ATMOS-03421; No of Pages 6

Atmospheric Research xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Atmospheric Research

journal homepage: www.elsevier.com/locate/atmos

Validation of a high-resolution precipitation database (CHIRPS) over Cyprus for a

30-year period
Dimitrios Katsanos a,⁎, Adrianos Retalis a, Silas Michaelides b
a
National Observatory of Athens, Greece
b
Department of Meteorology, Nicosia, Cyprus

a r t i c l e i n f o a b s t r a c t

Article history: A study for a 30-year period (1981–2010) for a new precipitation database is performed over the island of Cyprus.
Received 29 January 2015 Climate Hazards Group Infrared Precipitation with Station data (CHIRPS) is a more than 30-year quasi-global
Received in revised form 21 May 2015 rainfall dataset, spanning 50°S–50°N (and all longitudes). Starting in 1981 to near-present, CHIRPS incorporates
Accepted 22 May 2015
0.05° resolution satellite imagery with in situ station data to create gridded rainfall time series. In this study,
Available online xxxx
CHIRPS database is firstly compared to other precipitation databases over the Mediterranean basin. In the
Keywords:
following, this study focuses over Cyprus, where a dense and reliable network of rain gauges is available.
CHIRPS database CHIRPS data are compared for the first time with in situ measurements in this area, for the aforementioned 30-
TRMM year period. Monthly and annual comparisons are presented for each of the 0.05 × 0.05 degree cells overlaying
Cyprus climatology the island of Cyprus. Results showed good correlation between CHIRPS values and recorded precipitation,
Mediterranean climatology although an overestimation of the in situ rainfall data has been noted during the last decade.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction (i.e., southeastern Ethiopia, northern Kenya and southern Ethiopia). In

Climate Hazards Group Infrared Precipitation with Station data
(CHIRPS) is a new land-only climatic database of precipitation, made
available since early 2014; it encompasses three diverse types of infor-
mation: global climatologies, satellite estimates and in situ observations.
More specifically, this database incorporates monthly precipitation cli-
matology CHPClim (Climate Hazards Group Precipitation Climatology),
quasi-global geostationary thermal infrared satellite observations, Trop-
ical Rainfall Measuring Mission's (TRMM) 3B42 product, atmospheric
model rainfall fields from NOAA CFS (Climate Forecast System), and
precipitation observations from various sources, including national or
regional Meteorological Services (Funk et al., 2013). Nevertheless, the
key difference of this database with all other existing precipitation data-
bases is the high-resolution of the available data, since the inherent
0.05-degree resolution is a rather unique threshold; the majority of
the available global precipitation datasets have a resolution of 0.5° or
lower (Tapiador et al., 2012), while a mere few of them, like TRMM's
3B43 (Huffman et al., 2007) and E-OBS (Haylock et al., 2008) have a
0.25-degree resolution.
To date, studies based on CHIRPS data are limited in the literature
and those existing are mainly focused on Africa. For example, Shukla
et al. (2014) used CHIRPS data for the primary growing and rainy season
(March–April–May) to evaluate soil moisture forecasts generated by
a seasonal drought forecast system on the equatorial East Africa
⁎ Corresponding author. Tel.: +30 2108109136; fax: +30 2108103236.
E-mail address: katsanos@noa.gr (D. Katsanos).
their study for mapping population vulnerability to climate change in
Africa, López-Carr et al. (2014) used CHIRPS' standardized precipitation
index time series for Africa with regard to the main growing season.
Also, in a study focusing on Eastern Africa, Funk et al. (2013) showed
an almost identical precipitation average for CHIRPS and GPCC (Global
Precipitation Climatology Centre) datasets, regarding springtime
(March–May). On that count, the present study performs for the first
time validation of the CHIRPS dataset in the area of the Mediterranean
basin, especially using a detailed rain-gauge network.
Section 2 presents a climatology of CHIRPS database over the Medi-
terranean, together with a brief comparative assessment of this data-
base with other sources of precipitation data. Section 3 focuses over
Cyprus, making use of the extensive network of rain gauges of the is-
land. Conclusions are presented in Section 4.
2. CHIRPS Climatology
2.1. CHIRPS data climatology over the Mediterranean
The spatial distribution of the data gives a very detailed representa-
tion of precipitation climatology in the Mediterranean basin in which
elevation effects along with latitude and longitude have been taken
into consideration (Funk et al., 2013). This is obvious looking both at
the annual average map (Fig. 1) and the seasonal maps (Fig. 2), where
the characteristics of rainfall climatology for the area surrounding the
Mediterranean Sea can be observed.

http://dx.doi.org/10.1016/j.atmosres.2015.05.015
0169-8095/© 2015 Elsevier B.V. All rights reserved.

