AVHRR warm-season cloud climatologies under various synoptic regimes across the Iberian Peninsula...

INTERNATIONAL JOURNAL OF CLIMATOLOGYInt. J. Climatol. (2014)Published online in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/joc.4102

AVHRR warm-season cloud climatologies under varioussynoptic regimes across the Iberian Peninsula

and the Balearic IslandsCesar Azorin-Molina,a* Sergio-M. Vicente-Serrano,a Deliang Chen,b

Bernadette H. Connell,c María-Ángeles Domínguez-Durán,d Jesus Revueltoa

and Juan-Ignacio López-Morenoa

a Instituto Pirenaico de Ecología, Consejo Superior de Investigaciones Científicas (IPE-CSIC), Zaragoza, Spainb Regional Climate Group, Department of Earth Sciences, University of Gothenburg, Sweden

c Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, CO, USAd National Institute of Aerospace Technology, Canaries Space Centre, Gran Canaria, Spain

ABSTRACT: In this study we retrieved the spatial distribution of mid-afternoon clouds under various synoptic regimesacross the Iberian Peninsula and the Balearic Islands for the warm/convective-season, from May to October. Accurate dailycloud masks were derived by applying a daytime over land multispectral convective cloud detection algorithm spanning 15years (1997–2011) of Advanced Very High Resolution Radiometer (AVHRR) HRPT data. We processed a total of 2094afternoon overpasses (between 1230 and 1720 UTC) corresponding to the NOAA-14, NOAA-16 and NOAA-18 spacecrafts,and stratified daily cloud masks as a function of: (1) the automated circulation-typing scheme of Jenkinson and Collinson and(2) the prevailing wind field at 850 hPa. The AVHRR warm-season cloud climatology with high spatial resolution (1.1-km)identified six representative areas (regions of interest; ROIs) with intensified cloud activity (hot spots). The results also revealedthe typical spatial distribution of clouds for each synoptic regime across the whole region, identified the synoptic patterns andwind regimes under which high amounts of clouds occur for each ROIs, and showed that strong boundary layer winds ingeneral increase the frequency of clouds. The regional cloud climatology presented here could be useful, e.g. to improveconvective short-term forecasting by identifying active cloud areas for each atmospheric type.

KEY WORDS NOAA AVHRR; warm-season cloud climatologies; synoptic regimes; Iberian Peninsula and Balearic Islands

Received 8 April 2014; Revised 3 June 2014; Accepted 9 June 2014

1. Introduction

Mid-afternoon clouds are generally associated withnon-frontal convective processes and represent a commonfeature mainly over the mountainous areas of both theIberian Peninsula (IP) and the Balearic Islands (BI) duringthe warm season of the year (May to October) (Sánchezet al., 1998). Convective cells develop with high frequencyacross portions of this temperate mid-latitude region, inparticular northeastern Spain over the Pyrenees and theIberian System Mountains (Romero et al., 2001) wheresummertime maxima of precipitation occurs (Martin-Videand Olcina-Cantos, 2001). The diurnal cycle of cloudsis particularly strong over land surfaces because of thesmall heat capacity compared with the ocean surfaces(Grawoski et al., 2006). Thus convective clouds andprecipitation are mainly driven by the intense diabaticsolar heating and intensified far more quickly at hightemperatures (Berg et al., 2013). This thermodynamic

* Correspondence to: C. Azorin-Molina, Departamento de ProcesosGeoambientales y Cambio Global, Instituto Pirenaico de Ecología,Consejo Superior de Investigaciones Científicas (IPE-CSIC), AvenidaMontañana 1005, 50059-Zaragoza, Spain. E–mail: [email protected]

process enhances: (1) the destabilization of the boundarylayer by mixing out low-level inversions; (2) the develop-ment of sea breezes and other local winds (e.g. anabaticvalley wind circulations on heated south-facing mountainslopes, Azorin-Molina et al., 2011) in an area of complexterrain that enhances orographic lift and the transport ofAtlantic and Mediterranean warm moist surface air tomountains in the daytime (Millan et al., 2005) and (3)the establishment of a quasi-permanent thermal low overthe Iberian plateau (Hoinka and De Castro, 2003). Inorder for convective clouds to become thunderstorms itis necessary that conditional instability exists: e.g. fromthe passage of short-wave troughs which bring cold airaloft and sometimes passing cold fronts over the northernfringe of the IP (Romero et al., 2001).

