1. Introduction
Several researchers have presented the comparison of satellite estimates with rain-gauge measurements in various study areas. One of the most common difficulties in such an attempt is the fact that satellite estimates refer to precipitation amounts over areas of a few kilometers around a gauge station, whereas rain-gauges measurements refer to the precipitation recorded in an area of a few centimeters in diameter (point source).
A literature overview of precipitation estimates from satellite, ground-based and gridded data could be found in [
1,
2,
3,
4]. However, recently released IMERG (Integrated Multi-satellite Retrievals) is going to reduce this sampling issue such that Global Precipitation Measurement (GPM) constellation satellites will have improved spatial and temporal resolutions (i.e., 30 min) [
5,
6].
Contrary to other satellites, such as its ancestor Tropical Rainfall Measuring Mission (TRMM), that could not measure light rain or snowfall, GPM-IMERG uses numerous sensors that measure at various frequencies from different satellites to detect both light and heavy rain and also snowfall. GPM expands TRMM’s reach not only in terms of global coverage, but also through sophisticated satellite instrumentation, the inter-calibration of datasets from other microwave radiometers, coordinated merged precipitation datasets, reduced latency for delivering data products, simplified data access, expanded global ground-validation efforts, and integrated user applications [
6]. GPM improves on TRMM’s capabilities, although it has only two instruments versus the five instruments on TRMM. Dual-frequency Precipitation Radar (DPR) and the radiometer called GPM Microwave Imager (GMI) are some of the most advanced ever developed to measure precipitation from space. The DPR is the only dual frequency radar in space and is capable of creating 3-D profiles and intensity estimates of precipitation ranging from rain to snow and ice. The GMI has a greater frequency range than TRMM’s (13 channels versus 9 channels), which allows GPM to measure precipitation intensity and type through all cloud layers using a wider data swath.
Sharifi et al. [
7] compared precipitation estimates of GPM IMERG final run data sets (along with TRMM 3B42 and ERA-Interim) and with meteorological synoptic stations’ data over Iran to evaluate the GPM-based precipitation outputs for one-year period (March 2014 to February 2015). They found that IMERG underestimates precipitation slightly when considering days with observed precipitation. IMERG yields better results for detection of precipitation events on the basis of Probability of Detection (POD), Critical Success Index (CSI) and False Alarm Ratio (FAR) in areas with stratiform and orographic precipitation compared with other satellite/model data sets. In case of precipitation below 15 mm/day IMERG indicated better results, while for heavy precipitation (>15 mm/day) IMERG was far superior to the other products.
Manz et al. [
8] performed validation of the GPM IMERG Day-1 and TRMM TMPA products against rain gauge stations across Ecuador and Peru from March 2014 to August 2015 at different spatial (0.1°, 0.25°) and temporal (1 h, 3 h, daily) scales. They found that IMERG showed better detection and quantitative precipitation intensity estimation ability than TMPA, particularly in the high Andes. Furthermore, IMERG showed better characterization of gauge observations after separating precipitation detection and precipitation rate estimation.
Prakash et al. [
9] presented an assessment of IMERG, TMPA and Global Satellite Mapping of Precipitation (GSMaP, v.6) estimates against gauge-based observations over India at a daily scale for the southwest monsoon season (June to September 2014). Results indicated that the IMERG product represented well the mean monsoon precipitation and its variability, performing a clear improvement in missed and false precipitation bias over India. However, in areas that precipitation is dominated by orographic effects (northeast India) and over rain-shadow regions (southeast peninsular India) all three satellite-based precipitation estimates did not perform well (in terms of correlation coefficient, RMSE, bias).
Sungmin et al. [
10] presented an analysis of the IMERG (v.3) Early, Late, and Final (IMERG-E, IMERG-L, and IMERG-F) half-hourly precipitation estimates against gauge-based gridded precipitation data from a high-density climate station network in southeastern Austria for the period of April–October in 2014 and 2015. Results showed that IMERG-F precipitation estimates were in the best overall agreement with the gauge-based gridded precipitation data, followed by IMERG-L and IMERG-E estimates, particularly for the hot season.
