Introduction

Providing long-term daily rainfall observation data is very important for hydrological studies and flood event predictions. It is a big challenge in most arid and semi-arid regions, especially in the study area of Wadi (Valley) Ghoweiba watershed, due to the absence of established rain gauge stations. It is necessary to look for alternative sources to obtain the required rainfall data for hydrological study in this area. RS rainfall products or satellite-based precipitation products (SPP) measurements can provide such information. Nguyen et al. (2018) and Ghimire et al. (2022) indicated that SPP observations demonstrate that they are the most capable method to perfectly monitor rainfall everywhere on Earth. In the last two decades, several SPPs have been launched (Nguyen et al., 2018). For example, the Tropical Rainfall Measurement Mission (TRMM) was launched in 1997, followed by the global precipitation measurement (GPM) mission (Nguyen et al., 2018). SPP datasets are then provided to meet the requirements of long-term rainfall data. Recently, SPP datasets are provided by several agencies and institutes using different algorithms. Among these satellite-based datasets are various products, including the PERSIANN and NASA-POWER products, which are considered in this study.

Precipitation estimation from remotely sensed information using artificial neural networks is one of these products (PERSIANN) (Hsu et al., 1997; Hsu et al., 1999; Nguyen et al., 2018; Soroosh Sorooshian et al., 2000). The Center for Hydrometeorology and Remote Sensing (CHRS) at the University of California, Irvine, created this family of products in cooperation with NOAA, NASA, and the UNESCO program for the Global Network on Water and Development Information for Arid Lands (G-WADI) (Nguyen et al., 2019). PERSIANN, PERSIANN-CCS, and PERSIANN-CDR are the three SPPs in the PERSIANN family (Nguyen et al., 2019). In addition to the Dynamic Infrared Rain Rate near real-time (PDIR-Now) which was added recently. To meet the requirements of professionals, researchers, and the general public, the products are reachable through different web interfaces. These interfaces, which are available through the CHRS portal, enable various types of analysis, visualization, and download (Nguyen et al., 2019). The data for these products can be downloaded easily. The products have been extensively used by researchers in the areas of water resources, hydrology, and climate. For example, such studies compare the PERSIANN products to other SPPs, ground observations, and model simulations (Li et al., 2003; Mehran & AghaKouchak, 2014; Miao et al., 2015; Nguyen and Thorstensen et al., 2017b; S Sorooshian et al., 2002; Yilmaz et al., 2005). Various studies have been used these products for runoff prediction (AghaKouchak et al., 2010; Ashouri et al., 2016; Behrangi et al., 2011; Hsu et al., 2013; Liu et al., 2017), monitoring drought (AghaKouchak & Nakhjiri, 2012; Katiraie-Boroujerdy et al., 2017a), frequency analysis (Gado et al., 2017), precipitation forecasting (Zahraei et al., 2013), assimilation into climate models (Yi, 2002), trend analysis (Damberg & AghaKouchak, 2014; Nguyen and Sorooshian et al., 2017a), modeling soil moisture (Juglea et al., 2010), and tracking typhoons (Nguyen et al., 2014). According to a recent study conducted by Salehi et al. (2022), the PERSIANN-CCS, PERSIANN-CDR, and PDIR are capable of hydrological simulations as well as producing accurate precipitation estimates.

The other RS rainfall product is the NASA-POWER product, which offers unrestricted access to online agro-climatology archive data (Negm et al., 2017). It contains the extensive collection of weather parameters and fluxes of surface solar energy from NASA on a global scale. The rainfall data are available in time steps (e.g., hourly, daily, and monthly) which have been available since 1983 (Negm et al., 2017; Zhang et al., 2010). Various researches in both industry and academics have been carried out using the POWER-NASA database in different applications (Negm et al., 2017). For example, among of these researches, a study conducted to evaluate the POWER-NASA database's suitability for predicting daily meteorological factors and reference evapotranspiration by Negm et al. (2017), daily air temperature by Negm et al. (2018), Estimating wind speed probability distribution by Samal and Tripathy (2019), and Daily solar radiation by Sayago et al. (2020). A study conducted by Aboelkhair et al. (2019) in Egypt to evaluate NASA POWER datasets for temperature types and relative humidity found that the accuracy of NASA POWER's maximum temperature values is slightly over and underestimated in 80 and 90% of the stations. As well as, a significant underestimation of RH readings at the majority of stations (Aboelkhair et al., 2019). A recent research conducted in sultanate Oman by Marzouk (2021) indicated that the 2-m air temperature is accurately measured by NASA-POWER data, but the atmospheric pressure is underestimated by about 2.1 kPa. The 2-m relative humidity data are somewhat inaccurate, but the precipitation data are largely unreliable, with values that are markedly inflated when compared to actual recordings (Marzouk, 2021).

