Estimation of Rainfall Rate Cumulative Distribution in Indonesia Using Global Satellite Mapping of Precipitation Data
1. Introduction
Rain is a problem in telecommunication system. Raindrops cause significant attenuation of electromagnetic wave for the frequencies above 5GHz [14]. Attenuation is the function of rain rate and the operating frequency in which higher rain rates lead to the high attenuation. Therefore, accurate rain rate prediction is necessary for modeling the specific attenuation of rain.
International Telecommunication UnionRadiocommunication Sector (ITUR) provides mathematical model to predict the rain rate. The ITUR uses 40year ECMWF reanalysis (ERA40) data as input. ERA40 data have low spatial and temporal resolutions, namely, 1.125
∘
x 1.125
∘
and 6 hours, respectively. In this work, the use of ERA40 data was replaced with Global Satellite Mapping of Precipitation (GSMaP) data with better spatial and temporal resolutions. Replacing ERA40 data with Tropical Rainfall Measuring Mission (TRMM) PR 3A25 and TMPA 3B43 have been conducted for Malaysia [2]. The result shows that the use of TRMM data as input of ITUR P.837 model provide better estimation of rain rate. In Indonesia, replacing ERA40 data with TRMM data have been also conducted particularly using TRMM 3B42 and 3B43 data [5].
2. Methods
The main data used in this work are GSMaP reanalysis gauge data. GSMaP reanalysis gauge data contain rainfall data for tropical and subtropical areas (60
∘
N  60
∘
S and 180
∘
E  180
∘
W). Ten years GSMaP data were used in this work. The data have high spatial and temporal resolutions, i.e., 0.1
∘
x 0.1
∘
and 1 hour, respectively. GSMaP data are developed from several satellite observations that carry microwave radiometers with specified sensors. These sensors are in low earth orbit. The characteristics of microwave radiometers used in GSMaP data are given in Table 1 [6] and in GSMaP Data Format Description [7].
Table 1
Characteristics of satellites involved in GSMaP data.

Satellite

Altitude (km)

Sensor

Frequncies (GHz)

TRMM 
402 
TMI 
10,19,21,37,85 
AQUA 
705 
AMSRE 
7,10,19,24,37,89 
DMSPF11 
803 
SSM/I 
19,37,85 
DMSPF13 
803 
SSM/I 
19,37,85 
To calculate 1min integrated rain rate, ITUR P.8376 required Mt and P
0
, which are respectively mean yearly rain accumulaltion (mm) and average probability to have rain. In this work, those parameters are extracted from GSMaP data and compared with the TRMM and rain gauge data. The following equation is the way to calculate 1min rain rate available in the Annex 1 of the ITUR P.8376 [8], as in Azlan et al. [2], which is given by:
Nevertheless, P
r6h
and
β
are no longer needed, because P
0
was simply calculated by [2]:
where NR is the number of rainy pixel and NT is the number of total pixel. To validate the results, 7 years (2003, 2004, 2005, 2008, 2009, 2010, 2011) Optical Rain Gauge (ORG) data from Kototabang, West Sumatera, Indonesia (0.20
∘
S, 100.32
∘
E; 864 m above sea level) were used. Those data have availability more than 80% [1]. In addition to ORG data, the DBSG3 data also used to validate 1min rain rate. DBSG3 data contain radiowave propagation measurement data that have been submitted to and accepted by ITUR Study Groups 3. DBSG3 data are used by ITUR for testing related prediction methods contained in the Pseries of ITUR Recommendation – Radiowave Propagation such as ITUR P.8376 which is used in this work.
The accuracy of the rain rate of each model is examined using root mean square error (RMSE). The smaller the value of RMSE, the higher the accuracy of the method. RMSE1 denotes the accuracy between GSMaP data and the ORG, RMSE2 denotes the accuracy between TRMM and the ORG, and the RMSE3 denotes the accuracy between ITUR and the ORG.
3. Results
Fig. 1 shows the comparison of 1min rain rate derived using yearly Mt and P
0
from GSMaP, TRMM data, ITUR model and ORG at Kototabang. It can be observed that the ITUR provides better 1min rain rate especially for percentage of time
>
0.01%. On the other hand, for percentage of time
≤
0.01% GSMaP provides better 1min rain rate, while ITUR with default input underestimates the rain rate.
Figure 1
Comparison of 1min rain rate estimated using yearly Mt and P
0
of TRMM and GSMaP along with the result of ITUR model with default input and rain gauge data at Kototabang.
Figure 2
Same as Fig.1 but for average Mt and P
0
.
Fig. 2 shows the comparison of 1min rain rate derived from average Mt and P
0
from GSMaP data (10 years), TRMM data (17 years), ITUR, and the ORG (7 years) at Kototabang. It was found that 1min rain rate obtained from GSMaP data show the best performance for percentage of time
≤
0.01%. For percentage of time
>
0.01%, ITUR shows better performance in calculating the rain rate. However, for the entire data, GSMaP provides the best performance in calculating the rain rate, indicated by smaller RMSE (Fig. 2).
Figure 3
Same as Fig. 2 but the comparison with DBSG3 data for Bandung.
Fig. 3 shows 1min rain rate estimated by using average Mt and P
0
from GSMaP (blue line) and TRMM (green line), ITUR (red line), and the DBSG3 (cross) data for Bandung. Bandung #1, Bandung #2 and Bandung #3 are data from 1992 to 1994, the rest are data with unknown year period. It is found that the accuracy of model for all data vary from year to year. For Bandung#3, oneminute rain rate estimated by ITUR model with GSMaP input is in good agreement with DBSG3. However, for Bandung#2, oneminute rain rate estimated by ITUR model with default input is in good agreement with DBSG3.
In general, for GSMaP data the best accuracy is obtained for percentage of time
≤
0.01%, while for percentage of time
>
0.01% default ITUR model shows the best rain rate estimation. This condition may be due to the original data of ITUR to estimate rain rate. The ITUR data mostly come from middle and high latitudes such as Europe, North America and Japan which are dominated by light rain (high percentage of time) [9].
4. Conclusions
In general, the use of yearly and average Mt and P
0
derived from GSMaP data as input of ITUR model show better performance to calculate 1min rain rate than the original ITUR model, indicated by a smaller RMSE. However, the accuracy of 1min rain rate estimates from yearly data varies from year to year, as observed from the result which is validated using the DBSG3 data. An overall, the use of GSMaP data to calculate 1min rain rate shows the best performance for percentage of time
≤
0.01%, while the ITUR model provides the best result for the percentage of time
>
0.01%.
Acknowledgments
This study was supported by the 2017 and 2018 International Joint Collaboration and Scientific Publication grant from the Ministry of Research, Technology and Higher Education of the Republic of Indonesia (Contract No. 02/UN.16.1.17/PP.KLN/LPPM/2017 and 050/SP2H/LT/DRPM/2018). The authors thank to Japan Aerospace Exploration Agency and Goddard Space Flight Center (JAXA) and National Aeronautics and Space Administration (NASA) for providing the TRMM, and GSMaP data. The authors also thank to Dr. Toyoshi Shimomai for providing Optical Rain Gauge (ORG) data at Kototabang.