During dry summer months, clouds producing rain areoften the only source of precipitation over the region andare beneficial to agriculture directly as well as indirectlythrough the recharge of streamflow, aquifers and reser-voirs (Millan et al., 2005). However, more organized andlong-lasting convective thunderstorms (e.g. mesoscaleconvective systems) can become more intense andextensive and sometimes be extraordinarily severe, caus-ing heavy rain (i.e. flash floods), large hail, straight-line

© 2014 Royal Meteorological Society

C. AZORIN-MOLINA et al.

winds, lightning discharges and rare tornadoes, producingserious socioeconomic impacts with damages and eco-nomic loss (Llasat-Botija et al., 2007), preferably overthe Mediterranean region. On the evening of 7 August1996, rainfall in excess of 200 mm in a 3-h period causeda flash flood that killed 86 and injured 93 at a campingsite in Biescas (Huesca, Spain) (Romero et al., 2001) orthe ‘Montserrat-2000’ flash flood event that caused mate-rial damaged estimated at over 65 million euros (Llasatet al., 2003). Additionally, the short-term forecastingof the location, timing and intensity of isolated thun-derstorms represents a challenging task in numericalweather prediction (Azorin-Molina et al., 2014), mainlydue to (1) the uncertainties in the initial conditions, (2)the limited knowledge about the cloud microphysicalprocesses and (3) the difficulties in resolving low-levelconvergence and convection with fairly coarse horizontalresolution operational models (Mazarakis et al., 2009).Furthermore, isolated thunderstorm cells with severeweather can develop unexpectedly under weakly definedsynoptic-scale or mesoscale precursor disturbances (Wil-son, 2008), and may be missed by forecasts. Satellite

Table 1. Monthly number of NOAA-14, NOAA-16 andNOAA-18 AVHRR scenes used for the 6-month study period

May to October 1997–2011.

Satellite May Jun. Jul. Aug. Sep. Oct. Total

NOAA-14 82 85 100 101 87 86 541NOAA-16 89 94 118 116 90 101 608NOAA-18 155 144 155 138 176 177 945Total 326 323 373 355 353 364 2094

cloud monitoring could provide valuable informationfor improving the forecasts of these meteorologicalhazards and the climate monitoring of storms. More-over, knowledge of the distribution of mid-afternoonclouds under various synoptic regimes is useful for otherfields such as solar energy, hydrology and agricultureactivities.

Despite the benefits and risks associated with the devel-opment of mid-afternoon clouds and thunderstorms, theimpact of synoptic regimes on the spatial distributionof convective clouds using remote-sensing data across

Figure 1. (a) Map of the study area showing locations named in the text (rivers are shown as dashed lines) and (b) wind direction at the 850-hPalevel.

© 2014 Royal Meteorological Society Int. J. Climatol. (2014)

AVHRR WARM-SEASON CLOUD CLIMATOLOGIES UNDER VARIOUS SYNOPTIC REGIMES

Figure 2. Convective cloud frequency composites for mid-afternoon orbits for (a) May to October during 1997–2011. The number of images averagedis shown in the lower-left corner (n = sample size). Dashed lines represent (1) transect #1 (from South -ROIs5- to Northwest -ROIs1-) and (2) transect#2 (from South -ROIs5- to Northeast -ROIs2-). ROIs are shown by black squares of 50 × 50 pixels, except for the isle of Mallorca. Convective cloudfrequency composites for mid-afternoon orbits for (b) May, (c) June, (d) July, (e) August, (f) September and (g) October for 1997–2011. Maximumand minimum cloud frequency statistics for both the IP and the BI are shown in the lower-right corner for each figure. The inset graphs represent the

spatial distribution of clouds (solid line) in relation to the orography (black dotted line) for (1) transect #1 and (2) transect #2.

the IP and the BI has not been reported in the literature.Previous studies focussed on the location of areas withthe most frequent convection. For instance, Ramis andAlonso (1988) used geostationary visible imagery on 11May 1987 to show the development of a sea breeze frontin the isle of Mallorca. Pascual (1999) used geostationaryinfrared imagery to develop convection climatologies inthe northeastern IP. Pascual et al. (2004) examined polar1.1 km VIS and IR satellite imagery from NOAA-16 and

NOAA-17 to infer the existence of boundary layer con-vergence lines for sea breeze and coastal range interactionin the northeastern IP. More recently, Azorin-Molinaet al. (2009) used high-resolution daytime AdvancedVery High Resolution Radiometer (AVHRR) data forretrieving the spatial distribution of convective areasassociated with low-level sea breeze convergence andconsequently sea breeze front development, and providedstatistics about the impact of prevailing large-scale flows



Table 2. Mean cloud frequency (in %) from the convective composites shown in Figure 2 for (a) the Iberian Peninsula and the BalearicIslands and (b) the ROIs during the warm semester May to October (1997–2011).

May Jun. Jul. Aug. Sep. Oct. May to Oct.

