Evaluation of Satellite-based Rainfall Data in Flood Prediction

Rainfall-runoff transformation is a solution to the difficulty of obtaining observed discharge data in flood prediction analysis. Rainfall-runoff transformation requires observed rainfall data with a high rate of accuracy spatially. However, observed rainfall data is also often not available. Satellite rainfall data is commonly used to replace observed rainfall data. However, the accuracy of satellite rainfall data still needs to be tested. This study applied rainfall-runoff transformation to the observed rainfall data and the PERSIANN, GPM, and GSMaP satellite rainfall data in the Opak Watershed using GAMA I SUH method, which were then compared with the observed hydrograph at the AWLR Kretek during the flood event that occurred in Yogyakarta Province due to Cyclone Cempaka to evaluate their accuracy. The results showed that the GPM data generated a hydrograph that is the closest to the observed hydrograph, both the shape and the peak of the hydrograph.


INTRODUCTION
Yogyakarta City is the economic center of the Yogyakarta Province.Yogyakarta Province has quite high rainfall and several major rivers in the Opak Watershed flow through the Yogyakarta City.With high rainfall and many rivers that flow into the Opak Watershed, flooding will be more prone to occur.Considering that the level of vulnerability to flooding is quite high and some of the water flows through the Yogyakarta City, it is necessary to monitor and observe the behavior of the rivers in the Opak Watershed to overcome the adverse effects that can be caused by the flow of these rivers.Saputra et al. (2019) in their research stated that the Cempaka Cyclone event 27-30 November 2017 caused a maximum flood discharge of 2,185 m 3 /s and resulted in overflow in the Opak River at several points from the Opak -Oyo River confluence to Kretek.
One of the observations that needs to be made in the Opak Watershed is the observation of river flow discharge, but this often causes problems due to the unavailability of observed discharge data over a long time at the Automatic Water Level Recorder (AWLR) station.This problem can be overcome by utilizing observed rainfall data from rainfall gauge stations (ARR) in the watershed, which are usually available for a long period using certain analytical approaches to estimate unavailable discharge data.The transformation of rainfall data into an estimate of discharge can be done by constructing a flood hydrograph using a Unit Hydrograph (UH).
However, sometimes rainfall data is not always available or the rainfall gauge station is too far from the watershed.Therefore, a solution is needed to be able to overcome this problem.Due to their large coverage, high spatial resolution, and temporal frequency, satellite weather radars produce observations that adequately represent precipitation structure and evolution (Pidwirny, 2006 as cited in Mohamad et al., 2021).There are many types of satellite rainfall data that can be utilized in constructing a hydrograph.Gunawan (2008) and Natadiredja et al. (2018) states that satellite rainfall data has the potential to be used to fill in empty observed rainfall data.However, in general, satellite rainfall data has figures that tend to br different from actual precipitation due to the satellite radar beam blockage by obstacles, overshooting and partial beam filling, clutter, and the attenuation of the radar signals (Ryzhkov & Zrkić, 2019as cited in Mohamad et al., 2021).So that the discharge data from the transformation of satellite rainfall data also does not match the observed discharge data in the field.From https://journal.umy.ac.id/index.php/st/issue/view/1036the results of his research on the GSMaP (Global Satellite Mapping of Precipitation), CHIRPS (Climate Hazards Group InfraRed Precipitation with Station), and GPM (Global Precipitation Measurement) satellite rainfall data in South Lampung Regency, Pratama et al. (2022) state that for rainfall intensity, the three satellite rainfall data still have quite a large error over the observed data even though the ability to detect rainfall is good.From the results of research conducted by Trisantikawaty & Sepriando (2015), the Tropical Rainfall Measuring Mission (TRMM) satellite rainfall data is not good enough to be used for estimating daily rainfall, but good enough to be used for estimating monthly rainfall.
From the problems above, it is necessary to evaluate the accuracy of the satellite rain fall data.In previous studies, Ginting et al. (2019) analyzed the relationship between GPM and PERSIANN (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks) satellite rainfall data and the observed rainfall data reviewed at the ARR Kalibawang.In addition, Harsanto et al. (2021) compared direct runoff hydrographs from GPM 3IMERGHH satellite rainfall data in the upstream Winongo sub-watershed and observed data from other sub-watershed with similar characteristics.
This study evaluated the accuracy of observed and satellite rainfall data, namely PERSIANN, GPM, and GSMaP, compared to the observed discharge of the Opak Watershed, reviewed at the AWLR Kretek.In this research, the analysis of rainfall-runoff transformation using the GAMA I SUH method was carried out for the Opak Watershed, which is divided into several sub-watersheds.The hydrograph resulting from the rainfall data is compared with the AWLR Kretek hydrograph to see the level of accuracy of observed and satellite rainfall data.

