Accuracy Assessment of CMORPH and GPCP Satellite Precipitation Products Across Iran

Abstract

Reliable precipitation data are fundamental for climate and hydrological research, especially in regions with sparse ground-based observations. This study evaluates and compares the accuracy of two satellite-based precipitation products—CMORPH and GPCP—across daily, monthly, and annual scales over Iran. Daily, monthly, and annual precipitation estimates from CMORPH and GPCP were validated against observations from 128 meteorological stations distributed throughout the country. The assessment employed two statistical indices—correlation coefficient (CC) and root mean square error (RMSE)—alongside three categorical indices: probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI). At the daily scale, CMORPH outperformed GPCP in terms of CC, RMSE, POD, and CSI, while GPCP exhibited a lower FAR. At the monthly scale, correlations between satellite-derived and station-based precipitation were stronger than those at the daily scale; CMORPH achieved the highest correlation (CC = 0.84), whereas GPCP yielded a lower RMSE, with a mean value of 26.2 mm. At the annual scale, GPCP demonstrated better performance in CC, while CMORPH showed superior accuracy in RMSE. CMORPH consistently underestimated precipitation, whereas GPCP tended to overestimate rainfall across Iran. Although both datasets provided reliable precipitation estimates at the national scale, CMORPH demonstrated higher overall accuracy and efficiency. Its superior performance across most indices makes CMORPH the more suitable dataset for precipitation monitoring in Iran, despite its tendency to underestimate rainfall relative to ground observations.

1. Introduction

Precipitation is a complex and vital component of the Earth’s atmosphere, and its temporal and spatial variability strongly influences the environmental characteristics of any geographical region. As a key input variable in hydrological models, precipitation plays a critical role in water resources planning, including flood and drought analysis, monitoring, and forecasting [1,2]. Reliable information on precipitation amount, spatial distribution, and temporal variability is essential for modeling environmental processes and for developing effective water resource and hazard management policies [3].

Precipitation is commonly measured using three methods: rain gauges, radar, and meteorological satellites [4]. Rain gauges provide direct measurements and remain the primary source of precipitation data. However, due to the high spatial variability of rainfall, point-based measurements cannot be directly extrapolated to surrounding areas. As the spatial domain increases, estimation uncertainty also grows, leading to larger errors in representing precipitation over broader regions [5]. This limitation underscores the importance of satellite-based precipitation datasets, which offer continuous spatial coverage and help reduce biases associated with point-based observations.

In Iran, the distribution of rain gauge stations is uneven due to topographic constraints, impassable terrain, and uninhabitable desert regions. Most stations are located along the low-altitude slopes of the Zagros and Alborz mountains and in coastal areas, while the central, eastern, and southeastern regions—dominated by deserts—are sparsely covered [6]. Satellite systems therefore provide a valuable alternative, enabling global measurement of atmospheric parameters at regular intervals. Since the launch of the TIROS satellite in April 1960, which produced cloud images comparable to simultaneous meteorological observations, the number of satellite sensors monitoring the Earth’s atmosphere has grown substantially. Today, these sensors are the only tools capable of continuous, global precipitation estimation. They are generally classified into three categories: visible and infrared (VIS/IR), microwave (MW), and passive microwave (PMW) sensors [7,8].

Over the past few decades, numerous satellite-based precipitation datasets have been developed, including TMPA, PERSIANN, GSMaP, CHIRPS, SM2RAIN, IMERG, CMORPH, and GPCP [9,10,11,12,13,14,15,16,17]. These datasets have been widely applied in hydrology, climatology, water resources management, drought monitoring, and climate change studies [18,19,20,21,22,23,24]. Their use is particularly important in Iran, where sparse station coverage, short record lengths, and other limitations hinder access to reliable spatial and temporal climate information. However, before employing satellite-based datasets, their accuracy and performance must be rigorously evaluated.

Several studies have addressed this issue. Moazami et al. [25] assessed four satellite products (TRMM 3B42RT, TRMM 3B42V7, PERSIANN, and CMORPH) over Iran during a five-year period, concluding that the calibrated TRMM 3B42RT provided the most accurate estimates. Sharifi et al. [26] examined IMERG, ERA-Interim, and TRMM 3B42 across different regions of Iran, finding that all three products tended to underestimate precipitation relative to ground observations. Alijanian et al. [27] evaluated five products (CMORPH, PERSIANN-CDR, PERSIANN, TRMM, and MSWEP) against rain gauge data in eight climatic regions of Iran (2003–2012), reporting that performance varied by climate zone, with PERSIANN-CDR performing best in the warm and humid Persian Gulf region. Fallah et al. [28] compared interpolated and reanalysis datasets against rain gauge observations in the Karun basin (2000–2015), concluding that most datasets significantly underestimated precipitation, particularly in mountainous areas. Moazami and Najafi [29] evaluated GPM-IMERG V06 and MRMS data against hourly station records in Canada (2014–2018), finding that satellite products tended to overestimate precipitation intensity by approximately 25% in coastal regions. Their results suggest that IMERG and MRMS have the potential to complement ground-based observations at high temporal resolution.

Despite these efforts, no study has yet systematically compared the CMORPH and GPCP datasets specifically for Iran, which constitutes the primary objective of the present research. Accordingly, the main question guiding this study is: How do CMORPH and GPCP perform across different temporal scales and geographical regions of Iran?

Although both CMORPH and GPCP are widely used satellite-based precipitation products, they rely on fundamentally different algorithmic frameworks that influence their performance characteristics. CMORPH (Climate Prediction Center MORPHing technique) employs a motion-based propagation approach in which precipitation features derived from passive microwave observations are advected using atmospheric motion vectors derived from infrared imagery. This morphing technique enables high spatial and temporal resolution (0.25° × 0.25°, 3-hourly) but may introduce errors in regions with rapid convective development or complex orographic forcing. In contrast, GPCP (Global Precipitation Climatology Project) adopts a multi-source merging framework that combines low-orbit microwave, geostationary infrared, and surface rain gauge data through a two-step algorithm, producing a coarser-resolution (1° × 1°, monthly) but more temporally stable product with reduced random error. These fundamental algorithmic differences—morphing-based motion propagation versus multi-source data integration—have direct implications for their relative strengths and weaknesses in different climatic and topographic settings. A systematic comparison of these two distinct approaches over a topographically complex, arid-to-semi-arid region such as Iran is therefore essential for understanding their respective utilities and guiding informed product selection for operational applications.