(a) The Iberian Peninsula and the Balearic IslandsIP 38.6 27.0 17.0 20.5 29.6 40.4 28.6BI 28.7 20.1 16.0 19.6 34.7 38.3 26.2Mean 33.7 23.6 16.5 20.1 32.2 39.4 27.4(b) The ROIs during the warm semester May to October (1997–2011)ROIs1 59.3 51.9 46.5 46.0 41.7 49.2 48.9ROIs2 57.6 53.2 48.8 46.4 44.1 42.1 48.5ROIs3 39.7 22.7 12.1 18.2 28.6 40.0 26.6ROIs4 47.3 40.7 30.6 34.2 42.4 40.9 39.1ROIs5 39.7 23.1 9.6 15.5 34.8 39.9 26.8ROIs6 31.0 24.3 19.6 23.5 37.0 39.8 29.2

(a)

Figure 3. Cloud frequency composites for mid-afternoon orbits for (a) the nine anticyclonic and hybrid weather types, (b) the cyclonic and hybridweather types and (c) the directional weather types during the 15-year study period 1997–2011. The individual composites are arranged in a clockwisemanner, starting with the A type (upper left corner) and followed by the AN, ANE, AE, ASE, AS, ASW, AW and ANW (lower right corner). Thestandard deviation computed for the number of images averaged (n= sample size; in parenthesis the % of images with respect to the total 2094 cloud

masks), and the mean, maximum and minimum cloud frequency statistics for both the IP and the BI are shown for each composite.



(b)

Figure 3. Continued.

on the sea breeze convection. Furthermore, it should benoted that most studies on high-resolution regional cloudclimatologies have been computed by the Regional andMesoscale Meteorology Branch (RAMMB; http://rammb.cira.colostate.edu/research/satellite_climatologies/; lastaccessed 1 June 2014) for the United States, CentralAmerica, the Caribbean region and Spain.

The main goal of this study lies in objectively quantify-ing for the first time the impact of large-scale atmosphericcirculation on the spatial distribution of mid-afternooncloud frequency observed from high-resolution AVHRRsatellite-registered radiances; the stratification of dailycloud masks as a function of various synoptic regimesand the use of 15 years (1997–2011) of AVHRR data areadvances of the work published by Azorin-Molina et al.(2013). This article is organized as follows: Section 2describes the AVHRR data, and Section 3 shows the cloudfrequency statistics and the two synoptic classifications;Section 4 displays the mid-afternoon cloud frequencycomposites for various synoptic regimes across the IP andthe BI during the warm months (May to October) of the15-year period 1997–2011 and finally a summary andconclusions are drawn in Section 5.

2. Data

2.1. AVHRR data and pre-processing

AVHRR data from the NOAA-14 (AVHRR/2 instrumentonboard; 5-spectral channels), NOAA-16 and NOAA-18 (AVHRR/3 instrument onboard; 6-spectral channels)polar orbiting satellites were collected from the HighResolution Picture Transmission (HRPT) receivingground station placed at the National Institute forAerospace Technology (INTA, Maspalomas, GranCanaria, Canary Islands, Spain) and supplied by the Centrefor Reception, Processing, Archiving and Disseminationof Earth Observation (CREPAD programme; http://www.crepad.rcanaria.es/en/index-en.html; last accessed 1 June2014) for the warm-season May to October spanning the15-year study period 1997–2011. We chose these NOAAsatellites because their afternoon overpasses are ideal forcapturing the active hours of the life cycle of convection.AVHRR scenes used in this study were acquired from1230 UTC till 1720 UTC; NOAA-14 between 1310 and1720 UTC, NOAA-16 between 1244 and 1501 UTCand NOAA-18 between 1230 and 1437 UTC. The mostfrequent acquisition intervals occurred between 1300



(c)


and 1400 UTC. Regardless the instability of polar satel-lite orbits (Karlsson, 2003), we used all the availableAVHRR scenes since the time window of all of themcovers the active hours of convection. A fully automatedpixel-by-pixel pre-processing routine using the headerfiles (metadata) of L1B ESA SHARP data (binary file;SHARP is the European Space Agency – Earthnet formatfor AVHRR data) was designed in Environment for Visu-alizing Images (ENVI)+ Interactive Data Language (IDL)4.7 package, also including a geometric correction (fordetails refer Azorin-Molina et al., 2009). A set of 2094AVHRR scences (75.9% of the theoretically availablesatellite scenes) were analysed in this study. The 24.1%not used were thrown out because of (1) reception andtechnical-processing problems, (2) archiving-tape failures,(3) calibration errors, (4) georeferencing problems and (5)missed portions of the IP and the BI. Table 1 summarizesthe monthly number of AVHRR scenes for each satelliteduring the period 1997–2011.