Research Location
The location of this research is the Opak Watershed, which is administratively located in Yogyakarta Province, covering Sleman Regency, Bantul Regency, Yogyakarta City, and Gunung Kidul Regency as can be seen in Figure 1.The Opak Watershed empties into the south coast of Yogyakarta, precisely in Bantul Regency.
The location point for the review in this study is the AWLR Kretek, which is placed on southwest (downstream) of the Glondong (Kretek) Bridge and northeast (upstream) of the Kretek Weir, to be precise at 110°18'52.9"E and 7°59'25.9"S. The total area of the Opak Watershed studied at the AWLR Kretek location is ±1248 km 2 .

Research Steps
In general, the research steps that will be carried out in the evaluation of satellite rainfall data in the Opak Watershed are as follow.(1) Collecting sub-watersheds map data in the Opak Watershed and characteristics of the sub-watersheds.(2) Generating the average rainfall of the watershed and sub-watersheds for observed rainfall data using the Thiessen Polygon method.(3) Generating the average rainfall the watershed and subhttps://journal.umy.ac.id/index.php/st/issue/view/1036watersheds for satellite rainfall data based on the area of influence of the grid.(4) Processing land use data obtained from Regional Development Planning Agency of Yogyakarta (Bappeda DIY) and the Indonesian Topographical Map (RBI) using ArcMap 10.5 software to obtain the area of each land use for each subwatershed.( 5) Determining the type of soil in each sub-watershed using the Harmonized World Soil Database (HWSD) v 1.2 soil type map provided by the Food and Agriculture Organization (FAO) by entering boundary data for each sub-watershed in shapefile (shp) format.( 6 8) Generating unit hydrographs using the GAMA I method for each sub-watershed.( 9) Creating the sub-watershed and river model into the HEC-HMS 4.7.1 software from the upstream of the Opak River to the AWLR Kretek, with the river routing method using the Lag method, and the loss method using SCS-CN, and baseflow using constant flow.( 10) Entering data into the HEC-HMS 4.7.1 software, these data include characteristic data for each sub-watershed such as data on sub-watershed area, river length, hourly rainfall data, unit hydrograph, CN value, and b aseflow.(11) Conducting simulation/running after creating the watershed and river models.( 12) Obtaining the simulation results in the form of flood hydrograph data.( 13) Comparing the hydrographs of rainfall data to the hydrograph of AWLR Kretek.

Data Collection Administrative and topographical data
Administrative and topographical data for the Opak Watershed were obtained from the Serayu-Opak River Basin Center (BBWSSO, 2020).The land use map was obtained from Regional Development Planning Agency of Yogyakarta (Bappeda DIY, 2016).Indonesian Topographical Map data (RBI Map) obtained from the Geospatial Information Agency (BIG), which is provided online (BIG, 2019).The soil type and texture map was obtained from the Harmonized World Soil Database (HWSD) v 1.2 map provided by the Food and Agriculture Organization (FAO), which is provided online (FAO, 2013).Then the data is proce ssed using ArcMap 10.5 software.

Observed discharge data
The observed discharge data uses hourly water level recording data from AWLR Kretek obtained from Serayu-Opak River Basin Center (BBWSSO, 2021).The selected flood event is flood event on 27-30 November 2017 when Cyclone Cempaka occurred in Yogyakarta Province.

Observed rainfall data
The observed rainfall data in this study used hourly automatic rainfall recording data from rainfall stations around the Opak Watershed obtained from Serayu-Opak River Basin Center.19 rainfall stations data available around the Opak Watershed on 27-30 November 2017.The 19 rainfall stations are shown in Figure 3 (BBWSSO, 2021).
The satellite rainfall data devide into grids with specific spatial resolution, where each grid has a different precipitation value from the others.The grids of PERSIANN satellite rainfall data have a spatial resolution of 0.25°×0.25°.The grids of GPM and GSMaP satellite rainfall data have a spatial resolution of 0.1°×0.1°.https://journal.umy.ac.id/index.php/st/issue/view/1036

GAMA I Synthetic Unit Hydrograph
This study uses GAMA I for the analysis of the rainfall-runoff transformation.GAMA I was developed based on the hydrological behavior of 30 watersheds in Java Island in 1985 by Prof. Dr. Ir. Sri Harto Brotowiryatmo, Dip. H. (Brotowiryatmo, 2009).
The data needed in the analysis of GAMA I are.