Please cite this article as: Katsanos, D., et al., Validation of a high-resolution precipitation database (CHIRPS) over Cyprus for a 30-year period,
Atmos. Res. (2015), http://dx.doi.org/10.1016/j.atmosres.2015.05.015
2 D. Katsanos et al. / Atmospheric Research xxx (2015) xxx–xxx

Fig. 1. Spatial distribution of CHIRPS data over land areas surrounding the Mediterranean. Mean annual values (mm/year) for a 30 year period (1981–2010).

On the one hand, by looking at the map with the annual mean (see
Fig. 1), one can see the high precipitation values positioned over the
mountainous areas of the Alps and the Pyrenees and also over the coast-
al areas of the Balkan Peninsula. On the other hand, in the seasonal
mean (mm/month) maps (see Fig. 2), contrasting differences between
winter and summer can be observed. Summer (Fig. 2c) is topographical-
ly influenced, since the seasonal values are higher in the mountainous
regions of the Alps mainly, and also the Pyrenees in the west and the
Carpathians in the east. On the contrary, winter values (Fig. 2a) are
higher near the coasts of the Balkans, highlighting the influence of the
warmer sea and its temperature contrast with the cooler land surface.
Spring (Fig. 2b) has a “mixed” portrayal, with high values both over
mountainous areas and coastal zones, while autumn seems to be more
like winter than like summer, with the higher values being observed
primarily over coastal areas; however, high values are also observed
over some mountainous regions.
2.2. Comparison between CHIRPS and other datasets
CHIRPS data are compared with two different datasets over the
Mediterranean basin: TRMM's 3B43 product and E-OBS precipitation
data, both provided at a spatial resolution of 0.25 × 0.25°.
Level 3 TRMM 3B43 data, also called TRMM Multi-satellite Precipi-
tation Analysis (TMPA) products, have a spatial coverage that is quasi-
global. This is the first rain product, combining TRMM precipitation
radar (PR) and TRMM microwave imager (TMI) rain rates to calibrate
rain estimates from other microwave and infrared measurements

Fig. 2. Seasonal average values (mm/month) for CHIRPS data: a) December–January–February, b) March–April–May, c) June–July–August, d) September–October–November.

Please cite this article as: Katsanos, D., et al., Validation of a high-resolution precipitation database (CHIRPS) over Cyprus for a 30-year period,
Atmos. Res. (2015), http://dx.doi.org/10.1016/j.atmosres.2015.05.015
 D. Katsanos et al. / Atmospheric Research xxx (2015) xxx–xxx 3

(Huffman et al., 2007). The final product is an average value of the In the context of the comparisons performed in this section, the
monthly rain rate. correlation coefficient is calculated as:
The E-OBS data set was developed as part of the European Union's
X X X
6th Framework Programme ENSEMBLES project that covers the area N xy− x y
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r ¼ s ð1Þ
25°–75°N, 40°W–75°E (land only). The data set, comprising daily pre-
X X 2  X X 2 
cipitation, was constructed through interpolation of the most complete N x2 − x N y2 − y
collection of station data over wider Europe (Haylock et al., 2008; Klok
and Klein Tank, 2008; Hofstra et al., 2009). Precipitation data used in
the present study refers to version 8.0. where: N is the number of pairs of values, Σxy is the sum of the products
CHIRPS database is compared with TRMM and E-OBS for different of paired values, Σx is the sum of x values, Σy is the sum of y values, Σx2
periods, in respect to the years that each database is available. Therefore, is the sum of squared x values and Σy2 is the sum of squared y values.
the comparison with the 3B43 product of TRMM (Fig. 3a) is made for a On the one hand, CHIRPS correlates better with TRMM (Fig. 3 top)
15 year period (1998–2012), while the comparison with E-OBS (Fig. 3b) than E-OBS (Fig. 3 bottom); this should be expected since CHIRPS itself
covers a 30 year period (1981–2010). In order to have a common grid, contains TRMM data. Regarding CHIRPS and TRMM comparison (Fig. 3
CHIRPS data were rescaled to a 0.25 degree resolution, using the nearest top), it is noted that the values of correlation coefficient generally
neighbor method. Comparison with even coarser databases available exceed over 0.8 (this is quite a high value given the difference in resolu-
over the study area was not considered. tion). In the western parts of the Mediterranean, especially over Spain

Fig. 3. Correlation coefficient (r) between CHIRPS and TRMM-3B43 (top) and CHIRPS and E-OBS (bottom).