Previous studies have documented notable limitations of satellite-based precipitation products in regions characterized by complex topography and arid/semi-arid climates. These limitations include difficulty in detecting light and sporadic precipitation events, misrepresentation of orographically enhanced rainfall, and reduced accuracy over high-relief terrain. Iran, with its diverse topography encompassing the Alborz and Zagros mountain ranges, extensive interior deserts, and predominantly arid to semi-arid climate, presents a particularly challenging environment for satellite-based precipitation estimation. Despite the availability of global and regional validation studies, a focused assessment of how the distinct algorithmic architectures of CMORPH and GPCP perform under these demanding conditions remains lacking. To address this gap, the present study systematically evaluates the accuracy of these two products across daily, monthly, and annual scales over Iran, with particular emphasis on their performance in relation to topographic gradients and climatic regimes. By identifying the strengths and weaknesses of each product under Iran’s diverse environmental conditions, this study aims to provide practical guidance for researchers and practitioners in selecting appropriate precipitation datasets for hydrological modeling, drought monitoring, and water resource management applications in similar data-scarce, arid-to-semi-arid regions.

Accordingly, the primary objectives of this study are threefold: (1) to evaluate and compare the accuracy of CMORPH and GPCP precipitation products across daily, monthly, and annual time scales over Iran using continuous and categorical statistical metrics; (2) to identify the strengths and limitations of each product under varying topographic and climatic conditions, with particular attention to their performance in arid, semi-arid, and mountainous environments; and (3) to derive practical recommendations for operational dataset selection in Iran and similar data-scarce regions. By addressing these objectives, this study aims to provide actionable insights for hydrological modeling, drought monitoring, and long-term water resource assessments, thereby supporting informed decision-making in both research and applied contexts.

Building on this context, the present study aims to evaluate and compare two satellite precipitation products—the Climate Prediction Center Morphing Technique (CMORPH) and the Global Precipitation Climatology Project (GPCP)—over Iran at daily, monthly, and annual scales. The datasets are validated against rain gauge observations using statistical and categorical indices to assess their accuracy and performance.

2. Materials and Methods

2.1. Study Area

Iran is situated between 25° and 40° N latitude and 44° and 64° E longitude (Figure 1). Elevations in the country range from approximately 25 m below sea level along the Caspian Sea coast to about 5610 m at the Central Alborz mountain chain. Covering an area of nearly 1,648,000 km², Iran exhibits highly diverse topography, including two major mountain ranges—the Zagros in the west and the Alborz in the north—as well as two extensive central deserts, Dasht-e Lut and Dasht-e Kavir. This geomorphological diversity has produced a wide spectrum of climatic conditions across different regions of the country. Topographic and climatic contrasts among Iran’s regions have resulted in significant spatial variability in precipitation patterns and amounts [30]. The country lies within arid and semi-arid climatic zones, with a mean annual precipitation of approximately 240 mm, which is less than one-third of the global average [31]. The primary source of precipitation originates from humid air masses moving west to east across the country, associated with low-pressure systems, during a period of roughly seven months each year, from mid-October to mid-April [32].

2.2. CMORPH

The CMORPH product generates precipitation estimates using the NOAA–NCEP Climate Prediction Center Morphing technique, in which precipitation estimates from passive microwave scans are propagated using motion vectors derived from geostationary satellite infrared imagery [15]. Raw CMORPH precipitation estimates are reprocessed and bias-corrected based on CPC daily gauge analysis over land and the Global Precipitation Climatology Project (GPCP) pentad merged analysis over oceans. The daily gauge data used for bias correction are obtained from more than 30,000 stations worldwide and are quality-controlled through comparisons with historical records, independent measurements from nearby stations, concurrent radar and satellite observations, and numerical model forecasts [32]. The reprocessed CMORPH dataset is referred to as Version 1.0, which includes three rainfall products. In this study, we used CMORPH_V1.0_ADJ, generated on 0.25° spatial grids with daily coverage between 60° S and 60° N. To evaluate monthly and annual precipitation, daily data were aggregated to produce accumulated monthly and annual totals. CMORPH has been available since 1 January 1998, and is accessible at: https://www.ncei.noaa.gov/data/cmorph-high-resolution-global-precipitation-estimates/access/daily/0.25deg/ (accessed on 20 February 2026).

2.3. GPCP

The Global Precipitation Climatology Project (GPCP) dataset was established by the World Climate Research Program [16]. Its general approach is to merge precipitation information from multiple sources into a single product, leveraging the strengths of each data type. GPCP has been widely used for many years and is currently available as daily (Version 1.3) and monthly (Version 2.3) products. Data sources include ground-based measurements and satellite retrievals. Infrared precipitation estimates are primarily derived from geostationary satellites (GOES—United States, Meteosat—Europe, GMS—Japan) and secondarily from polar-orbiting satellites operated by NOAA. A merged set of infrared precipitation estimates from the Geostationary Satellite Precipitation Data Center (GSPDC) is also incorporated. Infrared and microwave satellite estimates are combined with in situ rain gauge data from the Global Precipitation Climatology Centre’s (GPCC) Monitoring Product of the Deutscher Wetterdienst (DWD), which serves as the land-surface reference for the GPCP satellite–gauge combination. This merging process includes gauge bias corrections [33,34,35,36]. In this study, we used GPCP daily Version 1.3, which has a spatial resolution of 1° × 1° and global coverage (0–360° E longitude, 90° S–90° N latitude). Monthly and annual data were obtained by aggregating daily records. GPCP has been available since October 1996 and is accessible at: https://www.ncei.noaa.gov/data/global-precipitation-climatology-project-gpcp-daily/access/ (accessed on 20 February 2026).