As a result of technical advancement in the usage of the RS products globally in the field of hydrology, the use of satellite data may be used instead of rain gauge stations measurements, which are not available or incomplete in arid and semi-arid regions such as the current case study (Wadi Ghoweiba watershed). The significance of this work is that it can be utilized to get precise maximum rainfall depth data for various return periods. Furthermore, offer an appropriate RS product for estimating the runoff created by strong rainstorms that cause flash floods in the target area. The target area is subjected to infrequent heavy storm events causing flash floods (Gado et al., 2019; Geriesh et al., 2004). These floods could affect the new development industrial project area and the Harbor of North El-Sokhna. This area is considered one of the most important areas in Egypt. It is located downstream of the study area watershed (Wadi Ghoweiba). Wadi Ghoweiba is considered one of the largest wadis in the Eastern region (Geriesh et al., 2004). Seasonal storm intensity of rainfall is produced in this Wadi in the winter season (Geriesh et al., 2004). The industrial project region and the Harbor of North El-Sokhna are therefore at risk, so it is critical to investigate extreme flood events and evaluate these threats. Therefore, it is very important to search for other reliable alternative solutions to provide continuous time series of daily rainfall by investigating the available remote sensing (RS) rainfall product that provides the time series of rainfall data.

Thus, the current study aims to investigate the appropriate RS rainfall product and use its data to develop frequency analysis and intensity–duration–frequency curve (IDF curve) at return periods of 5, 10, 25, 50, 100, and 200 years in the study area of Wadi Ghoweiba, Egypt. The assessment of the PERSIANN products with NASA-POWER datasets compared to the available Annual Maximum Daily Rainfall (AMDR) in the Suez Rain Gauge (SRG) station as a reference will be achieved.°

Data and Methods

Study Area

Wadi Ghoweiba Watershed (also named Ghweiba or Ghoweba) is one of the Red Sea streams that slopes toward the east to the Red Sea (Northern and Southern zones). The watershed lies at the North-Western portion of the Gulf of Suez between Jabal Al Jalalah Al Bahariyyah and Gebel Ataqa as shown in Fig. 1. It is located between latitudes (29° 11′ and 29° 45′) N and longitudes (31° 41′ and 32° 25′) E (Geriesh et al., 2004).

Fig. 1
figure 1

Location of Wadi Ghoweiba Watershed, delineated and extracted using WMS and ArcGIS softwares

Full size image

It is bordered by Jabal Ataqah from the North, Gulf of Suez from the east, Jabal Al Jalalah Al Bahariyyah from the south (which comprises the main catchment area of the basin) and some heights such as Jabal al Hayy, Jabal Sad El Naam and Gebel Ghurabun, from the west. It comprises the Harbor of North El-Sukhna and the most important newly developed  industrial area in Egypt. This watershed runs northwesterly, easterly, and westerly to debauch into the Gulf of Suez. It is located at the junction of three main asphaltic roads directed northward to the Suez City, southward to Hurgada and northwestward to Cairo.

The watershed area is approximately 2847 km2. The (WRRI, 2016) indicated that the basin area reaches about 2551 Km2 while its length reaches 58 km. It consists of a number of branches such as Shonah Wadi that divides into sub branches such as (Gharbah, Om-Gerfan, Al-Sheikh, Om-Garf, Seem, Al-Abyad, and Om-Reses) streams. The first branches start from Al-Galalah Mountain with a height of 1254 m and Om-Rewees Mountain with a height of 1260 m. In addition, the slope is steep at the tops of these mountains (WRRI, 2016).

