Precision of raw and bias-adjusted satellite precipitation estimations (TRMM, IMERG, CMORPH, and PERSIANN) over extreme flood events: case study in Langat river basin, Malaysia

Extremities in weather condition cause flooding, which is one of the most widespread of hydrometeorological hazards that can be particularly disruptive, leading to widespread collapse of infrastructure in most of the regions in the world (Scofield & Kuligowski 2003; Khan et al. 2011; Seyyedi et al. 2014). For instance, the floods that happened at the end of the year 2014 in Malaysia (also known as 2014/2015 flood events) (Akasah & Doraisamy 2015) have been described as the worst flood in decades. Although it is known that flood on the eastern coast of Peninsular Malaysia is an annual occurrence during the northeast monsoon (NEM), the magnitude and damage due to the flood are not as high as compared to the 2014/2015 flood events in which more than half of Peninsular Malaysia, including those regions in the central part and western side, were submerged and caused most of the rivers to reach dangerous levels. Also, this event caused millions of ringgit of property damage with thousands of people affected and life loss.

For a few decades, rainfall–runoff modeling studies have become widespread in order to figure out the processes in water movement (Shabalova et al. 2003; Collischonn et al. 2008; Su et al. 2008; Dadhwal et al. 2010; Gu et al. 2010; Verma et al. 2010; Behrangi et al. 2011; Roy et al. 2013; Choudhari et al. 2014; Elgamal et al. 2017). Precipitation data are one of the most sensitive model inputs required for rainfall–runoff modeling (Su et al. 2008; Mair & Fares 2010; Behrangi et al. 2011; Jiang et al. 2012). Traditionally, rain gauge networks have been employed as the primary source of ground-based precipitation estimates. However, they are susceptible to certain errors, such as size of collector, evaporative loss, out-splash, leveling, siting of gauges, the effect of wind, etc. (Strangeways 2004).

Satellite precipitation products (SPPs) have received increased attention in estimation precipitation due to their representation of high space-time variability of the precipitation field with quasi-global coverage, hence are beneficial over ungauged catchments, especially mountainous and oceanic regions (Collischonn et al. 2008; Tian et al. 2009; Behrangi et al. 2011; de Coning 2013; Moazami et al. 2013; Gado et al. 2017), and potentially attractive for hydrologic modeling studies in data-sparse regions. Various global high-resolution SPPs are operationally available, including the Integrated Multisatellite Retrievals for the Global Precipitation Measurement (IMERG) (Hou et al. 2014; Huffman et al. 2017), Tropical Rainfall Measuring Mission (TRMM) Multi-satellite Precipitation Analysis products (TMPA) (Huffman et al. 2007), National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center morphing technique product (CMORPH) (Joyce et al. 2004), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) (Hsu et al. 1997; Sorooshian et al. 2000), etc.

The aforementioned SPPs had been widely applied and investigated around the world (Yilmaz et al. 2005; Jiang et al. 2010, 2018; Behrangi et al. 2011; Yong et al. 2012; Chen et al. 2014; Mei et al. 2016; Liu et al. 2017; Soo et al. 2018; Tan & Santo 2018; Tan et al. 2018). However, these satellite estimations are still imperfect and prone to systematic and random errors associated with observations, sampling, and retrieval algorithms (Dinku et al. 2009; Villarini et al. 2009; Pereira Filho et al. 2010; Piani et al. 2010; Teutschbein & Seibert 2013). The models can augment or suppress rainfall biases to streamflow based on the response mode of the model (Segond et al. 2007; Habib et al. 2014; Fang et al. 2015). Several bias correction (BC) schemes have been developed to downscale the meteorological variables from any datasets or models, ranging from the simple scaling approach to sophisticated distribution mapping (Haerter et al. 2011; Teutschbein & Seibert 2012). In recent years, numerous studies to improve SPPs' estimations by BC have been done varying location, season, topography, climatology, and so on (Boushaki et al. 2009; Tesfagiorgis et al. 2011; Chen et al. 2013; Habib et al. 2014; Fang et al. 2015; Abera et al. 2016; Gumindoga et al. 2016; Pan et al. 2016; Valdés-Pineda et al. 2016; Worqlul et al. 2017).