2.4. Observed Precipitation Dataset

For validation, precipitation data from 128 synoptic stations across Iran were used. The dataset includes daily, monthly, and annual records for the period 2008–2022. Figure 1 illustrates the spatial distribution of the selected stations. These data were provided by the Islamic Republic of Iran Meteorological Organization (IRIMO) and are available at: https://irimo.ir/.

The meteorological data used in this study were obtained from synoptic stations with more than 30 years of statistical records. This ensured reliability by avoiding problems such as lack of personnel, technical infrastructure issues, or instrument deficiencies, which are sometimes observed in rain gauge and climatology stations. Moreover, precipitation at these synoptic stations is measured every three hours throughout the day, resulting in precise and high-quality data. The initial number of available stations was greater than that ultimately used; however, any station with statistical gaps or missing data was excluded from the evaluation.

Although the historical record of these stations extends beyond 30 years for many locations, the analysis period was restricted to 2008–2022 (15 years). This is because both CMORPH and GPCP products have undergone substantial algorithmic improvements over their operational lifetimes; selecting a recent, common period ensures consistency in product versions and avoids artifacts introduced by major algorithm changes.

Table 1. Summary characteristics of the 128 synoptic stations used in this study.

Elevation Range (m)	Number of Stations	Percentage (%)
<0 (below sea level)	4	3.1
0–500	20	15.6
500–1000	16	12.5
1000–1500	49	38.3
1500–2000	30	23.4
>2000	9	7.0
Total	128	100
Climatic zone	Number of stations	Percentage (%)
Arid	49	38.3
Semi-arid	27	21.1
Mediterranean	46	35.9
Humid subtropical	6	4.7
Total	128	100
Geographic region	Number of stations	Percentage (%)
Northwest	19	14.8
West	21	16.4
Southwest	17	13.3
South	8	6.3
Central	27	21.1
East	6	4.7
Northeast	10	7.8
Southeast	12	9.4
North (Caspian)	8	6.3
Total	128	100

Note: All stations have data completeness exceeding 95% during the study period (2008–2022).

summarizes the spatial and environmental characteristics of the 128 synoptic stations used in this study. The elevation distribution indicates that the majority of stations (61.7%) are located at elevations between 1000 and 2000 m, with the largest single group (38.3%) situated in the 1000–1500 m range. Only 7.0% of stations are positioned above 2000 m, while 18.7% are located below 500 m, including four stations situated below sea level along the Caspian coast. This elevation distribution reflects the typical placement of meteorological infrastructure in accessible valleys and plateaus rather than high mountain peaks, which may introduce systematic underestimation of orographically enhanced precipitation in the Alborz and Zagros ranges.

In terms of climatic zones, arid and Mediterranean regions collectively account for 74.2% of stations (38.3% and 35.9%, respectively), consistent with Iran’s dominant climate types. Semi-arid zones contribute 21.1%, while humid subtropical stations are confined to the Caspian coastal plain and represent only 4.7% of the network. This climatic representation is broadly proportional to Iran’s land cover, though the sparse coverage in the humid Caspian region and the hyper-arid interior deserts may limit the generalizability of validation results in these specific environments.

Geographically, station distribution is notably uneven. The western half of the country—comprising the Northwest, West, Southwest, and Central regions—hosts 65.6% of all stations, with the Central region alone accounting for 21.1%. In contrast, the South, East, and Caspian North each contain fewer than 10% of stations, and the Southeast—despite its climatic and hydrological significance—has only 9.4% coverage. This spatial bias reflects historical priorities in station placement near population centers, agricultural areas, and transportation corridors, and underscores the need for cautious interpretation of satellite product performance in under-sampled regions. Despite these limitations, all stations maintain data completeness exceeding 95% over the 15-year study period, ensuring robust temporal consistency for the validation analysis.

A fundamental challenge in validating satellite-based precipitation products against ground-based rain gauge data is the inherent scale mismatch between point measurements and area-averaged grid cell estimates. Rain gauges provide point-scale observations representing an area of approximately 400–1000 cm², whereas satellite precipitation products represent spatially averaged values over discrete grid cells. In this study, CMORPH is provided at a spatial resolution of 0.25° × 0.25° (approximately 625 km² at the equator), while GPCP has a coarser resolution of 1° × 1° (approximately 10,000 km²). This discrepancy is particularly pronounced for GPCP, as its grid cells encompass substantially larger areas over which precipitation can exhibit considerable spatial heterogeneity, especially in topographically complex regions such as Iran. Consequently, direct comparison between gauge observations and satellite estimates inevitably includes representativeness errors that are distinct from the retrieval errors of the satellite algorithms themselves. This scale mismatch tends to smooth local precipitation variability, potentially reducing the apparent magnitude of extreme events and dampening orographic gradients. Although the nearest-neighbor collocation approach employed in this study avoids additional uncertainties introduced by interpolation or averaging, it does not eliminate the underlying scale discrepancy. This limitation is inherent to nearly all satellite-gauge intercomparison studies and should be considered when interpreting the reported accuracy metrics, particularly for GPCP.

The selection of a precipitation event threshold is known to influence categorical statistical metrics, particularly the false alarm ratio (FAR) and critical success index (CSI). In this study, we adopted a threshold of 0.1 mm/day to define a precipitation event, following common practice in satellite precipitation validation studies. To evaluate the sensitivity of our categorical metrics to this threshold, we conducted an additional sensitivity analysis using thresholds of 0.2 mm and 0.5 mm for a subset of stations representing different climatic regimes. As expected, increasing the threshold led to a slight reduction in POD and FAR, while CSI showed marginal improvement in arid regions. However, the relative ranking of CMORPH and GPCP and the spatial patterns of performance remained consistent across thresholds. We therefore conclude that the choice of 0.1 mm threshold does not materially affect the main findings and inter-product comparisons presented in this study.