3. Methods

3.1. Cloud detection algorithm and frequency statistics

We used a newly proposed daytime over land algorithm forcomputing AVHRR convective cloud climatologies for the

IP and the BI (Figure 1(a)) (for details refer Azorin-Molinaet al., 2013). The convective cloud frequency compositespresented in Section 4 are based on the cloud frequency(in percentage) regarding the complete available scenes.Additionally, we used these high-resolution cloud fre-quency maps to identify active convective hotspots as thosedisplaying the maximum frequency values (i.e. regions ofinterests, ROIs; see Section 4.1) in relation to topograph-ical features such as altitude and land/water boundaries(Klitch et al., 1985). Centred over the selected ROIs, wecomputed cloud frequency statistics for a 50× 50 pixelarray, which Azorin-Molina et al. (2009) identified as anarea large enough to quantify the impact of various synop-tic regimes on cloud frequency.

Azorin-Molina et al. (2013) employed the term ‘con-vective’ cloud frequency composites because the algo-rithm was specifically defined for identifying convectionand most clouds developed within the mid-afternoon over-passes; these clouds are mostly associated with convec-tive processes during the warm-season of May to Octo-ber. However, it should be noted that composites presentedin this study not only include cloud masks that compriselarger Cumulus and Cumulonimbus clouds, but also con-tain convection embedded in multilevel cloud cover, andmid-afternoon clouds linked to other synoptic-scale situa-tions (e.g. frontal passages). A detailed description of the



Figu

re4.

Box

-and

-whi

sker

plot

sof

clou

dfr

eque

ncy

for

com

pari

son

betw

een

(1)

the

antic

yclo

nic

and

hybr

id(l

eftp

arto

fea

chfig

ure)

,(2)

the

cycl

onic

and

hybr

id(c

entr

e)an

d(3

)th

edi

rect

iona

l(ri

ght)

wea

ther

type

sfo

rea

chof

the

six

RO

Is.T

hem

ean

(gre

ylin

e),t

hem

edia

n(b

lack

line)

,the

25th

and

75th

perc

entil

era

nge

(box

es),

the

10th

and

90th

perc

entil

es(w

hisk

ers)

and

the

5th

and

95th

perc

entil

es(d

ots)

are

repr

esen

ted

for

each

syno

ptic

type

.



Table 3. Summary of the five cloudiest and cloudless weather types for each of the six ROIs.

Cloudiest weather types Cloudless weather types

ROIs1 CSW N NW CN CNW ASE AS SE AE CSROIs2 AN N ANE NE CNE ASW CSE S AS ASEROIs3 CSW CW W SW CNW AE ASE ANE A ANROIs4 CSW CNE CN CW CE ASW ASE A AE SEROIs5 CW CNE CSW SW S ASW ANW AN A ANEROIs6 CW CN CSW C CNE ASW CS ASE CSE ANW

daytime over land cloud-masking algorithm is found in thestudy by Azorin-Molina et al. (2013).

3.2. Synoptic regimes

With the aim of analysing the impact of atmospheric cir-culation on the spatial distribution of mid-afternoon cloudpatterns, the 2094 cloud masks were stratified resultingin cloud frequency climatologies as a function of: (1) theautomated circulation-typing scheme by Jenkinson andCollison (1977) based on sea level pressure (SLP) and(2) the prevailing low-level boundary layer wind field at850 hPa.

3.2.1. Automated circulation-typing scheme byJenkinson and Collison

We chose the Jenkinson and Collison’s (1977) circulation-typing scheme as an objective synoptic classificationmethod (hereafter JC) because it has been successfullyapplied to study the relationship between climate variablesand atmospheric circulation over the IP (Goodess andPalutikof, 1998; Trigo and DaCamara, 2000; Goodess,2000; Spellman, 2000; Martin-Vide, 2001; Vicente-Serrano, 2004; Azorin-Molina et al., 2011; amongothers) and the BI (Grimalt et al., 2013). The grid-pointused consisted on 16 points of daily SLP reanalysis data(p(n)) at a 5∘ latitude by 10∘ longitude (Jones et al.,1993; Chen, 2000; Linderson, 2001). The area is boundedby 30.0∘ and 50.0∘N, and 20.0∘W and 10.0∘E, essen-tially centred over the IP as found by Azorin-Molinaet al. (2011). The SLP dataset was obtained from theNational Centres for Environmental Prediction (NCEP)and the National Centre for Atmospheric Research(NCAR) reanalysis project (http://www.cdc.noaa.gov/cdc/reanalysis/reanalysis.shtml; Kalnay et al., 1996; lastaccessed 1 June 2014) for the 15-year study period(1997–2011). The JC method is based on objectiverules concerning the calculation of seven equations orcirculation indices (for details refer Azorin-Molina et al.,2011). The synoptic catalogue identifies 27 weather typesgrouped into four categories: (1) directional flow types(N, NE, E, SE, S, SW, W and NW); (2) anticyclonic(A) and cyclonic (C) types related to the rotation of theatmosphere; (3) hybrid types (AN, ANE, AE, ASE, AS,ASW, AW, ANW, CN, CNE, CE, CSE, CS, CSW, CWand CNW) and (4) the unclassified type (UD), which wasresolved and classified by extending the original grid basisof Spellman (2000) and Martin-Vide (2001). Therefore,

26 weather types were used for stratifying daily cloudfrequency composites.