Although the correction of climate variables can considerably improve hydrologic simulations under current climate conditions (Teutschbein & Seibert 2012), there is a major drawback whereby most methods follow the assumption of stationarity of model errors, which means that the correction algorithm and its parameterization for current climate conditions are assumed to also be valid for a time series of changed future climate conditions. Whether or not this condition is actually fulfilled for our future climate cannot be evaluated directly. This motivated us to address this issue and to test how well different correction schemes perform for conditions different from those used for calibration.

The present study attempts to improve four different SPPs (IMERG, TRMM 3B42 Version 7 (V7), CMORPH, and PERSIANN) so that more accurate prediction of extreme events (in terms of both rainfall and streamflow simulation) can be achieved. Langat river basin was chosen as flooding is common in this study area when it coincides with localized rainfall. According to Soo et al. (2018), TRMM, CMORPH, and PERSIANN performed satisfactorily at Langat river basin during the 2014/2015 flood events. Thus, in the present study, the selected SPPs were improved by three different BC schemes, which are linear scaling (LS), local intensity scaling (LOCI), and power transformation (PT) methods. IMERG estimation is included as this estimation comprises an international constellation of satellites that provide rainfall estimations with significant improvements in spatiotemporal resolution, compared to TRMM products. Recent studies highlighted that the IMERG estimations can adequately substitute for TRMM estimations both hydrologically and statistically, despite limited data availability (Liu 2016; Tang et al. 2016; Tan & Santo 2018). However, the application of this estimation in hydrological models for the Malaysia region is still limited. Later, both raw and improved SPP estimations are employed in a Hydrologic Engineering Center's Hydrologic Modeling System (HEC-HMS) to simulate the rainfall–runoff at Langat river basin during the 2014/2015 flood events.

The Langat river basin, located in the western part of Peninsular Malaysia (latitude 1°30′–2°10′N and longitude 103°20′–104°10′E) (Figure 1) has been selected as a case study based on its history of great floods (Saudi et al. 2017). This basin covers the state of Selangor and Negeri Sembilan and also a portion of the Federal Territory of Putrajaya, Kuala Lumpur and Klang, and Petaling Jaya district. It has a total catchment area of about 2,350 km². The larger part of the basin totaling 1,900 km² occupies the south and south-eastern parts of the state of Selangor. There are three major tributaries, i.e., Langat River (is the main river), Semenyih River, and Labu River (Lian et al. 2019). The Langat River has a total length of about 180 km, draining from the main range (Banjaran Titiwangsa) at the northeast of Hulu Langat district in a south–southwest direction into the Straits of Malacca. Both Langat River and Semenyih River originate from the hilly and forested areas on the western slope of Banjaran Titiwangsa, northeast of Hulu Langat. This water catchment is important as it provides a raw water supply and other amenities to approximately 1.2 million people within the basin. Important conurbations served include towns such as Cheras, Kajang, Bangi, Government Centre of Putrajaya, and others (Atiqah et al. 2017). There are two reservoirs (Semenyih and Hulu Langat) and eight water treatment plants (four of which operate for 24 hours/day) (Saudi et al. 2017), which provide clean water to users after undergoing treatment. In terms of climate, high rainfall and high humidity occur at various periods throughout the year. The mean areal annual rainfall of this basin is 1,994.1 mm.

As shown in Figure 1(b), daily data at 28 operating rain gauge stations and four streamflow stations in Langat river basin were collected from the Department of Drainage and Irrigation (DID), Malaysia. Four SPPs, including the TRMM, IMERG, CMORPH, and PERSIANN were employed in the present evaluation. The selected resolution for each satellite product is summarized in Table 1.

The TMPA (TRMM Multisatellite Precipitation Analysis) was produced by the National Aeronautics and Space Administration (NASA). This product is a combined microwave-infrared precipitation product (Huffman et al. 2007), providing precipitation for the spatial coverage of 50°N–50°S at the latitude–longitude resolution. The latest version of this product, 3B42V7, can be freely downloaded from Goddard Earth Sciences Data and Information Services Center (GES DISC) (https://disc.gsfc.nasa.gov/). In this study, the daily aggregated TRMM 3B42V7 observations at a spatial resolution of 0.25° were analyzed.