The study area generally experiences arid to desert weather characterized by high temperature, high rates of evaporation, and low rainfall. However, the region can occasionally experience dangerous flash flooding due to heavy rainstorms. The average maximum air temperature is between 20 and 35 °C, with the highest recorded temperature of 46 °C occurring in June 1965 (Geriesh et al., 2004). The average minimum air temperature is between 8 and 20 °C, with the lowest recorded temperature being 4.1 °C (in January 1973) (Geriesh et al., 2004). From September to March, the relative humidity rises (from 55 to 60%), and from April to August, it falls (from 55 to 50%) (Geriesh et al., 2004).

Data Collection

As mentioned earlier, historical rainfall data for Wadi Ghoweiba are not available because there is no meteorological or rain gauge station has been built in the site. The nearest ground meteorological station with accessible observed daily rainfall values is the SRG station located at Suez City (a distance of about 40 km from the study area). It is difficult to get the complete time-series of daily rainfall values of this station from the responsible authority. However, the AMDR data of this meteorological station are collected from Refaey et al. (2022) for the period of 1965–2020. The SRG station's coordinates are 29°57′49″ N, 32°33′46″ E as shown in Fig. 1.

Thus, the first step of the methodology was to collect the AMDR data of the ground SRG station as shown in the flowchart (Fig. 2). Then, the second step is to collect the daily rainfall values for the past period from the available RS satellite-based datasets products of PERSIANN family products and POWER-NASA database for the same site of SRG station and the site of Wadi Ghoweiba watershed. This step was conducted by collecting the daily rainfall data from the TRMM—NASA website (POWER-NASA) (https://power.larc.nasa.gov/data-access-viewer/) based on its availability for 40 years from 1 Jan 1982–31 Dec 2020. The data were collected in different 20 virtual points over the target area. Following the same procedure, the daily rainfall data were collected from satellite-based datasets of the PERSIANN family products website (https://chrsdata.eng.uci.edu/) based on its availability, during the periods of 2000–2021, 2003–2021, 1983–2021, and 2000–2021 for PERSIANN, PERSIANN-CCS, PERSIANN-CDR, and PDIR-Now, respectively. The PERSIANN and PERSIANN-CDR provide precipitation estimates at 0.25° × 0.25° resolutions. The PERSIANN-CCS and PDIR-Now provide precipitation estimates at 0.04° × 0.04° resolutions. After that, the AMDR values were analyzed and compared with the ground measured (observed) AMDR values at the ground SRG station.

Fig. 2
figure 2

Methodology flowchart

Full size image

The comparison criteria were done through the performance of the estimated value by all RS products which was compared with the actual value at the SRG station using commonly used error measures, root mean square error (RMSE), bias error (Bias), and correlation coefficient (CC). The error measures RMSE, Bias, and CC are given by the following equations (Eqs. 1–3), respectively.

$${\text{RMSE}} = \sqrt { \frac{1}{n}\mathop \sum \limits_{i = 1}^{n} \left( {S_{i} - R_{i} } \right)^{2} }$$
(1)
$${\text{Bias}} = \frac{1}{n} \mathop \sum \limits_{i = 1}^{n} \left( {S_{i} - R_{i} } \right)$$
(2)
$$CC = \frac{{\mathop \sum \nolimits_{i = 1}^{n} \left( {R_{i} - \overline{R}} \right)\left( {S_{i} - \overline{S}} \right)}}{{\sqrt {\mathop \sum \nolimits_{i = 1}^{n} \left( {R_{i} - \overline{R}} \right)^{2} } \times \sqrt {\mathop \sum \nolimits_{i = 1}^{n} \left( {S_{i} - \overline{S}} \right)^{2} } }}$$
(3)

where Ri is the actual value of observation rainfall at the SRG station. Si is the value of rainfall of the RS product, and n is the total number of observations. \(\overline{R }\) and \(\overline{S }\) are the average of the observation rainfall data and satellite-based datasets, respectively.