3.2.2. Wind field at 850 hPa

The prevailing direction and strength of wind in thelow-level troposphere have been used in satellite clima-tology for stratifying remote sensing data and investi-gate the impact of large-scale flows on cloud patterns(Combs et al., 2001; Connell et al., 2001; Azorin-Molinaet al., 2009). Previous studies found that the direction andstrength of low-level synoptic winds along with the shapeof coastline and complex terrain are the most influentialfactors for sea breezes and local winds (Estoque, 1962;Blanchard and López, 1984; Azorin-Molina and Chen,2009), which in turn directly control the development ofconvection and the evolution of storms over land (Gouldand Fuelberg, 1996).

The daily 850-hPa u-wind and v-wind data from theNCEP and the NCAR reanalysis project (http://www.cdc.noaa.gov/cdc/reanalysis/reanalysis.shtml; Kalnay et al.,1996; last accessed 1 June 2014) were used to designatethe synoptic-scale flow regimes described below. The gridat 40.0∘N− 5.0∘W was chosen as the most centred pointand therefore considered representative of large-scaleambient flow over the whole study area. The 850-hPawind was assumed to be representative of the low-levelsynoptic forcing because the layer 0–1.500 m is the onewhere local and mesoscale circulations trigger convectiveclouds (Banta et al., 1993; Helmis et al., 1995).

We classified synoptic-scale flows in 16 types in rela-tion to (1) the wind direction (eight directional flowtypes), i.e. N (337.6∘–22.5∘), NE (22.6∘–67.5∘), E(67.6∘–112.5∘), SE (112.6∘–157.5∘), S (157.6∘–202.5∘),SW (202.6∘–247.5∘), W (247.6∘–292.5∘), and NW(292.6∘–337.5∘) regimes and (2) the wind speed (twointensities), i.e. light to moderate (0–5.14 m s−1) andstrong (Str, >5.14 m s−1) categories according to Gouldand Fuelberg (1996) and Connell et al. (2001). A dia-gram of the near surface synoptic regimes is displayed inFigure 1(b).

4. Results

4.1. Seasonal cloud climatology and identification ofthe region of interest

Figure 2 displays the seasonal (i.e. May to October) andmonthly mid-afternoon cloud frequency composites for



Figure 5. Spatial distribution of clouds for (a) transect#1 (from South -ROIs5- to Northwest -ROIs1-) and (b) transect#2 (from South -ROIs5- toNorthest -ROIs2-) shown as dashed-line in Fig. 2a, for the 26 weather types in relation to the orography (black dotted lines). Note that the 26 weather

types have been grouped in pairs.




the period 1997–2011. The present study of warm-seasoncloud climatology (Figure 2(a)) is ideal for identifyingthe areas most likely to receive convection and there-fore paying particular attention by, e.g. forecasters. Apartfrom the marked latitudinal gradient found in cloudinesswith the maximum (73.8%) over the northernmost fringeof the Atlantic-Cantabrian area and the Pyrenees, partlyattributed to low-stratiform clouds linked to large-scale

northwesterly winds and the influence of cold fronts(Azorin-Molina et al., 2013), and the minimum in cloudi-ness (12.4%) over the southern region of the IP causedby the suppression of convection under the influence ofthe Azores high-pressure system; we identified six ROIsas those displaying high cloud frequency amounts. TheseROIs basically correspond to the main mountainous areasof the IP and the BI: (1) the Cantabrian Mountains (ROIs1;



(a)

Figure 6. Cloud frequency composites for mid-afternoon orbits for (a) the NE, E, SE and S and (b) the SW, W, NW and N synoptic-scale flowsduring the 15-yr study period 1997–2011. The standard deviation computed for the number of images averaged (n= sample size; in parenthesis the% of images with respect to the total 2094 cloud masks), and the mean, maximum and minimum cloud frequency statistics for both the IP and the

BI are shown for each composite.



(b)




Figu

re7.

Box

-and

-whi

sker

plot

sof

clou

dfr

eque

ncy

fore

ach

ofth

e16

syno

ptic

-sca

leflo

wty

pes.