IMERG (Integrated Multisatellite Retrievals for the Global Precipitation Measurement (GPM)) was launched on 14 February 2014 by the United States National Aeronautics and Space Administration (NASA) and the Japan Aerospace Exploration Agency (JAXA). This algorithm is intended to inter-calibrate, merge, and interpolate every satellite microwave precipitation estimate, together with microwave-calibrated infrared (IR) satellite estimates, precipitation gauge analyses, and potentially, other precipitation estimators at fine time and space scales for the TRMM and GPM eras over the entire globe. There are three main IMERGs: (1) IMERG Early Run which is a near-real time product with latency of 6 hours, (2) IMERG Late Run which is reprocessed near-real time with latency of 18 hours, and (3) IMERG Final Run, which is gauged adjusted with latency of four months. These IMERG products are made available in 30 minutes, daily and monthly temporal dimensions. According to Huffman et al. (2017), the IMERG Final Run is more accurate compared to other IMERG products as it is bias corrected using the Global Precipitation Climatology Centre (GPCC) precipitation gauges. The IMERG Final Run daily version 6 products at spatial resolution of 0.1° were employed in this study.

The CMORPH product (Joyce et al. 2004) is a pure satellite precipitation product using only satellite infrared information about the spatial and temporal evolution of rain clouds and not the rainfall estimates themselves. This product provides precipitation for the spatial coverage of 60°N–60°S. In the latest CMORPH Version 1.0, bias correction was conducted by adjusting the satellite estimates against a daily rain gauge analysis, and can be accessed from ftp://ftp.cpc.ncep.noaa.gov/precip/global_CMORPH/. Three spatial and temporal resolutions can be selected: 8 km/30 min, 0.25°/3 hourly, and 0.25°/daily. In this study, the 0.25°/daily bias-corrected Version 1.0 CMORPH data were analyzed.

The PERSIANN product estimates the rainfall rate from satellite observations by combining the infrared and passive microwave data using the artificial neural network function (Hsu et al. 1997; Sorooshian et al. 2000). This product can provide precipitation data for the spatial coverage of 60°N–60°S. In this study, the NOAA Climate Data Record (CDR) of PERSIANN data (PERSIANN-CDR) Version 1 Revision 1, which maintain the total monthly precipitation estimation with GPCP (Global Precipitation Climatology Project), at the spatial resolution of 0.25° and daily temporal resolution were downloaded from the following website (http://www.ngdc.noaa.gov/).

This study attempts to improve the SPP estimations by adopting BC schemes including LS (Lenderink et al. 2007), LOCI (Schmidli et al. 2006), and PT (Leander & Buishand 2007) methods to produce more accurate prediction before the data are ready to be input in hydrologic modeling. It was found that studies regarding the BC on SPP estimations in Malaysia appear to be limited. Finally, a simulation process was carried out to simulate the rainfall–runoff during the 2014/2015 flood events based on the raw and improved (LS, LOCI, and PT) SPP estimations (TRMM, IMERG, CMORPH, and PERSIANN). The HEC-HMS was employed to validate the performance of the raw and bias-adjusted SPP simulated flows with rain gauge model parameters. Figure 2 shows the overall procedure of this study.

Rain gauge (RG) measurement is considered as a point precipitation measurement and it cannot represent the volume of precipitation falling over a given catchment area. Therefore, a high density of RG stations spatially distributed over a catchment is crucial as a true representative precipitation of the area. However, often these true representative criteria are practically difficult to find in most countries. When a limited number of RG is compared to the satellite products, a point-to-grid precipitation is insufficient for the large variability of RG associated with the spatial and temporal resolution of satellite products. Therefore, conversion to a gridded surface from RG data at the same resolution of the satellite data by interpolation method is applied to overcome the large variability issue (Lo Conti et al. 2014). According to Soo et al. (2018), the areal precipitation pattern is almost similar, regardless of any spatial interpolation methods applied on the RG data. However, looking deeper into the results presented by them, excluding the result of arithmetic mean rainfall (as the mean precipitation computed is equally distributed to the whole basin), we noticed that inverse distance weighting (IDW) (Di Piazza et al. 2011; Ly et al. 2011, 2013; Wagner et al. 2012) performed slightly better compared to other rainfall interpolation methods. Dirks et al. (1998) reported that the IDW method is more accurate and feasible compared to other interpolation methods as it gives consideration to both complexity and calculating time. Thus, the point-based RG observations were interpolated by IDW, and these interpolated observations will be utilized in comparison with spatial-based satellite estimations and as input for hydrological modeling.