Please cite this article as: Katsanos, D., et al., Validation of a high-resolution precipitation database (CHIRPS) over Cyprus for a 30-year period,
Atmos. Res. (2015), http://dx.doi.org/10.1016/j.atmosres.2015.05.015
4 D. Katsanos et al. / Atmospheric Research xxx (2015) xxx–xxx

and Portugal, with the exception of some mountainous areas, values are
even higher than 0.9. This is normal since the topography effect is not
considered in TRMM, while in CHIRPS there is an effort to be taken
into account. On the other hand, the correlation between CHIRPS and
E-OBS is poor over some parts of central and eastern Mediterranean,
highlighting the possible lack or limited use of station data in the E-
OBS database over these areas (Fig. 3 bottom). Furthermore, in some
mountainous regions the coefficient is close to or even lower than 0.5.

3. Focus over Cyprus

Cyprus, located in the north eastern corner of the Mediterranean

Sea, has a typical eastern-Mediterranean climate, which means winters
are wet and summers are dry, in terms of precipitation. The highest av-
erage annual precipitation values are recorded at the meteorological
stations located on the mountainous areas of the island and the lowest
at the coastal ones (Fig. 4). This is very similar to the distribution
Fig. 5. Mean annual precipitation (CHIRPS data) over Cyprus for the years 1981–2010.
found by analyzing the CHIRPS database for a 30 year period (1981–
2010) over the island (Fig. 5).
Indeed, the mean annual precipitation map matches the climatic re- (300–500 m), Semi Mountainous (500–1000 m), and Mountainous
gime that is known for Cyprus (Michaelides et al., 2009) for the period (N1000 m). This is the official grouping of stations as adopted by
1961–1990, where the higher values of precipitation are distributed the Cyprus Department of Meteorology, based on climatological
over the central Troodos mountain range. and geomorphological information.

3.1. Climatological station network 3.2. Results

Cyprus Department of Meteorology operates a very reliable and well
distributed network of meteorological stations (most of them function-
ing for more than 50 years). It adequately covers the island and in the
framework of this study 140 stations were used (Fig. 6). The criterion
that was adopted here was the matching of CHIRPS grid with at least
one rain gauge (taking into account the fact that there is no data avail-
able for the northern part of the island).
The outcome was 100 grid cells of CHIRPS matching with these 140
stations from the available network; the majority of these cells are asso-
ciated with a single station (i.e., falling within its respective boundary),
while very few of them are associated with two or more stations.
Subsequently, the stations were classified in five groups consider-
ing their elevation: Coastal (0–150 m), Inland plain (0–300 m), Hilly
Table 1 presents the average correlation coefficient values between
CHIRPS grid cells and the associated stations, along with the corre-
sponding minimum and maximum values, per group of stations. In
the case of cells with two or more adjacent stations, the average month-
ly values of observed precipitation were used for the comparison with
the corresponding CHIRPS values. The correlation values presented in
this table concern the whole time series (e.g., 360 monthly values),
mean annual and mean monthly values for 9 months. Values for the
summer dry period (June, July, and August) were not included be-
cause for most cells the comparison is impossible during these
months; many of the stations used, did not record any precipitation
throughout the whole 30 year period for the aforementioned
months, while some of them recorded precipitation only for a few

Fig. 4. Climatological average annual precipitation over Cyprus for 1961–1990.

Please cite this article as: Katsanos, D., et al., Validation of a high-resolution precipitation database (CHIRPS) over Cyprus for a 30-year period,
Atmos. Res. (2015), http://dx.doi.org/10.1016/j.atmosres.2015.05.015
 D. Katsanos et al. / Atmospheric Research xxx (2015) xxx–xxx 5

Fig. 6. The location of the stations used. Stations are grouped in terms of their elevation.