The

mea

n(g

rey

line)

,the

med

ian

(bla

cklin

e),t

he25

than

d75

thpe

rcen

tile

rang

e(b

oxes

),th

e10

than

d90

thpe

rcen

tiles

(whi

sker

s)an

dth

e5t

han

d95

thpe

rcen

tiles

(dot

s)ar

ere

pres

ente

dfo

rea

chsy

nopt

ic-s

cale

flow

.



maximum cloud amount 66.2%), (2) the Pyrenees (ROIs2;67.6%), (3) the Central System (ROIs3; 41.5%), (4) theeastern Iberian System Mountains (ROIs4; 44.5%), (5)the Betic Mountains (ROIs5; 43.9%) and (6) the centreof the isle of Mallorca (ROIs6; 41.2%) as a result of thelow-level convergence of sea breezes in the centre of theisle (Ramis and Alonso, 1988). All the ROIs are shownby squares in Figure 2(a), except for ROIs6 where cloudfrequency statistics are computed for the whole isle of Mal-lorca because the 50× 50 square includes sea pixels thatare out of our area of interest. Along with these six convec-tive hotspots, these high-resolution cloud frequency mapsreveal other areas with high cloud frequency amounts thatare important for forecasting. In contrast to these zones,large cloud-free regions dominate around southern coastalareas, bays and particularly across main river valleys inrelation to subsident divergent flows over nearby waterbodies (Connell et al., 2001). Note that the spatial distri-bution of cloud frequency shown in Figure 2(a) is simi-lar to that of the seasonal (JJA) mean precipitation fromobserved and simulated grid datasets revealed by Car-doso et al. (2013). The inset transects #1 and #2 shown inFigure 2(b) are manifestations of orography on the cloudfrequency. It is noteworthy that the cloud frequency linesdisplay the same shape as the orography lines for both tran-sects; i.e. high cloud amounts over the mountainous areas(ROIs; 55–60%) and low cloud activity along river valleys(e.g. Guadalquivir, Tajo and Ebro; 15–20%).

Figure 2 also shows a pronounced intermonthly cyclein cloudiness. The statistics computed in Table 2 con-firm that May (Figure 2(b)) is the second cloudiest monthwith a mean cloud frequency of 33.7% (max. 87.1%and min. 16.9%), due to the development of widespreadclouds associated with convective processes during spring.The mean cloud frequency sharply decreases in June(Figure 2(c)) with 23.6% (max. 86.4% and min. 5.3%).This declining trend continues and the mean frequencyreaches the lowest level in July (Figure 2(d)) with 16.5%(max. 76.9% and min. 1.3%), as a consequence of thesubsident influence of the Azores high pressure system.A gradual increase in cloud activity was found in August(Figure 2(e)) with a mean cloud frequency of 20.1% (max.74.8% and min. 3.6%). September (Figure 2(f)) is thethird cloudiest month with a mean cloud frequency of32.2% (max. 68.1% and min. 11.2%). Finally, October(Figure 2(g)) represents the cloudiest month with a meancloud frequency of 39.4% (max. 63.4% and min. 22.7%)in accordance with the occurrence of cold fronts thatrepresents an instability approaching from the AtlanticOcean (Azorin-Molina et al., 2013), and the develop-ment of mesoscale convective systems over the Mediter-ranean area.

4.2. Cloud composites according to weather type

Figure 3 shows the 26 cloud frequency composites cor-responding to (1) the anticyclonic and hybrid weathertypes (Figure 3(a)), (2) the cyclonic and hybrid weathertypes (Figure 3(b)) and (3) the directional weather types

(Figure 3(c)). The box-and-whisker plots shown inFigure 4 summarize the cloud frequency statistics foreach synoptic pattern for comparison between the sixidentified ROIs. Both figures are particularly helpfulin identifying the typical spatial distribution of cloudsassociated with the weather types. Overall, there is a cleardifference in the total amount and the spatial patterns ofcloudiness among the three groups. The anticyclonic andhybrid weather types are generally characterized by lowcloud amounts across the region (with exceptions, seedescription for each ROIs below), whereas the cyclonicand hybrid weather types are associated with an unsta-ble atmosphere and widespread high cloud amounts.Furthermore, the directional weather types present thewell-defined cloud composites because the spatial distri-bution of mid-afternoon clouds is driven by flow directionand corresponding moisture sources (Gimeno et al.,2010). Next we focus on analysing the favourable weathertypes to develop (or inhibit) cloud development for eachROIs.