Despite considerable progress and evolution in recent years, SPPs exhibit large systematic and random errors which may cause large uncertainties in hydrologic modeling. Moreover, the models can augment or suppress rainfall biases to the streamflow based on the response mode of the model (Segond et al. 2007; Habib et al. 2014; Fang et al. 2015). Several bias correction schemes have been developed to downscale the meteorological variables from any datasets or models, ranging from the simple scaling approach to sophisticated distribution mapping (Teutschbein & Seibert 2012). However, these schemes have not been investigated for Malaysia, especially during a particular scenario or event. Thus, it is necessary to apply the BC schemes on the latest SPP estimations and evaluate them in terms of rainfall and streamflow simulation for Malaysia. In the present study, all SPPs were bias corrected utilizing three BC schemes, i.e., LS (Lenderink et al. 2007), LOCI (Schmidli et al. 2006), and PT (Leander & Buishand 2007) methods. In this study, quantile mapping, which is known as the best effective correction scheme, was not selected as this scheme ignores the correlation between raw ensemble forecasts and observations (Zhao et al. 2017). A more detailed description of the selected methods is presented below.

Similar to the LS scheme, the scaling factor was calculated and applied separately for every selected event.

To implement this method, there are two scaling factors to be calculated, a and b. b is calculated iteratively so that the coefficient of variation (CV) of the satellite daily precipitation time series matches that of the gauged precipitation time series. Next, a is calculated such that the mean of the transformed precipitation value matches that of the gauged precipitation. Finally, these two scaling factors are applied to each uncorrected daily satellite observation corresponding to that month to generate the corrected daily time series.

HEC-HMS is hydrologic modeling software developed by the US Army Corps of Engineers Hydrologic Engineering Center (HEC). This physically based and conceptual semi-distributed model is designed to simulate the rainfall–runoff processes in a wide range of geographic areas, such as large river basin, water supply, and flood hydrology to small, urban, and natural watershed runoff. The system encompasses losses, runoff transform, open channel routing, analysis of meteorological data, rainfall–runoff simulation and parameter estimation. HEC-HMS uses separate models to represent each component of the runoff process, including models that compute runoff volume, models of direct runoff, and models of base flow. Each model run combines a basin model, meteorological model, and control specifications with run options to obtain results. A schematic diagram for the setup of HEC-HMS for the Langat river basin is shown in Figure 3. The selected methods for each component of runoff process such as runoff depth, direct runoff, base-flow, and channel routing in event-based hydrological modeling are discussed in the following section.

The Muskingum method for channel routing is chosen. Under this method, the X and K parameters must be evaluated. Theoretically, parameter K is the time of passing of wave in reach length and parameter X is a constant ranging from 0 to 0.5. The parameters can be estimated with the help of observed inflow and outflow hydrographs. Parameter K is estimated as the interval between similar points on the inflow and outflow hydrographs. Once K is estimated, X can be estimated by trial and error (USACE-HEC 2008).

This study utilizes all RG observations, raw and bias-adjusted (LS, LOCI, and PT-adjusted) SPP datasets starting from 1 November 2014 to 20 January 2015 for the flood simulation in Langat river basin. As discussed in the section ‘Spatial interpolation of rain gauge observation’, we adopted the IDW-interpolated RG data to drive the HMS model and optimize the parameter values by comparing the simulated RG streamflow with the observed streamflow gauge station. Finally, the model is then forced and used to run the model with the RG optimized parameters. Figure 4 summarizes the overall hydrologic process simulation for this study.