Tan and Duan [
11] evaluated GPM IMERG and TRMM Multisatellite Precipitation Analysis (TMPA) 3B42 and 3B42RT precipitation data over Singapore on daily, monthly, seasonal and annual scales from April 2014 to January 2016. They found that both IMERG and TMPA products were well correlated with gauges data at monthly scale, but moderately at daily scale (IMERG had the lowest systematic bias). However, IMERG had better performance in the characterization of spatial precipitation variability and precipitation detection capability compared to the TMPA products. All satellite precipitation products overestimated moderate precipitation events (1–20 mm/day) and underestimated light (0.1–1 mm/day) and heavy (>20 mm/day) precipitation events. They concluded that in overall, IMERG showed a slight improvement compared to the TMPA products over Singapore.
Asong et al. [
12] evaluated GPM IMERG (V03 Final Run) data against ground-based measurements at the 6-hourly, daily, and monthly time scales over different terrestrial ecozones of southern Canada during March 2014 to January 2016. They found that considering the mean daily precipitation IMERG and ground-based records showed similar regional variations, while IMERG was overestimated during higher monthly precipitation amounts. Regarding daily and 6-hourly time scales IMERG and ground-based records found to be in agreement for the months of June–September, unlike months November–March. Considering precipitation extremes, it was found that generally IMERG captured well the distribution of heavy precipitation amounts, although correlation was spatially varied.
Tang et al. [
13] evaluated GPM IMERG Day-1 product over Mainland China from April to December 2014 (hourly data) against TRMM TMPA (3B42V7) product (3-hourly and daily data) data and hourly ground-based observations. They found that Day-1 IMERG showed better performance than 3B42V7 at both sub-daily and daily timescales, with significant differences between the two products noticed at the sub-daily resolution; IMERG especially performed better at the mid- and high-latitudes and at relatively dry climate regions; Regarding ability of capturing precipitation intensity, IMERG reproduced better the probability density function (PDF), especially in the low ranges.
Tian et al. [
14] investigates the dependency of the evaluation of the GPM IMERG precipitation product on the gauge density of a ground-based rain gauge network as well as precipitation intensity over five sub-regions in mainland China during two warm seasons (April–October 2014 and 2015). They found that the denser the rain gauge network the better the evaluation metrics. In addition, IMERG tends to overestimate (underestimate) light (heavy) precipitation events.
Xu et al. [
15] evaluated two satellite precipitation products (GPM IMERG and TRMM 3B42V7) in southern Tibetan Plateau region focusing on the effects of topography and precipitation intensity, for the rainy season from May to October 2014. Results indicated that GPM product presented better ability in detecting daily precipitation accumulation at all spatial scales and elevation ranges. It also tends to underestimate the amount of light rain (0–1 mm/d) events, while showing better detecting ability for light precipitation (0–5 mm/d) events; however, the detecting ability at high-elevation (>4500 m) regions was not defined. In terms of the ability of precipitation detection, false alarm ratio (FAR) of GPM (14%) was lower than that of TRMM (21%), while missing ratio of GPM (13%) was larger than that of TRMM (9%).
Dezfuli et al. [
16] evaluated GPM IMERG (V04A) precipitation data using in situ observations from selected stations in West and East Africa, along with some comparisons between IMERG and TMPA. Results indicated temporal (by season) and spatial (by region) variation of correlation between the satellite-based products and stations data. They also found that precipitation diurnal cycle is relatively better captured by IMERG than TMPA, while both products demonstrated better agreement with stations data in East Africa and humid West Africa than in the southern Sahel. Regarding annual cycle, differences were found during the short rains in East Africa. A direct comparison between IMERG and TMPA over the entire continent was also established concluding that IMERG and TMPA presented large differences over mountainous regions.
Liu [
17] commented on the comparison between the IMERG Final Run monthly product and the TMPA monthly product (3B43) in the boreal summer of 2014 and the boreal winter of 2014/15 on a global scale. It was found that IMERG monthly product could capture major heavy precipitation regions in the Northern and Southern Hemispheres reasonably well. However, differences between IMERG and 3B43 varied with surface types and precipitation rates in both seasons, with systematic differences found much smaller over land. When low precipitation rates were considered positive relative differences (IMERG > 3B43) were noticed in contrast to high precipitation rates (IMERG < 3B43). Over ocean, under all precipitation rates negative differences were found. Tan and Santo [
18] studied the performance of six satellite precipitation products (SPPs), GPM IMERG products (IMERG-E, IMERG-L and IMERG-F), TMPA 3B42 and 3B42RT products and the PERSIANN-CDR product, against precipitation gauges for an almost two years period (March 2014 to February 2016), over Malaysia. Results indicated that all the SPPs performed well in annual and monthly precipitation measurements. However, moderate correlation was found at daily precipitation estimations.