At the end, the suitable RS product of the annual maximum daily rainfall data that have the best performance was selected. The maximum daily rainfall data of that product was collected for the study area of Wadi Ghoweiba and then used for frequency analysis.

In consequence, the AMDR data for the ground SRG station and the RS data (PERSIANN family products and POWER-NASA database) at the site of the station and at the same coordinates of the study area (Wadi Ghoweiba) were collected. Then, the analysis of the data was done and the comparison criteria was conducted. The second comparison criteria were achieved through the frequency analysis in order to select the appropriate RS product that can be used for hydrological and flood risk studies at Wadi Ghoweiba.

AMDR Analysis

Frequency analysis aims to use statistical probability distributions to calculate the amount and frequency of rainfall during different return periods. The procedure followed in this study include the following. First, the frequency analysis was conducted for the SRG station and the appropriate selected RS satellite-based dataset at Wadi Ghoweiba Watershed from the first comparison criteria of performance analysis (error measurements). Then, the AMDR values were adjusted over 24 h with the distributions of generalized extreme value distribution (GEV), Gumbel, Normal, Pearson T3, and Log-Pearson T3 using the Hyfran-Plus software for the 5-, 10-, 25-, 50-, 100-, and 200-year return periods.

In the case of goodness of fit measurement, the Chi-squared test was used to determine whether there is a significant difference between the anticipated frequency distribution and the observed frequency distribution (Coronado-Hernández et al., 2020). The value of the Chi-squared test can be determined using the following formula (Eq. 4).

$$x^{2} = \frac{{\mathop \sum \nolimits_{i}^{n} \left( {R_{i} - M_{i} } \right)^{2} }}{{M_{i} }}$$
(4)

In which, X2: The Chi-squared test value, Ri: The value of recorded, and Mi: The modeled value. In this study, the Chi-squared test values for each probability distribution were used as comparison criteria to choose the appropriate distribution that give the best fit. The design rainfall depth for the selected distribution is used to produce the IDF curves for the study area at the selected return periods.

IDF-Curves Analysis

This section demonstrates the procedure of IDF-curves development. The IDF-curves for the study area were generated utilizing the selected probability distributions presented in Sect. “AMDR Analysis” and short duration (sub-daily) rainfall data which are the main data required for the estimation of the rainfall intensity. However, the short duration data are not accessible in the available data, so the other alternative option is to the use SCS (NRCS) type II storm distribution. Most previous studies have been conducted using the “published depth period ratios or full synthetic storm distributions, such as the ratio of 24 h to 1 h precipitation depths contained in the SCS (NRCS) type II storm distribution, which is known to be the most conventional among SCS distributions” ( Awadallah & Younan, 2012). These depth duration ratios have been widely used and tested in different studies. For rainfall stations where extreme storms last more than 3 h on average, the SCS type II synthetic storm distribution can be used securely. (Awadallah & Younan, 2012; Awadallah et al., 2018). Since, the distribution of SCS type II storm is record per 10 min (10–1440 min), thus this distribution depth duration ratio will be used in this study to derive the IDF curve and calculate the rainfall intensity at the maximum rainfall intensity for considered durations at 5-, 10-, 25-, 50-, 100- and 200-year return periods.

Results

Rainfall Characteristics

The historical rainfall data obtained from the SRG station and RS (PERSIANN family and POWER-NASA database) were analyzed at daily scales. The results show that the rainfall pattern is varied over the study area. Figure 3 demonstrates the values and distributions of a sample daily rainfall in the study area for PERSIANN family and POWER-NASA products with the SRG station site which was extracted by the ArcGIS software. The rainfall distribution rate is based on the temporal and spatial resolutions of the PERSIANN family products (PERSIANN, PERSIANN-CCS, PERSIANN-CDR, and PDIR-Now) (Fig. 3). The rainfall amounts are different in the four products as shown in Fig. 3. In POWER-NASA dataset product, the rainfall distribution is extracted and presented in terms of points rainfall. The rainfall pattern is also varied over the study area as PERSIANN family products. In general, the highest rainfall amount is found to be over areas of high elevation, while it seems to be lower in coastal areas.