ROIs1 representing the Cantabrian Mountains receivesthe highest cloud amounts under the CSW type (meancloud frequency 79.2%) followed by the directional N(74.3%) and NW (73.8%) types, the hybrid cyclonicsCN (72.5%) and CNW (70.0%), among others. As anexception, high cloud amounts also develop under hybridanticyclonic weather types such as the ANW (61.2%) andthe AN (60.7%). It is therefore clear that mid-afternoonclouds are mostly formed over the Cantabrian Mountainswhen weather types with southwesterly, westerly, north-westerly and northerly circulations bring high Atlanticmoisture levels into the region. On the contrary, under theASE (12.2%) and the AS (16.3%) convective clouds areunlikely to develop as dry and warm tropical air movesfrom the north of Africa and the subsidence exerted by theAzores high-pressure system inhibits lifting. The ROIs2located in the eastern Pyrenees displays the highest cloudamounts under the AN (68.0%), the N (63.6%), the ANE(63.3%), the NE (62.4%) and the CNE (62.2%) weathertypes; i.e. mid-afternoon clouds develop when coldnortherly and humid northeasterly advections destabilizethe atmosphere. Along with the ROIs1, this mountainousregion also represents an exception of showing high cloudamounts under anticyclonic hybrid circulation types dueto the upward deflection of large-scale horizontal flow bythe orography. In contrast, the ASW (15.2%), the CSEand the S (21.6%) represent the circulation types withlowest cloud amounts over the Pyrenees. The ROIs3 inthe Central System Mountains receives the highest cloudamounts under the CSW (82.7%) and the CW (71.4%)types, followed by the directional W pattern (63.2%).Westerly and southwesterly circulations favour Atlanticmoisture advection towards the centre of the IP and theupward vertical propagation of moist air up by the orog-raphy help convective clouds to develop. The most stableAE (7.8%) and ASE (10.8%) weather types, i.e. easternMediterranean circulations, almost inhibit the develop-ment of mid-afternoon clouds over this mountainousregion.



Table 4. Summary of the five cloudiest and cloudless large-scale synoptic flows for each of the six ROIs. The suffix _str indicatesstrong wind categories.

Cloudiest wind regimes Cloudless wind regimes

ROIs1 W_str N_str NW_str SW_str W E_str E SE_str NE SROIs2 NW_str NW N_str W_str N SE_str E_str S_str S EROIs3 SW_str S_str W_str NW_str SE_str NE W E SW NROIs4 NW W_str N_str NW_str NE SE_str E SE S SWROIs5 NE_str S_str E_str SE_str NE W NW N SW NW_strROIs6 N_str NE_str N NE E_str S_str SE_str SW W SE

Cloud development in the ROIs4 over the eastern IberianSystem Mountains is basically influenced by cyclonictypes from both the Atlantic and the Mediterranean mois-ture sources, i.e. the CSW (68.9%), the CNE (62.5%) andthe CN and the CW (61.9%). It is also characterized by pre-senting moderate to high cloud amounts for almost everyweather circulation, except for the ASW (10.3%). Themost southern ROIs5 representing the Betic System Moun-tains is much affected by the influence of stable air masses(i.e. proximity of Africa and the subtropical anticyclonebelt) and mean cloud amounts only show moderate val-ues for the CW (56.2%), the CNE (53.3%) and the CSW(50.2%). Additionally, the stability induced by subsidenceunder the ASW type (6.0%) produces the lowest meancloud amount recorded among all ROIs. Finally, the ROIs6representing the whole of the isle of Mallorca displays lowto moderate mean cloud amounts for all weather types,only being greater than 50% for the CW type (57.1%), fol-lowed by the CN circulation pattern (44.8%). The lowestmean cloud frequency value is recorded under the ASWtype (9.6%) and the CS (11.6%). A summary of the fivecloudiest and cloudless weather types for each ROIs isshown in Table 3, and transects #1 and #2 for all the 26weather types are displayed in Figure 5 showing the spatialdistribution of clouds in relation to the orography.

4.3. Cloud composites according to background flowdirection and speed

Figure 6 displays the 16 cloud frequency maps as a func-tion of (1) wind direction and (2) wind speed, i.e. lightto moderate (left panel) and strong (right panel) intensi-ties. Figure 7 summarizes the cloud frequency statisticsin form of box-and-whisker plots for each synoptic-scaleflow type and ROI. On the one hand, the NE, E, SE andS (i.e. more Mediterranean winds) synoptic-scale flowcomposites (Figure 6(a)) are characterized by developingmid-afternoon clouds in the eastern part of the IP, partic-ularly over the mountainous areas under light to moderatewinds, whereas the strong flows surpass the high easternIberian mountain barriers bringing moisture and enhanc-ing the development of more and much widespread cloudswestward of the IP, even though mountains lead to someFöhn effect to the western parts of the IP. On the otherhand, the N, NW, W and SW (i.e. more Atlantic winds)synoptic-scale flow composites display a clear latitudinalgradient in the spatial distribution of clouds, with higheramounts in the northern part and lower amounts in the

southern–southwestern parts of the region under light tomoderate flows. The mid-afternoon clouds are more fre-quent and more widespread over the whole region understrong boundary layer winds.