Regarding precipitation detection ability, all the SPPs underestimated light (0–1 mm/day) and extreme (>50 mm/day) precipitation events, while overestimated moderate and heavy (1–50 mm/day) precipitation events. The IMERG demonstrated better ability in detecting light precipitation (0–1 mm/day) compared to the other SPPs.
In the literature, only a few studies with evaluation of satellite precipitation products over Cyprus are reported based on gauge observations and a few satellite precipitation products.
Retalis et al. [
19] performed an analysis of precipitation data from satellite data TRMM 3B43 (versions 7 and 7A) over Cyprus and compared them with the corresponding gauge observations and E-OBS gridded data for a 15-year period (1998–2012). They found that TRMM and E-OBS estimates were well fitted with the gauge data, although they were both influenced by the variation of elevation.
A study for a 30-year period (1981–2010) for the precipitation database Climate Hazards Group Infrared Precipitation with Station data (CHIRPS) was performed over the island of Cyprus [
20]. CHIRPS database was evaluated against the dense and reliable network of rain gauges. Results showed good correlation between CHIRPS values and recorded precipitation; although an overestimation, in the CHIRPS values, of the in-situ precipitation data has been noted during the last decade.
Katsanos et al. [
21] in their study calculated a number of climatic indices for the CHIRPS precipitation dataset and records from rain gauges on high spatial (0.05°) and temporal (daily) resolution for a period of 30 years (1981–2010). Results indicates a quite promising performance regarding indices related to daily precipitation thresholds, resulting in high correlation scores. However, for indices referring to number of days, results showed medium or no correlation, probably due to the criteria used for the identification of a wet (rainy) day on the CHIRPS dataset.
Retalis et al. [
22] proposed a novel downscaling method of TRMM 3B42 product in order to attain high resolution (1 km × 1 km) precipitation datasets, by correlating the CHIRPS dataset with elevation information and the normalized difference vegetation index (NDVI) from satellite images at 1 km × 1 km, utilizing artificial neural network models. The resulted product was validated with precipitation measurements from the corresponding rain gauge network of the Cyprus Department of Meteorology. It was found that the new product was able to capture the spatial precipitation variability, especially in the mountainous areas. Its overall accuracy in the annual products was considered rather significant, while the monthly product, although characterized by variability in the range of correlation (r) values, indicated rather significant values when mean monthly products were considered.
Katsanos et al. [
23] analyzed the daily precipitation (rain gauges) database for the island of Cyprus for period of 50 years. They calculated a number of climatic indices for precipitation, including the Standardized Precipitation Index (SPI for the recent 30-year period 1981–2010 and compared the results with those of the period 1961–1990. The results showed an increase in the occurrence of heavy precipitation events and a slight decrease in extreme drought events.
The aim of this study is to assess the accuracy of the new GPM IMERG satellite precipitation product over Cyprus on a monthly basis and during extreme events. To our knowledge, no such report exists yet to study the GPM constellation satellites data over Cyprus.
4. Conclusions
This study, as the title indicates clearly, is a first attempt to validate a specific GPM precipitation product by using in situ surface observations over the island of Cyprus. In comparing the two datasets, neither an upscaling of the rain-gauge measurements to the GPM grid nor an interpolation of the GPM measurements to the station level was attempted. The satellite product was validated against the gauge data by simply comparing the former with the latter within each satellite cell. The limitation of the data span is not encouraging more rigorous validation approaches. Upscaling the rain-gauge measurements or interpolating the satellite product to the station sites could be an alternative as more data become available but in such a case associated issues should also be considered [
26].
An evaluation of the IMERG product in daily and monthly scales against rain-gauges from the dense and reliable network operated by the Cyprus Department of Meteorology in Cyprus from April 2014 to February 2017 was performed. Based on the statistical analysis presented, it was concluded that, in overall, a very good agreement on monthly level was established, although IMERG tend to be underestimated as elevation goes higher, especially for areas higher than 1000 m elevation. Regarding daily evaluation, in terms of extreme recorded events, it seems that there exists a case dependency, while no specific correlation with elevation was established.
In summary, that accurate estimation by satellite-based precipitation products remains still a challenge. However, form this preliminary assessment on the IMERG product, it is highlighted that it could be useful for use in hydrological studies of small basins and flood studies due to its finer spatial and temporal resolution than TMPA.
The research in the field is still open. The authors aim to further evaluate the product, while implementing the new released version of TMPA product to the spatial resolution of IMERG.