Fig. 3
figure 3

Sample of daily rainfall distribution in the study area for PERSIANN family and POWER-NASA data products with SRG station extracted by ArcGIS

Full size image

The analysis of the data presented in Fig. 3–POWER-NASA shows that 12 points out of the 20 have the same rainfall pattern (points from 1 to 12). These points are concentrated in the main catchment area of the Wadi Ghoweiba watershed. The rainfall pattern in the other points is slightly different. The average annual rainfall was in the range of 44.5–62.9 mm. The analysis of daily rainfall data shows that the AMDR amount was in the range of 1.1–86.6 mm/day over the study area during the period 1982–2021 using the data of RS products, while it reaches to 49.6 mm/day in Suez between 1965 and 2020. Heavy storm rainfall events have been recorded over the study area during the last period especially in some years (e.g., 1965, 1984, 1986, 1987, 2016–2021). These records of events, indicate that the study area is under risk of flash floods. Initially, the daily rainfall data obtained from these websites were compared with some available data in a published study (Refaey et al., 2022). The results of this comparison show that these recorded data need more checks. More comparisons with the observation data that recorded by the SRG station are conducted. The results are presented below.

Figure 4 shows a comparison between the maximum daily annual rainfall data from the SRG station and satellite precipitation products of PERSIANN, PERSIANN-CCS, PERSIANN-CDR, PDIR-Now, and POWER-NASA, respectively. It can be noticed that there is a wide variation between rain gauge station data and satellite data. So, to judge which satellite product is the better, statistical analysis were carried out.

Fig. 4
figure 4

Maximum daily historical rainfall data from both gauged site (SRG station) and satellite precipitation products of PERSIANN, PERSIANN-CCS, PERSIANN-CDR, PDIR-Now, and POWER-NASA database

Full size image

The error measures presented in Eqs. 1–3 were used in statistical analysis. The results obtained from the determination of RMSE, Bias, CC, and R2 error measures are presented in Table 1. The results are presented based on two cases (case I and case II) of purpose. The first case (case I) was based on the usage of all available rainfall data for each type of the five rainfall products with the gauged site (SRG station) data, while the second case (case II) was based on the usage of 30 years of rainfall values (70% of the available data) after excluding the overestimates data in some years for the two satellite precipitation products of PERSIANN-CDR and POWER-NASA. There are 38 years of the available maximum daily rainfall data from 1983 to 2020 in these two products. It is adequate to use 20—30 years of these data after excluding the overestimated rainfall values and make sure that these data gave the best fit and CC between these data and observed rain gauge data. The usage of this amount of maximum daily rainfall data is enough in hydrological and flash flood studies.

Table 1 Comparison metrics for the PERSIANN products with POWER-NASA datasets against the SRG station data
Full size table

In Case I, the results of RMSE, Bias, CC, and R2 show that PERSIANN-CDR gave a minimum value of RMSE than other PERSIANN family products as shown in Table 1. The RMSE values were about 10.19, 8.94, 17.30, and 11.60 mm for PERSIANN family products of PERSIANN, PERSIANN-CCS, PERSIANN-CDR, and PDIR-Now, respectively. The RMSE value for the POWER-NASA product was about 15.56 mm. The PERSIANN-CDR and POWER-NASA gave a minimum value of Bias and high values of CC than other products as shown in Table 1.

The Bias values were about -2.37 and 5.34, while the CC values were approximately 0.35 and 0.44 for PERSIANN-CDR and POWER-NASA, respectively. The POWER-NASA product gave better CC and R2 values than PERSIANN-CDR and other satellite precipitation products. The R2 values were low for POWER-NASA and PERSIANN-CDR with about 0.12 and 0.19, respectively. The R2 values for all other satellite products were very low as shown in Table 1. This also is demonstrated in Fig. 5.