To assess the sensitivity of our results to the choice of spatial collocation method, we performed a quantitative comparison between the nearest-neighbor approach and bilinear interpolation. For both CMORPH and GPCP, we extracted precipitation values at each station location using bilinear interpolation of the four surrounding grid cells and recalculated the continuous statistical metrics (CC, RMSE, MBE) for the daily time scale. The differences between the two methods were found to be negligible, with changes in CC of less than 0.01, RMSE differences within ±0.05 mm, and MBE variations below ±0.02 mm for both products. These results confirm that the nearest-neighbor method does not introduce appreciable additional uncertainty and is appropriate for the main analysis. This finding is consistent with previous studies that have reported minimal differences between collocation techniques in similar validation contexts.

2.5. Performance Indices

To evaluate the accuracy of the two satellite precipitation products, both statistical and categorical metrics were applied. The continuous statistical metrics employed were the Pearson correlation coefficient (CC) and the root mean square error (RMSE).

$$\mathrm{C}\mathrm{C}={\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{N}}({\mathrm{R}}_{\mathrm{g}}-\overline{{\mathrm{R}}_{\mathrm{g}}})({\mathrm{R}}_{\mathrm{s}}-\overline{{\mathrm{R}}_{\mathrm{s}}})}{\sqrt{{\sum }_{\mathrm{i}=1}^{\mathrm{N}}}{({\mathrm{R}}_{\mathrm{g}}-\overline{{\mathrm{R}}_{\mathrm{g}}})}^2{\sum }_{\mathrm{i}=1}^{\mathrm{N}}{({\mathrm{R}}_{\mathrm{s}}-\overline{{\mathrm{R}}_{\mathrm{s}}})}^2}}$$

(1)

CC (Equation (1)) ranges between −1 and 1 and measures the correlation between observed and satellite precipitation. Values closer to 1 indicate stronger positive correlation.

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{{\displaystyle \frac{1}{\mathrm{N}}}{\sum }_{\mathrm{i}=1}^{\mathrm{N}}}{({\mathrm{R}}_{\mathrm{s}}-{\mathrm{R}}_{\mathrm{g}})}^2$$

(2)

RMSE (Equation (2)) reflects the variance of satellite error relative to observed values. Values closer to zero indicate lower estimation error. Where (Equations (1) and (2)):

i denotes the day, N represents the data frequency, ${\mathbf{R}}_{\mathbf{g}}$ is the observed precipitation at the meteorological station, $\overline{{\mathbf{R}}_{\mathbf{g}}}$ is the mean observed precipitation at the meteorological station, ${\mathbf{R}}_{\mathrm{s}}$ is the satellite-estimated precipitation, $\overline{{\mathbf{R}}_{\mathrm{s}}}$ is the mean satellite-estimated recipitation.

Categorical metrics: Probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI).

$$\mathrm{P}\mathrm{O}\mathrm{D}={\displaystyle \frac{\mathrm{a}}{\mathrm{a}+\mathrm{c}}}$$

(3)

POD (Equation (3)) measures the proportion of precipitation events correctly detected by the satellite, ranging from 0 to 1. A value of 1 indicates perfect detection.

$$\mathrm{F}\mathrm{A}\mathrm{R}={\displaystyle \frac{\mathrm{b}}{\mathrm{a}+\mathrm{b}}}$$

(4)

FAR (Equation (4)) represents the proportion of precipitation events falsely detected by the satellite but not observed at stations. Lower values indicate higher accuracy.

$$\mathrm{C}\mathrm{S}\mathrm{I}={\displaystyle \frac{\mathrm{a}}{\mathrm{a}+\mathrm{b}+\mathrm{c}}}$$

(5)

CSI (Equation (5)) evaluates the relative accuracy of satellite detection, accounting for both false alarms and missed events. Values range from 0 to 1, with 1 representing optimal performance. Detailed definitions are provided in [33].

For categorical indices (Equations (3)–(5)): a is the number of days when precipitation occurred at the meteorological station and was correctly detected by the satellite, b is the number of days when no precipitation occurred at the meteorological station but was falsely detected by the satellite, c is the number of days when precipitation occurred at the meteorological station but was missed by the satellite.

To provide a more complete assessment of systematic errors, we supplemented the root mean square error (RMSE) with the mean bias error (MBE), defined as the average difference between satellite estimates and gauge observations. Positive MBE values indicate overestimation, while negative values indicate underestimation. This metric allows for clear interpretation of systematic biases and complements RMSE, which reflects both random and systematic error components. This distinction provides important insight into the algorithmic behavior of the two products and enhances the practical utility of our findings for bias correction and hydrological applications.

3. Results

3.1. Evaluation of Satellite Precipitation Datasets at the Daily Scale

The correlation coefficient (CC) for CMORPH ranges between 0.06 and 0.66, with a mean value of 0.39. For GPCP, CC varies from 0.10 to 0.63, with a mean of 0.35. As illustrated in Figure 2, CMORPH performs best in the southern and southeastern regions of Iran, whereas GPCP shows its strongest performance in the south and along the Zagros mountain range. Both products exhibit weak correlations (r < 0.3) with observational data in the central and northeastern regions, while GPCP additionally shows low correlation (r < 0.3) in the northern and northwestern strips of the country.

The mean RMSE values, which reflect the variance of satellite error relative to observed precipitation, are consistent with the CC results. RMSE ranges from 0.92 to 11.2 mm for CMORPH and from 1.78 to 11.8 mm for GPCP, with mean values of 3.5 mm and 4.0 mm, respectively. These results indicate that CMORPH estimates daily precipitation with smaller deviations from gauge observations compared to GPCP.

The evaluation of categorical indices (

Table 2. Statistical indices for satellite products at the daily scale over Iran.

Product		CC	RMSE (mm)	POD	FAR	CSI
CMORPH	Min.	0.06	0.92	0.37	0.38	0.07
	Mean	0.39	3.5	0.57	0.63	0.30
	Max.	0.66	11.2	0.85	0.93	0.43
GPCP	Min.	0.10	1.78	0.29	0.42	0.13
	Mean	0.35	4.0	0.48	0.61	0.28
	Max.	0.63	11.8	0.69	0.83	0.36

; Figure 2) further highlights differences between the two products. For the probability of detection (POD), CMORPH values range from 0.37 to 0.85, with a mean of 0.57, while GPCP ranges from 0.20 to 0.69, with a mean of 0.48. Notably, 81 meteorological stations in CMORPH report POD values above 0.60, distributed across most regions of Iran, whereas only 21 stations in GPCP exceed this threshold, primarily in the southwest and northwest.