This section focuses on the rainfall comparison at basin scale between raw SPP estimations with IDW-interpolated RG observations for 2014/2015 flood events in Langat river basin. Apart from that, further comparison with the bias-adjusted SPP estimations were included to observe the effect of bias correction on rainfall datasets at basin scale. Figure 5 shows the comparison of the daily and accumulated rainfall data of every raw and bias-corrected dataset over the focused on study period at Langat river basin, accompanied by statistical analysis tabulated in Table 2. Generally, comparing the IDW interpolated areal rainfall of RG observations with every original SPP estimations, TRMM, IMERG, and PERSIANN overestimated the overall rainfall series by 17.34, 16.06, and 21.71%, and CMORPH underestimated the total rainfall by 36.43%. After applying the BC schemes on every SPP it was noted that the bias of every SPP was greatly reduced regardless of any BC scheme applied. LS-corrected rainfall estimates predict the overall gauged rainfall very well. It is noted that the LOCI scheme was not really suitable for TRMM, IMERG, and PERSIANN estimations. Originally, TRMM overestimated the overall areal rainfall by 17.34%. However, after performing the LOCI scheme, it underestimated the overall areal rainfall by 11.40%. On the other hand, the LOCI scheme caused deterioration of the IMERG and PERSIANN performances, whereby both LOCI-IMERG and LOCI-PERSIANN exacerbate the overall rainfall over the basin by about 20% overestimation. This might be due to the setting of the rainfall threshold (1 mm) to ensure the threshold exceedance matches the wet-day frequency of the observation.

The section ‘Bias correction’ describes the methods of the BC that are employed to fit the mean, standard deviation (SD), and coefficient of variation (CV) for the precipitation data. Figure 6 shows several scatter plots for the fitting statistics of all events which shows the observed statistics are plotted versus those of the uncorrected and corrected satellite data. The detailed statistical performances are shown in Table 3. Based on the scatter plots, generally, it is observed that the LS scheme matches the mean precipitation of every satellite estimation, but it does not correct the biases in SD and CV. When applying a higher degree BC scheme, such as LOCI and PT schemes, significant improvement on the SD and CV were noted as the data points in the scatter plots almost match the gauged observations. PT exhibits greater improvement compared to LOCI. These results are considered as good, as the method of BC schemes applied for this study was only intended to correct the aforementioned statistical parameters.

In order to assess the runoff predictions obtained from the RG and selected SPP datasets, the daily IDW-interpolated RG rainfall data were first used to drive the HEC-HMS model and optimize the parameter values by comparing the simulated RG streamflow with the observed streamflow gauge stations. The objective of the model calibration is to match the RG simulated flow with the observed streamflow from DID and maximize the Nash–Sutcliffe efficiency (NSE) (at least 0.8). Among the parameters selected to calibrate the model were the curve number (CN), Muskingum factors K and X. As HEC-HMS is an event-based model, the overall study period was divided into five sub-events, namely, sub-events A, B, C, D, and E, and the simulation run and calibrated separately. The optimized parameters of every sub-event are listed in Table 4 and Figure 7 showing the comparison of simulated and calibrated runoff hydrograph, accompanied by Table 5 showing the model performance before and after calibration. It is observed that the optimized parameter in the HEC-HMS model of every sub-event gave values of different runoff hydrograph parameters close to the observed streamflow than that before optimization, with NSE ranging from 0.81 to 0.93. From Figure 7, sub-event B is the peak event of the overall study period. The statistical analysis of this sub-event gave values of MAE and RMSE of 96.9 m³/s and 106.7 m³/s, respectively. After the optimized values are considered, the performances of MAE and RMSE have improved to 33.5 m³/s and 38.9 m³/s, respectively. A similar case is noted for all other events, and the statistical analysis reveals that the optimized model parameters listed in Table 4 should be considered in the model to simulate the runoff hydrograph.

As discussed previously, the RG precipitation data were first used to derive the HEC-HMS model and optimize parameters against observed streamflow at the outlet. The model was then forced by raw and bias-adjusted TRMM, IMERG, CMORPH, and PERSIANN rainfall data with the RG optimized parameter values listed in Table 4. Figure 8 shows the comparison of hydrograph for every raw and bias-adjusted SPP estimation simulated flow. The statistical performance of each raw and bias-adjusted TRMM, IMERG, CMORPH, and PERSIANN simulated flow is presented in Tables 6–9.