Fig. 5
figure 5

Scatter plots of SRG station rainfall data Vs. PERSIANN, PERSIANN-CCS, PERSIANN-CDR, PDIR-Now, and POWER-NASA database

Full size image

Figure 5 shows the comparison between gauged site (SRG station) observation values and the five satellite rainfall products. It is clear from Fig. 5 that the PERSIANN-CDR and POWER-NASA gave a best fit than other satellite precipitation products. This means that these two satellite products are suitable to be used in the frequency analysis for another check for their applicability since their rainfall values are close to the observation values at SRG station in most years. In Case II, the results show that the RMSE and Bias were decreased from 8.94 to 2.63 and from 2.73 to 0.31 for the PERSIANN-CDR product; from 15.56 to 4.23 and from 5.34 to -1.01 for POWER-NASA product, respectively (Table 1). This means that the PERSIANN-CDR product looks better than the POWER-NASA product. The results also show that the fit in case II is enhanced and becomes better than the fit in case I as shown in Figs. 5 and 6. The CC and R2 were increased from 0.35 to 0.83 and 0.12 to 0.69 for the PERSIANN-CDR product, and from 0.44 to 0.89 and 0.19 to 0.79 for the POWER-NASA product, respectively.

Fig. 6
figure 6

Scatter plot of SRG station rainfall data Vs. PERSIANN-CDR and POWER-NASA products using 70% of rainfall data

Full size image

Frequency Analysis

Based on the outcomes in the previous section, the AMDR data at Wadi Ghoweiba Watershed for PERSIANN-CDR and POWER-NASA satellite products were selected to be used in the frequency analysis. As well as, the maximum daily observation rainfall data in the SRG station were also used in order to check the applicability of using these data for the Wadi Ghoweiba Watershed. All these data were used to determine and select the suitable design rainfall depth at different return periods among the RS products and SRG station.

The results of the frequency analysis using Hyfran-Plus are presented in Tables 2, 3, and 4. Since it helped choose the appropriate distribution when using the Chi-squared test. The results of the Chi-squared test for the three datasets PERSIANN-CDR, POWER-NASA, and SRG station are presented in Table 2 utilizing the five statistical probability distributions of GEV, Gumbel, Normal, Pearson T3, and Log-Pearson T3. The minimum Chi-squared test value is reached by the GEV distribution (best agreement) followed by the Log-Pearson T3 distribution for PERSIANN-CDR and POWER-NASA, while the minimum Chi-squared test value is reached by the Log-Pearson T3 distribution followed by the GEV distribution for SRG station (Table 2). However, the GEV distribution maximum likelihood graphic gave a better fit than the Log-Pearson T3 distribution and other distributions for the SRG station. Thus, based on the Chi-squared test and graphic, the GEV distribution presents the best fit comparing to the other distributions with Chi-squared value of 5.1, 6.4, and 7.54 for PERSIANN-CDR, POWER-NASA, and SRG station, respectively (Table 2). It is obvious from Table 2 that the POWER-NASA product gave an overestimate of Chi-squared test values than PERSIANN-CDR and SRG station at all statistical distributions except at GEV distribution. It is lower than the Chi-squared test value in the SRG station but it does not give a good fit in the graphic of maximum likelihood.

Table 2 Values of Chi-squared test at different statistical probability distributions
Full size table
Table 3 Basic statistical characteristics for PERSIANN-CDR, POWER-NASA, and SRG station rainfall data
Full size table
Table 4 Values of AMDR (rainfall design depth) for multiple statistical distributions at different return periods
Full size table

Based on the results of this comparison criteria the PERSIANN-CDR product is better than the POWER-NASA product. As well as, the GEV statistical distribution is the best distribution that gave the best fit. The graphics of maximum likelihood of this distribution are presented in Figs. 7 and 8. Thus, it is selected as the appropriate statistical distribution. The statistical characteristics using this distribution are determined and demonstrated in Table 3.