years (less than 5). Nevertheless, CHIRPS always shows rainfall
occurrence (even a few mm per month), since a main concern of
the algorithm used is to avoid “false zeros” (Funk et al., 2013), lead-
ing to the exclusion of these months for the comparison.
The correlation coefficients regarding monthly values for the whole
period are quite high for all five station groups, with values around 0.85
but none of the station groups shows a clear dominance. The corre-
sponding values for the annual mean are lower, ranging between 0.70
and 0.74, with the mountainous stations showing a slightly higher cor-
relation. Regarding the mean monthly values, January is clearly the
month with the highest correlation, while the values for February,
March, and December are considered as agreeable; also, the mountain-
ous stations seem to have slightly or clearly higher correlation coeffi-
cient values.
Fig. 7 presents the difference between the stations' recorded average
annual precipitation and the corresponding values from CHIRPS data-
base, for each station group, separately. The same difference considering
all the stations together is shown in Fig. 8. An increase in the difference
for all stations is evident, but also in each group, especially after the mid-
90's, because of an increase in the average annual precipitation calculat-
ed from the CHIRPS data. This trend in CHIRPS data is in contrast to the
general trend of precipitation in the area of the eastern Mediterranean
and especially in the island of Cyprus. Michaelides et al. (2009) have
shown a decrease in the annual precipitation in Cyprus, while the
same conclusion is reached by analyzing the station data for the period
of this study (not shown). Possibly, this difference between the trend
of the station precipitation and the trend in CHIRPS data is due to the in-
corporation of TRMM estimates in the latter, which tend to overestimate

Table 1
Average correlation with minimum and maximum values for the five station groups, separately. All period refers to the whole time series monthly values — 360 months.

Group Coastal 0–300 m 300–500 m 500–1000 m N1000 m

All period 0.85 0.84 0.85 0.85 0.85

0.78–0.92 0.79–0.97 0.78–0.89 0.80–0.91 0.83–0.87
Annual 0.71 0.70 0.70 0.72 0.74
0.54–0.89 0.52–0.82 0.56–0.82 0.62–0.80 0.65–0.81
January 0.86 0.88 0.87 0.87 0.87
0.68–0.93 0.73–0.94 0.81–0.91 0.82–0.92 0.81–0.91
February 0.74 0.78 0.74 0.76 0.78
0.56–0.90 0.59–0.87 0.58–0.81 0.60–0.86 0.69–0.86
March 0.76 0.75 0.77 0.80 0.81
0.67–0.86 0.56–0.86 0.69–0.83 0.70–0.86 0.75–0.87
April 0.62 0.61 0.61 0.66 0.70
0.43–0.84 0.43–0.77 0.33–0.76 0.45–0.75 0.54–0.78
May 0.70 0.62 0.51 0.62 0.65
0.30–0.90 0.33–0.89 0.21–0.77 0.26–0.86 0.48–0.86
September 0.63 0.57 0.67 0.69 0.72
0.30–0.88 0.15–0.89 0.40–0.88 0.48–0.86 0.53–0.89
October 0.65 0.62 0.68 0.68 0.67
0.36–0.80 0.46–0.77 0.58–0.79 0.52–0.79 0.60–0.89
November 0.68 0.69 0.68 0.70 0.72
0.48–0.82 0.53–0.86 0.49–0.82 0.60–0.87 0.62–0.79
December 0.70 0.68 0.73 0.69 0.72
0.46–0.85 0.44–0.85 0.64–0.82 0.49–0.85 0.61–0.78

