low correlation in rainfall between IS0902 and TRMM 3B42 along the Caspian Sea.A single rain band was evident in the southern part of the Caspian Sea in IS0902, but is separated into two distinct rain bands in TRMM 3B42; one along the Caspian Sea and the other over the Alborz Mountains.This may be related to the complicated topography in the north of Iran and the lack of sufficient rain gauge data over the Alborz Mountains.
We have shown that spatial distribution of mean annual and seasonal precipitation over Iran has two main patterns along the Caspian Sea and the region along the Zagros Mountains.Comparison of spatial distribution of gridded precipitation data and TRMM 3B42 showed good accordance in pattern and amount of precipitation.The scatter plot between mean annual precipitation estimated by TRMM 3B42 and mean annual gridded precipitation using synoptic network data the nine year period from 1998 to 2006 has been derived for the three analysis regions (entire country, Caspian Sea region, Zagros Mountains region) showed had www.adv-geosci.net/25/119/2010/Adv.Geosci., 25, 119-125, 2010   spatial correlation coefficients of 0.77, 0.57, 0.75, respectively.TRMM(3B42) underestimated precipitation by 0.17, 0.39 and 0.15 mm day −1 , respectively.The calculated correlation coefficients are in accordance with Xie et al. (2007), who e derived correlation coefficients of ∼0.7 between the Chinese gauge-based daily precipitation and satellite estimates based on TRMM 3B42.The spatial distribution of temporal correlation coefficient between mean annual precipitation of IS0902 and TRMM 3B42 is 0.8 in the Zagros Mountains region, and 0.6 along the Caspian Sea and in the desert regions.
Future research will compare high-resolution gridded precipitation (IS0902) data with other satellite data such as PERSSIAN, and CMORPH further determine the optimum remote-sensing information to combine with surface observations of precipitation.It will also be necessary to merge gridded surface-based rain gauge data and space-based rainfall estimates to achieve more accurate monitoring of precipitation over Iran.
they.Satellites-based data sets need to be examined and verified by comparison with ground-based rain gauge data.Xie et al. (2007) verified Published by Copernicus Publications on behalf of the European Geosciences Union.
Iran lies in the world's dry belt.Iran (25 • -40 • N, 45 • -60 • E) is one of the most mountainous countries bordering the Gulf of Oman, the Persian Gulf, and the Caspian Sea.Sixty percent of Iran is covered by mountains.The central parts of the country comprise two dry deserts: the Dasht-e-Kavir and the Dasht-e-Lut.The country's topography is dominated by two mountain ranges.The Alborz range in the north, close to the Caspian Sea extends east-west with a maximum altitude of ∼5000 m.The Zagros Mountains cross the country

Fig. 3 .
Fig. 3. Time series of the number of synoptic gauges used in this study.

Fig. 7 .
Fig. 7. Scatter plots of annual average of rainfall from IS0902 versus TRMM 3B42 for each 0.25 • ×0.25 • grid cell (a) the entire country, (b) the Caspian Sea Region, (c) the Zagros Mountains.

Figure 8
Figure 8 shows a comparison between time series of monthly areal average of precipitation (hereafter MAAP) over Iran (25 • -40 • N, 45 • -60 • E) in IS0902 and TRMM 3B42 V6 from 1998 to 2006.The fluctuations in variation of MAAP in IS0902 and TRMM-3B42 accord well with each other.Maximum MAAP occurs in winter, and minimum rainfall occurs each summer in both IS0902 and TRMM 3B42 (Fig.9).Comparison of areal rainfall average shows that TRMM 3B42 underestimates rainfall compared with IS0902; for example, the maximum of areal average is 2.5 mm day −1 in IS0902 and is 2.0 mm day −1 in TRMM 3B42 for the winter of2003/2004.

Fig. 8 .
Fig. 8. Spatial distribution of temporal correlation of average annual rainfall in IS0902 and TRMM 3B42 from 1998 to 2006.

Table 1 .
Statistical parameters of linear regression and spatial correlation coefficient in Fig.7."*" denotes significance level of derived spatial correlation is 99%.

Table 1 .
Statistical parameters of linear regression and spatial correlation coefficient in Fig.7.
* * denotes significance level of derived spatial correlation is 99%.