Fig. 7
figure 7

Distribution graph for statistical probability distributions

Full size image
Fig. 8
figure 8

Distribution graph of GEV distribution “maximum likelihood” for PERSIANN-CDR, POWER-NASA, and SRG station

Full size image

Table 3 presents the values of basic statistical characteristics for PERSIANN-CDR, POWER-NASA, and SRG station rainfall data using the GEV distribution. POWER-NASA product gave an overestimate of most statistical characteristics than PERSIANN-CDR and SRG station. For example, the AMDR values were about 35, 86.3, and 49.6 mm/day for PERSIANN-CDR, POWER-NASA, and SRG station, respectively. The AMDR value of 86.3 mm is very significant in POWER-NASA comparing to the AMDR observed in the SRG station as reference. The AMDR value of 35 mm in the PERSIANN-CDR is noted to be eligible comparing to the AMDR of 49.6 mm that observed in the SRG station. The comparison criteria confirmed that the PERSIANN-CDR product is better than the POWER-NASA product using GEV statistical distribution when comparing with the statistical values obtained for the SRG station as a reference. We conclude from Table 3 that the values of statistical coefficients for the PERSIANN-CDR product is closer to the SRG station values than the POWER-NASA product.

The results of AMDR obtained from the GEV, Gumbel, Normal, Pearson T3, and Log-Pearson T3 distributions are shown in Table 4 for the PERSIANN-CDR, POWER-NASA, and SRG station as reference. Based on the comparison criteria (Chi-squared test) of statistical distributions (Table 2), GEV is the best distribution for fitting the AMDR at the PERSIANN-CDR comparing to the values obtained by the SRG station but POWER-NASA gave higher values for all return periods as shown in Table 4. For PERSIANN-CDR, the AMDR values are 10.8, 15.2, 22.7, 30.3, 40, and 52.2 mm for the return periods of 5, 10, 25, 50, 100, and 200 years, respectively (Table 4). For the SRG station, the AMDR values corresponding to the return periods of 5, 10, 25, 50, 100, and 200 years are 11.3, 17.1, 27.6, 38.6, 53.1 and 72.5 mm, respectively. The AMDR values for the return periods of 50 and 100 years are 30.3 and 40 at the PERSIANN-CDR, 38.6 and 53.1 mm at the SRG station. This finding agrees with the results obtained by WRRI (WRRI, 2019).

In the series of AMDR measured for the periods 1983—2020 and 1965—2020 in PERSIANN-CDR product and SRG station, the values of 30.3—38.6 and 40—53.3 mm corresponding to the biggest flood reported in 1965 and 2020 have a return period of 50—100 years, classifying it as an extreme event.

Intensity Duration Frequency Curves (IDF)

The IDF curves were developed for the study area of Wadi Ghoweiba Watershed based on the results obtained from the frequency analysis. The best-fitted distribution for PERSIANN-CDR product and SRG station were GEV distribution. Accordingly, the AMDR values corresponding to the return periods of 5, 10, 25, 50, 100, and 200 years obtained from this distribution were used to generate short-duration rainfall depths over periods from 10 to 1440 min (24 h). The results of the IDF curves are demonstrated in Table 5 and Figs. 9 and 10. Wadi Ghoweiba Watershed can be represented either by the IDF curves developed by the AMDR of the PERSIANN-CDR product or SRG station. Figure 9 presents the generated IDF values for the PERSIANN-CDR product, while Fig. 10 presents the generated IDF values for the SRG station. The IDF curves that were generated show that the intensity of the rainfall increases as the return period increases. On the other hand, the intensity decreases for the same return period and in all return periods as the duration increases. For example, the rainfall intensity increases from about 12.1 to 58.8 mm/hr during the return period of 5 to 200 years at 10 min duration and then decreases gradually until reached about 0.5 to 2.2 mm/hr at 24 h (PERSIANN-CDR product—Fig. 9). The same behavior is obvious in Fig. 10 for SRG station in which the rainfall intensity increases from about 12.7 to 81.1 mm/hr during the return period of 5 to 200 years at 10 min duration and then decreases gradually until reached about 0.5 to 3 mm/hr at 24 h. In comparison between Figs. 9 and 10, it is found that the PERSIANN-CDR product gave values slightly lower than the values obtained by the SRG station. This means that the IDF curve obtained by PERSIANN-CDR can be effectively used better than other satellite products when there is no observation data.