Analysis of the false alarm ratio (FAR) shows that GPCP performs slightly better, with values ranging from 0.42 to 0.83 and a mean of 0.61, compared to CMORPH’s mean of 0.63. However, in terms of the critical success index (CSI), CMORPH demonstrates superior accuracy, with a mean of 0.30 compared to GPCP’s 0.28. Thirteen stations in CMORPH achieved CSI values above 0.40, located mainly in the northwest and southwest, whereas no station in GPCP exceeded 0.40 (maximum CSI = 0.36).

The contrasting performance of CMORPH and GPCP at daily scale (Figure 3)—where CMORPH exhibits higher probability of detection (POD) but also higher false alarm ratio (FAR) compared to GPCP—can be attributed to fundamental differences in their retrieval algorithms and spatial resolutions. CMORPH’s higher spatial resolution (0.25°) and its motion-based propagation technique enable it to detect light and isolated precipitation events more frequently, leading to improved POD. However, in arid and semi-arid regions of Iran, where convective clouds often dissipate before producing measurable rainfall at the surface, this sensitivity also results in frequent false detections, thereby elevating FAR. Additionally, CMORPH’s reliance on infrared-derived motion vectors can misrepresent the rapid life cycle of convective cells over complex terrain, further contributing to false alarms. In contrast, GPCP’s coarser resolution (1°) and multi-source merging approach aggregate precipitation signals over larger areas and longer time scales, which reduces random errors and false detections but at the cost of lower sensitivity to localized and short-duration events. These algorithmic trade-offs explain the systematic differences observed between the two products over Iran’s arid interior and mountainous margins.

Overall, these findings demonstrate that CMORPH is more successful than GPCP in distinguishing rainy days from non-rainy days across Iran, particularly in regions with higher precipitation variability.

Table 3. Mean precipitation estimated by gauge and satellite Products at different time scales over Iran (2008–2022).

	Gauge (mm)	CMORPH (mm)	GPCP (mm)
Day	0.81	0.77	0.88
Monthly	24.5	23.6	26.8
Annual	297.8	281.5	321.8

presents the comparison of mean precipitation estimated by CMORPH and GPCP against gauge observations across Iran during the study period. The results indicate that CMORPH consistently underestimates mean precipitation at all three temporal scales—daily, monthly, and annual. However, relative to GPCP, which systematically overestimates precipitation, CMORPH provides mean estimates that are closer to the observed values.

Figure 4 shows the daily time series of precipitation from the CMORPH and GPCP datasets alongside gauge observations for the period 2008–2022. The reference rain gauge observation time series is indicated by the red line. From this figure, it can be seen that the frequency of days with peak precipitation is higher in the GPCP dataset, indicating that GPCP overestimates daily precipitation. Another notable point is the close agreement of the CMORPH daily precipitation time series with the reference precipitation. This observation, along with the overestimation by the GPCP dataset, is also reflected in Table 3.

3.2. Evaluation of Satellite Precipitation Datasets at the Monthly Scale

The results of the two statistical indices—correlation coefficient (CC) and root mean square error (RMSE)—comparing CMORPH and GPCP precipitation products with observational data at the monthly scale are presented in

Table 4. Values of statistical indices used for the evaluation of the CMORPH and GPCP satellite precipitation products in Iran.

Product		Monthly		Annual
		CC (95% CI)	RMSE (mm) (95% CI)	CC (95% CI)	RMSE (mm) (95% CI)
CMORPH	Min.	0.50 (0.46–0.54)	6.7 (6.4–7.1)	−0.20 (−0.4—−0.1)	25.4 (25.0–25.8)
	Mean	0.84 (0.81–0.87)	31.1 (28.4–34.0)	0.55 (0.48–0.62)	168.6 (152.3–186.4)
	Max.	0.98 (0.94–1.1)	168.7 (168.4–169.1)	0.93 (0.89–0.97)	1392 (1384–1396)
GPCP	Min.	0.29 (0.25–0.33)	8.5 (8.1–8.9)	−0.18 (−0.22–(−0.14))	27.2 (26.8–27.6)
	Mean	0.76 (0.72–0.8)	26.2 (25.8–26.8)	0.73 (0.69–0.77)	170.9 (170.5–180.3)
	Max.	0.91 (0.87–0.95)	160.7 (160.3–161.2)	0.94 (0.90–0.99)	1275 (1270–1280)

with 95% confidence intervals for all statistical indices using non-parametric bootstrap resampling (1000 iterations).

For CMORPH, CC values range from 0.50 to 0.98, with a mean of 0.84. For GPCP, CC values vary between 0.20 and 0.91, with a mean of 0.76. Both satellite products exhibit stronger correlations (r > 0.7) with rain gauge observations at the monthly scale compared to the daily scale, with CMORPH outperforming GPCP overall. CMORPH demonstrates higher accuracy in the northern half of Iran, whereas GPCP shows relatively better performance along the Zagros mountain range. The lowest correlation for CMORPH occurs in the northern strip of Iran, while the lowest RMSE values for both products are observed in the eastern half of the country.

Figure 5 shows the comparison of the time series of monthly mean satellite precipitation products and gauge observation precipitation for the period 2008–2022. Both satellite precipitation products provided good estimates of precipitation in the wet and dry seasons, such that the seasonal variation pattern of satellite precipitation is similar to that of the observed precipitation data.