Please cite this article as: Katsanos, D., et al., Validation of a high-resolution precipitation database (CHIRPS) over Cyprus for a 30-year period,
Atmos. Res. (2015), http://dx.doi.org/10.1016/j.atmosres.2015.05.015
6 D. Katsanos et al. / Atmospheric Research xxx (2015) xxx–xxx
Fig. 7. Difference in annual precipitation, between CHIRPS and rain gauges, for each station
group.
Fig. 8. Difference in annual precipitation, between CHIRPS and rain gauges, for all stations.
rainfall; this fact results in increasing differences throughout the last
decade of the study period.
4. Conclusions
The present study focused on validating CHIRPS, a new high-
resolution precipitation climatology database, over the Mediterranean
with particular emphasis on the island of Cyprus. The period covered
by the study starts from 1981 and extends to 2010. Initially, cross corre-
lation with two other precipitation databases was performed over the
Mediterranean basin on a 0.25° × 0.25° grid. CHIRPS data was correlated
with TRMM and E-OBS rainfall products, for the time periods that these
two databases were available. The comparison with TRMM resulted in
relatively high correlation, with the highest values over the Iberian
Peninsula. The corresponding values for the correlation with E-OBS
were lower.
Subsequently, the study focused on the area of Cyprus, where a
dense and reliable network of rain gauges is available. Precipitation
Please cite this article as: Katsanos, D., et al., Validation of a high-resolution precipitation database (CHIRPS) over Cyprus for a 30-year period,
Atmos. Res. (2015), http://dx.doi.org/10.1016/j.atmosres.2015.05.015
data recorded at the stations which are located within each CHIRPS
grid cell were averaged and then compared with the corresponding
CHIRPS values, for five groups of stations based on their elevation. The
correlation between the two sets of data is higher for the winter
months, especially January, while December, February, and March
have also notable values. Also, the stations located in mountainous
zones seem to correlate better with CHIRPS.
Finally, it was found that the mean difference between rain gauges
and CHIRPS is increasing over time, especially during the last decade,
possibly due to the incorporation in the latter of TRMM estimates,
which tend to overestimate rainfall.
Acknowledgments
The authors acknowledge the usage of the CHIRPS data set from the
Climate Hazards Group (http://chg.geog.ucsb.edu/data/chirps). The E-
OBS data were acquired from the website of the ECA&D project
(http://www.ecad.eu). The TRMM data were acquired from the respec-
tive website at http://trmm.gsfc.nasa.gov. Ground-based data were pro-
vided by the Cyprus Department of Meteorology.
References
Funk, C.C., Husak, G., Michaelsen, J., Shukla, S., Hoell, A., Lyon, B., Hoerling, M.P., Liebmann,
B., Zhang, T., Verdin, J., Galu, G., Eilerts, G., Rowland, J., 2013. Attribution of 2012 and
2003–12 rainfall deficits in Eastern Kenya and Southern Somalia. Special Supplement
to the Bulletin of the American Meteorological Society. Vol. 94 (No. 9).
Haylock, M.R., Hofstra, N., Klein Tank, A.M.G., Klok, E.J., Jones, P.D., New, M., 2008. A
European daily high-resolution gridded dataset of surface temperature and precipita-
tion. J. Geophys. Res. 113 (D20119, http://dx.doi.org/10.1029/2008JD010201).
Hofstra, N., Haylock, M., New, M., Jones, P.D., 2009. Testing E-OBS European high-
resolution gridded data set of daily precipitation and surface temperature.
J. Geophys. Res. 114 (D21101, http://dx.doi.org/10.1029/2009JD011799).
Huffman, G.J., Adler, R.F., Bolvin, D.T., Gu, G., Nelkin, E.J., Bowman, K.P., Hong, Y., Stocker,
E.F., Wolff, D.B., 2007. The TRMM Multisatellite Precipitation Analysis (TMPA): quasi-
global, multiyear, combined-sensor precipitation estimates at fine scales.
J. Hydrometeorol. 38–55 (http://dx.doi.org/10.1175/JHM560.1).
Klok, L., Klein Tank, A.M.G., 2008. Updated and extended European dataset of daily cli-
mate observations. Int. J. Climatol. 29, 1182–1191 (http://dx.doi.org/10.1002/joc.
1779).
López-Carr, D., Pricope, N.G., Aukema, J.E., Jankowska, M.M., Funk, C., Husak, G.,
Michaelsen, J., 2014. A spatial analysis of population dynamics and climate change
in Africa: potential vulnerability hot spots emerge where precipitation declines and
demographic pressures coincide. Popul. Environ. 35, 323–339. http://dx.doi.org/10.
1007/s11111-014-0209-0 (http://dx.doi.org/110.3133/ds832).
Michaelides, S., Tymvios, F.S., Michaelidou, T., 2009. Spatial and temporal characteristics
of the annual rainfall frequency distribution in Cyprus. Atmos. Res. 94, 606–615
(http://dx.doi.org/10.1016/j.atmosres.2009.04.008]).
Shukla, S., McNally, A., Husak, G., Funk, C., 2014. A seasonal agricultural drought forecast
system for food-insecure regions of East Africa. Hydrol. Earth Syst. Sci. 18, 3907–3921.
http://dx.doi.org/10.5194/hess-18-3907-2014.
Tapiador, F.J., Turk, F.J., Petersen, W., Hou, A.Y., García-Ortega, E., Machado, L.A.T., Angelis,
C.F., Salio, P., Kidd, C., Huffman, G.J., de Castro, M., 70-97, 2012. Global precipitation
measurement: methods, datasets and applications. Atmos. Res. 104-105, 70–97
(http://dx.doi.org/10.1016/j.atmosres.2011.10.021).