Table 5 Intensity duration frequency curves (IDF) for PERSIANN-CDR product and SRG station (Wadi Ghoweiba Watershed) using GEV distribution at different return periods
Full size table
Fig. 9
figure 9

Intensity Duration Frequency curves (IDF) for PERSIANN-CDR product (Wadi Ghoweiba Watershed) using GEV distribution

Full size image
Fig. 10
figure 10

Intensity Duration Frequency curves (IDF) for SRG station (Wadi Ghoweiba Watershed) using GEV distribution

Full size image

Discussion

According to the previous results, the PERSIANN-CDR product is selected as the appropriate RS product. It is found that the PERSIANN-CDR product is mainly perfect for analyzing statistical annual maximum daily rainfall due to its extended historical records. The product shows an underestimate of extreme rainfall data than the SRG station and POWER-NASA. It has a reasonable level of accuracy. This finding agrees with the findings of previous studies conducted in different places worldwide. A study conducted by Katiraie-Boroujerdy et al. (2017b) using the PERSIANN-CDR product found that the product exhibits reasonable performance and shows relatively high correlations for intensity indices. However, the product generally underestimates extreme values. This finding agrees with the results obtained by Miao et al. (2015) which indicate that the values of extreme heavy precipitation are slightly underestimated. Ombadi et al. (2018) found that an annual maximum series obtained from the product over the CONUS region exhibits underestimation. Nguyen et al. (2018) indicated that the product is basically ideal for analyzing statistical trends of past extreme rainfall occurrences due to its lengthy historical record. The product shows an underestimate of extreme rainfall events and has a reasonable level of accuracy. When using the dataset to analyze the frequency of extreme rainfall occurrences and create IDF curves, this should be taken into consideration. Moreover, Liu et al. (2017) used the product dataset as an input to a hydrologic model to simulate historical stream flow. The results show that the simulated stream flow is more accurate than the use of interpolated data from ground rain gauge stations. In addition, the findings of a study conducted by Mosaffa et al. (2020) suggest that the product of PERSIANN-CDR is a good source of information in low-density gauge network locations, capturing the temporal and spatial variation of rainfall. A study conducted by Santos et al. (2021) emphasized the ability of the product to capture the rainfall pattern over the analyzed period when compared with the rain gauge-measured. Consequently, the PERSIANN-CDR product can be used in frequency analysis and extreme rainfall estimation.

There are some advantages and disadvantages to using the PERSIANN-CDR product. The advantages include the following: easy access to data, visualization, and download. The rainfall data are available to the public for free and provide the data in daily, monthly, and yearly time series from January 1, 1983, to the present. The other advantage is that the rainfall data can be used in arid and semi-arid regions in which there are no established rain gauge stations. The product is basically ideal for analyzing the statistical trends of past extreme rainfall occurrences due to its lengthy historical records. The disadvantage of the product is that it sometimes shows underestimates extreme values. This should be taken into consideration when using the data.

Conclusion

This study investigated the appropriate RS precipitation product that can be used for study extreme flood events and evaluate these threats on the downstream area of Wadi Ghoweiba Watershed, Egypt. The evaluation of the RS (satellite-based rainfall measurements) PERSIANN family products (PERSIANN, PERSIANN-CCS, PERSIANN-CDR, and PDIR-Now) with NASA-POWER datasets compared to available annual maximum daily rainfall in SRG station as a reference were conducted. The selection of the appropriate product was conducted through comparison criteria of rainfall characteristics analysis, frequency analysis, and IDF curves development. The frequency analysis was conducted utilizing the five statistical probability distributions of GEV, Gumbel, Normal, Pearson T3, and Log-Pearson T3 at different return periods of 5, 10, 25, 50, 100, and 200 years. The results show that the PERSIANN-CDR product is the appropriate satellite-based rainfall dataset product based on the outcomes of the comparison criteria of RMSE, Bias, CC, and R2 error measures of rainfall characteristics analysis and Chi-squared test of distributions.

Based on the comparison criteria (Chi-squared test) of distributions and graphic, GEV is the best distribution for fitting the AMDR at the PERSIANN-CDR comparing to the SRG station as a baseline but POWER-NASA gave significant values for return periods.

Despite the great results of method suitability in rainfall estimation in arid and semi-arid regions. The available rain gauge station data have to be given due consideration to increase its usefulness. The study recommends that more work should be carried out to improve the method and to include other parameter in the prediction.