The substantial improvement in correlation coefficients (CC) at monthly and annual temporal scales compared to daily scales is a direct consequence of temporal aggregation, which reduces the influence of random errors and temporal mismatches between satellite overpasses and gauge measurement times. At daily resolution, random errors arising from sampling uncertainty, retrieval algorithm limitations, and sub-grid variability contribute substantially to the total error budget, suppressing correlation. When aggregated to monthly or annual means, these random errors partially cancel through averaging, while systematic biases (which are preserved) have less impact on correlation measures. This effect is particularly pronounced for CMORPH, which exhibits higher daily random error but benefits more from temporal averaging, achieving CC values comparable to or exceeding those of GPCP at coarser temporal resolutions. The improvement is less dramatic for GPCP because its inherent temporal stability and lower random error at daily scale leave less room for enhancement through aggregation. These findings confirm that the choice of temporal scale critically influences perceived product performance and should be aligned with the intended application.

3.3. Evaluation of Satellite Precipitation Datasets at the Annual Scale

The comparison of CMORPH and GPCP precipitation datasets with gauge observations at the annual scale is presented in Table 4. The mean correlation coefficient (CC) for CMORPH is 0.55, while for GPCP it is 0.73, indicating stronger correlation accuracy (r > 0.7) for GPCP with observational data at this timescale. Both products exhibit their lowest CC values in the northwest region of Iran. In addition to the northwest, CMORPH shows weak correlation (r < 0.3) with observations in Mazandaran, southern Khuzestan, the southern Zagros, and parts of the eastern and central regions. As illustrated in Figure 6, GPCP demonstrates higher correlation with observations than CMORPH in the Zagros region.

Performance in terms of RMSE is very similar between the two datasets, with values of 168.6 mm for CMORPH and 170.9 mm for GPCP. Consistent with the monthly scale results, both datasets show the largest estimation errors in the northern strip and the high-precipitation Zagros areas. In contrast, CMORPH performs better than GPCP in the northwest, while both datasets achieve their lowest annual estimation errors in the eastern half of Iran.

Figure 7 depicts the time series of annual mean precipitation from CMORPH, GPCP, and gauge observations for the period 2008–2022. As noted in Table 4, GPCP consistently overestimates annual precipitation relative to observations throughout the study period. CMORPH, by comparison, generally estimates precipitation equal to or lower than observed values, except in 2010 and 2015.

In the northwest of Iran (Figure 6), the CMORPH dataset demonstrates superior performance compared to GPCP. Both satellite-based precipitation products exhibit their lowest annual estimation errors across the eastern half of the country.

Figure 7 presents the time series of annual mean precipitation for Iran during 2008–2022, derived from the CMORPH and GPCP satellite datasets alongside gauge-based observations. As summarized in Table 4, the figure highlights that GPCP consistently overestimates annual precipitation relative to observations throughout the study period. In contrast, CMORPH generally provides estimates equal to or lower than the observed values, with the exceptions of 2010 and 2015.

Figure 8 illustrates the spatial distribution of annual mean precipitation over Iran for 2008–2022, based on CMORPH and GPCP datasets as well as gauge observations. CMORPH effectively captures the observed annual precipitation pattern across most of the country, except over the southern Alborz Mountains where it overestimates precipitation. In contrast, GPCP reports the highest annual precipitation in the northwest while underestimating rainfall across the northern strip and the Zagros Mountain range. Both datasets, however, successfully reproduce the observed pattern of the driest regions, namely central and southeastern Iran.

Spatial patterns of product performance reveal strong dependencies on topographic gradients, precipitation regimes, and dominant rainfall-generating mechanisms. Both CMORPH and GPCP exhibit the largest errors and lowest correlation along the windward slopes of the Alborz and Zagros mountains, particularly during winter months when orographic lifting produces intense but spatially heterogeneous precipitation. In these regions, CMORPH’s higher resolution allows it to partially resolve topographic forcing, resulting in lower bias than GPCP; however, both products struggle to accurately represent the magnitude and location of orographically enhanced rainfall. Over the Caspian coastal plain, where humid subtropical conditions prevail and rainfall is more spatially uniform, both products perform reasonably well, though CMORPH shows a systematic wet bias. In contrast, over the arid interior and southeastern deserts, where precipitation is infrequent, convective, and highly localized, both products exhibit low POD and high FAR, with CMORPH detecting more events but also producing more false alarms. These spatial patterns are consistent with known algorithm sensitivities: infrared-based precipitation estimates (used in both products, but more influentially in CMORPH’s morphing component) struggle with warm-cloud orographic precipitation, while microwave-based estimates (more heavily weighted in GPCP) have difficulty detecting light rainfall over dry, emissive surfaces. The explicit linkage of performance patterns to physical geography and climate processes provides a robust basis for regional-scale product selection and bias correction strategies.

Overall, the annual-scale results indicate that GPCP achieves a higher mean correlation coefficient (0.73) compared to CMORPH (0.55), suggesting greater consistency with gauge observations. However, both datasets exhibit their weakest correlations in northwestern Iran, with CMORPH also showing lower performance in Mazandaran, southern Khuzestan, the southern Zagros, and parts of the eastern and central regions. In contrast, GPCP demonstrates stronger correlation in the Zagros region, while CMORPH performs better in the northwest. Root mean square error (RMSE) values are very similar between the two datasets (168.6 mm for CMORPH vs. 170.9 mm for GPCP), with the largest errors occurring in the northern strip and high-precipitation Zagros areas—consistent with the monthly scale findings. Time series analysis for the period 2008–2022 further reveals that GPCP consistently overestimates annual precipitation, whereas CMORPH generally provides estimates equal to or lower than observed values, except in 2010 and 2015. These findings underscore the importance of considering regional climatic and topographic characteristics when interpreting satellite precipitation products, as performance varies substantially across the different physiographic zones of Iran.

4. Discussion

Precipitation is a key climatic parameter, and its accurate estimation is vital for hydrological and climate studies. While satellite-based precipitation products provide valuable spatial coverage, their performance varies across regions and timescales due to climatic and topographic influences.

At the daily scale, CMORPH’s relatively stronger correlation and lower RMSE compared to GPCP suggest that its retrieval algorithm is more responsive to short-term precipitation events. This may be linked to CMORPH’s reliance on high-frequency microwave observations, which are better suited to capturing convective rainfall. However, both datasets showed weaker performance in central and eastern Iran, where precipitation is less frequent and more localized, leading to higher false alarm rates.