Generally, with the rain gauge optimized parameters, the streamflow simulations from all three original SPPs do not show comparable results with the RG-calibrated streamflow. The simulations of the raw IMERG, TRMM, and PERSIANN overestimated the overall streamflow series by 2.4%, 23.5%, and 19.9%, respectively, due to their systematic overestimation of precipitation that was identified in the previous section. On the other hand, CMORPH simulation flow shows an underestimation of 45.6%. Generally, among the four SPP simulated flows, IMERG showed the best performance (as it shows the lowest RMSE and MAE) (Table 7), followed by CMORPH simulated flows. According to Bitew & Gebremichael (2011), CMORPH performed well in hydrological simulation compared to TRMM and PERSIANN, thus the result obtained in the present study is consistent. However, referring to Figure 8(a) and Table 6, it is noted that the original TRMM predicted the peak event (sub-event B) well, with NSE = 0.45, MAE = 75.1 m³/s, and RMSE = 95.7 m³/s. The other three SPPs (IMERG, CMORPH, and PERSIANN) (Tables 7–9) did not predict well the peak streamflow, as negative NSE and larger MAE and RMSE values are shown.

Comparing with other studies, TRMM rainfall has been shown to perform well in certain regions (Tian & Peters-Lidard 2007; Javanmard et al. 2010; Ochoa et al. 2014; Moazami et al. 2016). However, there are also some regions that do not reflect the performance of TRMM in hydrological simulation (Dinku et al. 2008; Haile et al. 2013). Haile et al. (2013) identified that the latest version of TRMM was improved (or bias-adjusted) based on the data from the GPCC (Zulkafli et al. 2014), instead of based on rain gauge data. The distribution of the GPCC and the number of stations per grid is scarce and therefore further adjustment has to be done to use TRMM 3B42 rainfall products. On the other hand, the results obtained for PERSIANN streamflow are consistent, according to Miao et al. (2015) in a similar study conducted in China. Liu et al. (2017) indicated that the PERSIANN-CDR rainfall product has good potential to be a reliable dataset and an alternative information source to a limited gauge network for conducting long-term hydrological and climate studies on the Tibetan Plateau, China.

The performance of the simulated flow using bias-adjusted SPP indicated an improved performance for all three SPPs. Based on the above analysis for every SPP, it is found that the BC schemes are able to improve the streamflow simulation, especially on the peak events of the study period.

For TRMM simulated flows, as shown in Table 6, it is noted that LOCI-corrected TRMM estimations (LOCI-TRMM) were found to be the best estimations compared to LS-TRMM and PT-TRMM, whereby the NSE, MAE, and RMSE had improved from −0.27 to 0.43, 70.1 to 49.8 m³/s, and 101.8 to 68.0 m³/s, respectively. Based on Figure 8(a), this estimation (LOCI-TRMM) matched the two highest peaks of the overall study period, i.e., in sub-event B (20^h November 2014–31 November 2014) and sub-event D (20 December 2014–4 January 2015).

On the other hand, for IMERG simulated flows, as shown in Table 7, LS-corrected IMERG estimations (LS-IMERG) were found to be the best estimations among all corrected SPP estimations, with NSE 0.44. Referring back to the rainfall analysis (Tables 2 and 3), it can be confirmed that the direct bias correction method (LS) is sufficient to improve the IMERG estimations.

For CMORPH simulated flows (Figure 8(c) and Table 8), the bias-adjusted simulated flows are improved equally regardless of any bias correction scheme. Simulation of bias-adjusted CMORPH rainfall estimate using the RG optimized parameters performs well with NSE around 0.30–0.40, and as for the peak event (sub-event B), the NSE ranges from 0.45 to 0.50. However, unlike the bias-adjusted TRMM, all three bias-adjusted CMORPH estimations deteriorate for the intermediate simulation flow in sub-events C and D. PERSIANN flows exhibit the lowest improvement regardless of which bias correction scheme was adopted on the rainfall estimations (Figure 8(d) and Table 9). However, surprisingly, the performance of LOCI-PERSIANN indicated a great improvement for sub-event B with NSE = 0.70, MAE = 70.3 m³/s, and RMSE = 57.7 m³/s.

The effect of bias correction of rainfall data on the simulation flow was also evaluated using the daily flow duration curves to assess the ability in simulating different ranges of streamflow and its probability of occurrence. The flow duration curves for streamflow simulated with RG, raw and bias-adjusted rainfall were plotted as shown in Figure 9.