Lag Time
The simplest method for flood tracing is the lag method provided by HEC in HEC-HMS 4.7.1 software.The Lag method is generally represented by the lag time equation as below.
https://journal.umy.ac.id/index.php/st/issue/view/1036t c is the concentration time, which is the time required for water to flow from the upstream point of the river to the control point of the river.The equation of Kirpich concentration time is shown below (Lydia & Mutia, 2015).
= 0,0663 0.77  −0,385 , (hour) (13) HEC-HMS 4.7.1 HEC-HMS (Hydrologic Center -Hydrologic Modeling System) is a model that can be used to transform rainfall into event flow or continuous flow through a watershed system and can perform flow routing analysis facilitated by hydrologic routing models (Scharffenberg, 2016).
The main components in the HEC-HMS 4.

Sub-watesheds
From the generating and merging of sub-watersheds using ArcMap 10.5 software, 27 sub-watersheds were obtained which were then modeled in HEC-HMS 4.7.1 software.The sub-watersheds map is shown in Figure 2. The area (A) of each sub-watershed is shown in Table 1.

Rainfall Data
Observed rainfall data for each sub-watershed that has been processed based on the Thiessen polygon method as shown in Figure 3.The PERSIANN, GPM, and GSMaP rainfall data for each sub-watershed which has been processed based on the coefficient of the area of influence of the grid on the sub-watersheds as shown in Figure 4 and Figure 5.
https://journal.umy.ac.id/index.php/st/issue/view/1036Rainfall-runoff tranformation analysis was carried out by modeling using the HEC-HMS 4.7.1 software, which is the result of input data for each sub-catchment that had been prepared previously.These data include subwatershed area, CN values, baseflow, rainfall data, and unit hydrographs.In addition, there is also input data in the form of lag time derived from the concentration time (t c ) of the Kirpich method for several sub-DAS where there are flood routings as presented in Table 4. of the rainfall-runoff tranformation reviewed at the AWLR Kretek produced hydrographs for each -respective data shown in Table 3 and Figure 7.Each hydrograph in Figure 7 is the accumulation of all sub-watersed hydrographs that are derived from each rainfall data using GAMA I SUH method and are flood-routed to AWLR Kretek point with HEC-HMS 4.7.1 software.From visual observations in Figure 7, it can be seen that the GPM hydrograph has a peak discharge value that almost matches the peak discharge value of the AWLR Kretek hydrograph.In addition, the GPM hydrograph also forms 2 (two) peaks like the AWLR Kretek hydrograph, while hydrographs from other rain fall data do not form 2 (two) peaks.To be more precise in seeing the difference in the peak discharge value of each hydrograph of rainfall data to the AWLR Kretek hydrograph, the difference values are presented in Table 4. From Table 4, it can be seen that the difference in the value of the peak discharge from the hydrograph of the observed rainfall data analysis to the peak discharge of the AWLR Kretek is 353.0 m 3 /s.The difference in the peak discharge value from the PERSIANN hydrograph to the the peak discharge of the AWLR Kretek is 1035.9m 3 /s.The difference in the peak discharge value from the GPM hydrograph to the peak discharge of the AWLR Kretek is 14.9 m 3 /s.While the difference in the value of the peak discharge from the GSMaP hydrograph to the peak discharge of AWLR Kretek is 164.4 m 3 /s.

CONCLUSION
From the results of this study, it can be concluded that from the analysis of the rainfall-runoff tranformation of observed rainfall data, PERSIANN rainfall data, GPM rainfall data, and GSMaP rainfall data, the GPM rainfall data produces a hydrograph that is closest to the shape and the peak discharge of the AWLR Kretek hydrograph during the flood event on 27-30 November 2017.

Figure 1 .
Figure 1.Research Location (BBWSO, 2020) ) Processing land use data and soil type data into CN (Curve Number) values and CN composite values.(7) Determining the baseflow discharge for each sub-watershed based on the area of influence of the sub-watershed on the Opak Watershed.( ) -Number of order 1 (O1) -Number of all orders (O) -Total length of order 1 (L O1 ) -Total length of all orders (L O ) -Watershed width at 0.75L from the control point (W U ), (km) -Watershed width at 0.25L from the control point (W L ), (km) -Upstream watershed area (A U ), (km 2 ) -River bed slope (S) -Number of river confluence (JN) -Source factor (SF)= the river order used in GAMA I is the Strahler method.Equations of peak time and peak discharge in GAMA I are.-Peaktime (T R )

Table 1 .
BaseflowEstimating the baseflow is important in order to predict the discharge that has the potential to generate flood.The baseflow (Q B ) data used in the input for flood routings in the Opak Watershed are shown in Table1(BBWSSO, 2021). https://journal.umy.ac.id/index.php/st/issue/view/1036

Table 2 .
Lag time values for some sub-watershedsHydrographsRainfall-runoff tranformation was carried out on observed rainfall data and GPM, PERSIANN, and GSMaP rainfall data that occurred in the Opak Watershed during the flood event on 27-30 November 2017.The results https://journal.umy.ac.id/index.php/st/issue/view/1036

Table 4 .
Hydrograph peak discharge values Peak