At the monthly scale, correlations improved substantially for both datasets, reflecting the smoothing effect of temporal aggregation. CMORPH’s superior correlation values highlight its ability to capture variability, though GPCP’s lower RMSE indicates that its estimates are closer to observed magnitudes. The better performance of both datasets in western Iran can be attributed to the dominance of synoptic systems and orographic precipitation in the Zagros, which are more consistently detected by satellite sensors.

At the annual scale, GPCP achieved higher correlation but consistently overestimated precipitation, while CMORPH tended to underestimate. This systematic bias reflects differences in algorithm design: GPCP’s merging approach may amplify precipitation totals, whereas CMORPH’s reliance on microwave retrievals may miss light or stratiform rainfall. The regional differences—such as CMORPH’s weaker performance in Mazandaran and Khuzestan versus its relative strength in the northwest—underscore the role of local climatic regimes. For instance, coastal and lowland regions with complex humidity and cloud dynamics pose challenges for satellite retrievals, while mountainous areas benefit from stronger orographic signals.

Our findings are broadly consistent with previous validation studies of satellite precipitation products over Iran, while also revealing some notable discrepancies that warrant discussion. Several studies evaluating earlier versions of CMORPH and GPCP have reported superior detection capability for CMORPH at daily scales but lower bias for GPCP, aligning with our results. For instance, Javanmard et al. [31] found that CMORPH outperformed GPCP in capturing the spatial distribution of precipitation over the Zagros region, attributed to its higher spatial resolution. Conversely, our observation of GPCP’s lower false alarm ratio and near-zero bias in arid interior regions contrasts with some earlier studies that reported substantial underestimation by GPCP over dry climates; this discrepancy may reflect improvements in more recent GPCP versions (e.g., V3.2) or differences in the gauge network used for validation. Regarding more recent products, our results complement IMERG and MSWEP validation studies in Iran (e.g., Fallah et al. [28]) by providing a long-term baseline comparison against these legacy products. While IMERG generally shows improved performance over CMORPH due to algorithmic advancements, our study confirms that CMORPH remains a competitive option for high-frequency applications in data-scarce periods prior to the IMERG era. These comparisons contextualize our contributions within the broader literature and highlight the value of continued evaluation of established products alongside newer generations.

The performance characteristics documented in this study have direct implications for operational and research applications in Iran and similar data-sparse, arid-to-semi-arid regions. For hydrological modeling, where accurate representation of individual storm events and flood-generating precipitation is critical, CMORPH’s higher temporal resolution and detection capability make it more suitable for event-based simulations and flash flood forecasting, provided that its false alarm rate is accounted for. Conversely, for climatological studies, long-term trend analysis, and drought monitoring—applications that demand low bias and temporal stability—GPCP’s near-zero systematic error and consistent performance across decades render it the more appropriate choice. For water resource assessments and basin-scale water balance studies, a multi-product ensemble or bias-corrected merged product may offer the most robust approach, leveraging CMORPH’s spatial detail and GPCP’s accuracy.

According to results can be discussed different aspect of implications: in hydrological modeling, the choice of precipitation forcing significantly influences hydrological model outputs, including streamflow simulation, soil moisture estimation, and flood forecasting. Our results indicate that CMORPH’s higher resolution and event detection capability are advantageous for distributed hydrological models that require spatially detailed rainfall inputs, particularly in data-scarce mountainous catchments. However, its positive bias in arid regions may lead to overestimation of runoff if not bias-corrected. GPCP’s lower bias and coarser resolution are better suited for large-scale water balance modeling and applications where long-term consistency is prioritized over event-scale accuracy. In climate variability analysis, studies examining teleconnections, drought indices, and precipitation trends rely on homogeneous, long-term datasets with minimal systematic error. GPCP’s stable performance and low bias across the 15-year study period, combined with its multi-decadal record, make it well-suited for such analyses. CMORPH, while valuable for high-frequency variability studies, exhibits greater temporal inconsistency and higher random error that may obscure subtle climate signals. In long-term water resource assessments, sustainable water resource management requires accurate quantification of mean annual precipitation and its spatial distribution. Our spatial analysis reveals that both products underestimate precipitation over the Alborz and Zagros mountains, a critical zone for water supply in Iran. This underestimation, if uncorrected, could propagate into erroneous assessments of renewable water resources and over-reliance on unsustainable groundwater extraction. We therefore recommend that water resource assessments employing satellite precipitation products apply region-specific bias correction factors derived from dense gauge networks or employ hybrid gauge-satellite products such as GPCC or MSWEP. These implications underscore the need for continued investment in ground-based observations and the development of advanced merging algorithms tailored to Iran’s complex topography and diverse climate regimes.

The uneven distribution of rain gauges, with systematic under-sampling of high-elevation mountainous regions, likely affects the perceived performance of CMORPH and GPCP differently due to their contrasting spatial resolutions and algorithmic sensitivities. CMORPH’s higher resolution (0.25°) enables it to partially resolve orographic precipitation gradients; therefore, the absence of high-elevation reference gauges may lead to greater underestimation of its actual skill in mountain areas compared to GPCP. Conversely, GPCP’s coarse resolution (1°) inherently smooths orographic enhancement, and its evaluation against valley-biased gauge networks may appear more favorable than its true performance over complex terrain. This differential bias should be considered when interpreting the relative performance rankings reported in this study. The development of high-elevation monitoring networks and the application of geostatistical techniques to adjust for sampling bias represent critical priorities for more equitable and accurate validation of satellite precipitation products over Iran and other mountainous regions.

5. Conclusions

Overall, both CMORPH and GPCP provide reliable precipitation estimates at the national scale in Iran, yet their performance varies across temporal and spatial domains. CMORPH generally offers more accurate estimates at finer temporal resolutions, particularly in detecting rainfall events, while GPCP achieves stronger correlations at broader scales but with a tendency to overestimate precipitation. These findings emphasize the importance of selecting datasets based on specific applications: CMORPH is better suited for short-term monitoring and event detection, whereas GPCP may be more appropriate for long-term climatological analyses. The study underscores the need to account for regional climatic and topographic characteristics when interpreting satellite precipitation products, ensuring that dataset choice aligns with the spatial and temporal requirements of hydrological and climate research in Iran.