Based on the flow duration curves of the original TRMM and its bias-adjusted plotted datasets (Figure 9(a)), it is observed that the streamflow simulated using LOCI-TRMM data followed closely the RG flow distribution even though there is a tendency to overestimate the streamflow at a range of 150–250 m³/s. This overestimation of streamflow could be due to the inaccurate simulation in the wet-day frequencies during bias correction (Smitha et al. 2018). The original IMERG seemed to be matched well with the RG duration curve, with a slight deviation at a range of 150–200 m³/s (Figure 9(b)). Comparing all bias-corrected IMERG distributions, it was noted for LOCI-IMERG that the streamflow distribution is deteriorated and not matched with the observed distribution.

For CMORPH (Figure 9(c)), the streamflow distributions simulated with the LS, LOCI, and PT methods are almost similar. This proves that the streamflow analysis is correct. Based on Figure 9(c), all bias-adjusted CMORPH streamflows indicate an overestimation at high stream flows (more than 200 m³/s). As for PERSIANN (Figure 9(d)), the LOCI and PT-PERSIANN streamflow distribution deteriorated, whereby the tendency of overestimation of streamflow is higher compared to the original PERSIANN distribution. In this case, LS-PERSIANN is the best among the three bias-adjusted PERSIANN data.

Based on the general result, there is room for improvement in order to adopt these bias-adjusted SPPs' estimations for flood prediction. Li et al. (2018) commented that the calibrated parameters of models have a tendency to be affected by the correlations between model parameters and observed data. Thus, it is recommended to use all raw and bias-adjusted SPPs as the forcing inputs to recalibrate the HEC-HMS model and then for validation in the same periods aimed at examining the influence of satellite precipitation datasets' uncertainty on streamflow simulations. Apart from that, we may examine the difference between the RG optimized model parameters and all raw and bias-adjusted SPP optimized model parameters.

Accurate and reliable precipitation data are the basis for hydro-climatological studies. SPP estimations provide alternative precipitation data for regions with sparse rain gauge measurements. Despite the continuing great efforts to develop fine resolution SPPs, the errors of SPP estimates cannot be removed completely because the characteristics of the retrieval errors vary in different climatic regions, seasons, and surface conditions (Sorooshian et al. 2011). In the present study, the capability of raw and bias-adjusted SPP estimations with rain gauge model parameters in the HEC-HMS model for the 2014/2015 flood events in Langat river basin was investigated.

The weather in Malaysia can be characterized by two monsoon regimes, namely, the southwest monsoon (SWM) from late May to September and the northeast monsoon (NEM) from November to March. Normally, the NEM causes massive heavy downpours of rain, particularly on the east coast states of Peninular Malayia, but the 2014/2015 flood crisis was regarded as one of the more devastating floods to hit Malaysia in recent decades (Akasah & Doraisamy 2015). The magnitude and damage of this flood crisis was high, whereby more than half of Peninsular Malaysia, including the Langat river basin (the current selected study area) and those regions at the central part and western side, were submerged, causing most of the rivers to reach dangerous levels.

Comparing the four original SPP estimations' (TRMM, IMERG, CMORPH, and PERSIANN) simulated flow with RG optimized parameters, the simulations of the raw TRMM, IMERG, and PERSIANN overestimated the overall streamflow series and CMORPH simulation flow showed an underestimation of 45.6%. TRMM had the potential to predict the peak streamflow although CMORPH show the best performance in general.

Next, we simulated the rainfall–runoff by replacing it with the bias-adjusted SPP estimations. Precipitation correction methods have more significant influence during high rainfall events, especially LS-adjusted IMERG and LOCI-adjusted TRMM. For PERSIANN-simulated flow, the BC schemes were able to improve the discharge simulation but only to a certain extent. Based on the general result, it is indicated that the current level of uncertainty in SPP estimations are still imperfect to be used in operational flood forecasting systems at the basin scale as the calibrated parameters are affected by correlations between model parameters and observed data. To avoid the calibration effects of different datasets, cross validation of different datasets is required.

The authors would like to greatly acknowledge the University of Malaya, Kuala Lumpur, Malaysia for financial support (FP039-2014B & PG194-2015B). The authors also would like to acknowledge the Department of Irrigation and Drainage Malaysia for providing the daily precipitation data as well as the developers of all SPPs for providing the downloadable data.