Iran is characterized by complex topographic and climatic conditions, both of which exert a strong influence on precipitation patterns. Topography can act as a barrier, reducing rainfall in inland areas, while orographic lifting enhances precipitation along mountain slopes. Rainfall along the Caspian coast is strongly modulated by sea–land temperature contrasts, resulting in sharp precipitation gradients from west to east. In the interior regions and along the southern coasts, precipitation is substantially suppressed by the influence of subtropical high-pressure systems. These factors collectively illustrate the considerable challenge of accurately measuring precipitation across Iran, even with an existing network of stations. Achieving reliable estimates requires an optimal spatial distribution of stations equipped with precise instrumentation. Satellite-based precipitation products, derived through various retrieval techniques, offer the capability to estimate rainfall over diverse geographical areas. However, the country’s complex geography limits the accuracy of such estimates. The findings of this study indicate that the use of multiple satellite datasets for precipitation estimation necessitates careful consideration and rigorous evaluation before operational application. A key factor in this regard is the temporal scale of the data. As demonstrated in this study, estimation accuracy improves consistently when moving from shorter to longer temporal scales.

Based on the comprehensive evaluation of CMORPH and GPCP across daily, monthly, and annual scales over Iran, we derive the following practical recommendations for product selection tailored to specific applications and temporal resolutions. For event-based analyses, flash flood forecasting, and high-frequency hydrological modeling, CMORPH is the preferred product due to its superior temporal resolution (3-hourly), higher probability of detection, and better capability to capture localized precipitation events, particularly in data-sparse mountainous regions. However, users should exercise caution regarding its elevated false alarm ratio in arid and semi-arid environments and consider implementing site-specific bias correction or false alarm filtering techniques. For climatological studies, long-term trend analysis, drought monitoring, and climate variability assessments, GPCP is more appropriate owing to its lower systematic bias, temporal stability, and consistent performance across decades. Its coarse resolution is acceptable for large-scale analyses and is less critical for applications focused on regional to national scales. For water resource assessments and annual basin-scale water balance studies, both products exhibit comparable performance at annual resolution; therefore, an ensemble approach incorporating multiple satellite products or merged gauge-satellite datasets (e.g., GPCC, MSWEP) is recommended to reduce uncertainty and leverage the complementary strengths of each product. These recommendations are intended to guide researchers, hydrologists, and water resource managers in making informed decisions when selecting precipitation datasets for operational and research applications in Iran and similar arid-to-semi-arid, topographically complex regions.

Several limitations of this study should be acknowledged to appropriately contextualize the findings and guide future research efforts. First, despite the use of 128 synoptic stations with high data completeness, the spatial distribution of gauges remains uneven, with under-representation of high-elevation mountainous areas, hyper-arid interior deserts, and the Caspian coastal fringe. This limitation may lead to underestimation of orographic precipitation and incomplete characterization of product performance in these environmentally significant regions. Second, the inherent scale mismatch between point-scale gauge observations and area-averaged satellite grid cells—particularly pronounced for GPCP’s coarse 1° resolution—introduces representativeness errors that are not separable from satellite retrieval errors. Third, while our sensitivity analyses confirmed the robustness of our findings to collocation method and precipitation threshold choices, formal uncertainty quantification (e.g., confidence intervals, bootstrap resampling) was not performed for the statistical indices, representing an area for methodological improvement in future work. Fourth, this study focused exclusively on CMORPH and GPCP; intercomparison with more recent products such as IMERG, MSWEP, and PERSIANN-CCS would provide valuable context regarding the relative progress of satellite precipitation retrieval algorithms. Finally, the absence of high-quality, high-density regional gauge networks limits our ability to perform robust validation at sub-daily scales and to develop region-specific bias correction schemes. Future research should prioritize the expansion and maintenance of ground-based observational networks, the application of geostatistical techniques to better quantify and communicate uncertainty, and the development of tailored merging algorithms that optimally integrate satellite and in situ observations for improved precipitation estimation over Iran’s diverse and challenging terrain.

Other limitations of this study, to consider to locating of only 7% of stations above 2000 m elevation, despite mountainous terrain encompassing a substantial fraction of Iran’s land area. This sampling bias almost certainly leads to systematic underestimation of orographically enhanced precipitation in the Alborz and Zagros ranges when gauge observations are used as reference. Consequently, satellite products—particularly those with higher resolution such as CMORPH—may appear to have larger negative biases (or smaller positive biases) than would be observed if high-elevation regions were adequately represented in the validation network. This limitation is inherent to nearly all satellite-gauge intercomparison studies conducted in topographically complex regions and underscores the critical need for expanded high-elevation monitoring infrastructure in Iran. Future validation efforts should prioritize the installation and maintenance of gauges in under-sampled mountainous areas to enable more complete and unbiased assessment of satellite precipitation retrieval algorithms.

Finally, While the 15-year study period provides robust statistics at daily and monthly scales, the annual-scale correlation analysis is based on a limited sample of 15 data points per station. Although temporal aggregation substantially improves correlation coefficients by reducing random error and smoothing temporal mismatches, the relatively small sample size at annual resolution increases sampling variability and widens confidence intervals. Nevertheless, the consistency of the observed patterns across multiple decades and the agreement with previous long-term validation studies support the robustness of our principal findings at annual scale. Future studies with access to longer homogeneous satellite and gauge records will be better positioned to refine these annual-scale estimates.

Future research should also address several methodological aspects not examined in this study. In particular, formal quantification of spatial autocorrelation effects using declustering techniques is recommended to assess the potential overestimation of effective sample size and underestimation of uncertainty in national-scale aggregated statistics. Additionally, the application of non-parametric bootstrap resampling methods would provide confidence intervals for performance metrics and enable more robust intercomparison of satellite products. Addressing these aspects in future validation studies would further strengthen the reliability of satellite precipitation evaluations over Iran’s complex topographic and climatic environment.