The frequency of mid-afternoon clouds in the ROIs1 isgreater for strong W (mean cloud frequency 72.3%), N(67.0%) and NW (65.3%) flows that bring Atlantic mois-ture (e.g. in form of cold front passes or low-stratiformclouds) to the Cantabrian Mountains, whereas the Mediter-ranean circulations from the E (23.4%) and SE (31.3%)directions inhibit cloud development resulting in low cloudfrequencies due to the adiabatic heating that operates overthe ROIs1 located in the leeward side of the IP. In Figure 7,for all wind regimes, the strongest strength of the low-levelwind is linked with high amount of clouds, except for theabovementioned E and SE flows that exhibit fewer cloudsfor strong intensities than those under light to moderateflows. The ROIs2 presents a similar pattern with enhancedcloud development under the dominance of both the lightto moderate (68.3%) and strong (70.7%) NW direction,followed by the strong N (62.1%) and W (61.2%) types.In contrast, as occurred for the ROIs1, the strong SE(28.8%) and E (30.0%) flows represent the types showingthe minimum mean of cloudiness due to the dissipatingeffect exerted by the adiabatic heating; however, convec-tive clouds can develop just over the Pyrenees Mountainsas shown in Figure 6(a). It is also noteworthy over theROIs2 that for N, NW and W directions (i.e. favourabletypes for clouds to develop), except for the SW, the strongflows develop much more clouds, whereas for all the strongNE, E, SE and S flows (i.e. unfavourable types) the oppo-site occurs. In the case of the ROIs3, the most character-istic feature corresponds to the low mean cloud amountsfor almost all regimes, hypothetically due to its locationin the centre of the IP, i.e. far from both the Atlantic andMediterranean moisture sources. The strong SW (43.8%),S (43.2%) and W (42.5%) flows, which bring Atlantic trop-ical moisture flux, are the cloudiest wind regimes, whereasthe light to moderate NE (13.1%) flows are the cloud-less ones. Note that for all wind regimes strong intensitiesdevelop much more widespread clouds, with noticeabledifferences compared with the light to moderate winds forthe S, SW, W and NW types.

Over the ROIs4 in the Iberian System, moderate cloudamounts are found for all wind regimes. The highestcloud frequency observed under the light to moderateNW type (51.8%) is associated with the large surfaceconvergence between the Mediterranean sea breezes



(a)

(b)

Figu

re8.

Spat

ial

dist

ribu

tion

ofcl

ouds

for

the

light

tom

oder

ate

(bla

ckso

lidlin

e)an

dst

rong

(gre

yso

lidlin

ew

ithdo

ts)

win

dre

gim

esat

850

hPa

inre

latio

nto

the

orog

raph

y(b

lack

dotte

dlin

e)fo

r(a

)tr

anse

ct#1

(fro

mSo

uth

-RO

Is5-

toN

orth

wes

t-R

OIs

1-)

and

(b)

tran

sect

#2(f

rom

Sout

h-R

OIs

5-to

Nor

thea

st-R

OIs

2-)

show

nas

dash

edlin

ein

Figu

re2(

a).



and the NW flows, which strengthen vertical motion forconvection initiation (Azorin-Molina et al., 2014). Forinstance, convective clouds are well-organized parallel tothe Iberian Mediterranean coast as shown in Figure 6(b)for the light to moderate NW regime, whereas strongNW flows slightly weaken (45.8%) low-level sea breezeconvergence (Azorin-Molina et al., 2009). The oppositeoccurs for the strong W (47.3%), N (46.5%) and SW(41.5%) flows, which show greater amounts of cloudsthan light to moderate winds. In the case of the ROIs5, thedevelopment of mid-afternoon clouds is mainly associatedwith low-level moisture fluxes linked to the strong NE(57.1%), S (37.5%), E (37.3%) and SE (33.9%), whereasthe remaining four wind regimes tend to inhibit convec-tion, e.g. light to moderate W winds (15.9%). For the BeticMountains, we also detected much more clouds understrong flows for all the eight wind directions. Finally, overthe ROIs6 strong N (61.9%), NE (58.4%) and E (41.0%)flows bring Mediterranean moisture and basically unstablecold air aloft enhancing cyclogenesis and development ofclouds over the western Mediterranean basin. The cloud-less wind regimes corresponds to the strong S (19.2%)and SE (19.9%) flows that bring tropical air (i.e. warminversion layer) from north of Africa. The inversion actsas a cap inhibiting mid-afternoon convection. A summaryof the five cloudiest and cloudless synoptic-scale flows foreach ROIs is shown in Table 4, and transects #1 and #2 forall the 16 synoptic-scale flows are displayed in Figure 8.

5. Summary and conclusions

In this study we revealed the location of mid-afternoonclouds (1230–1720 UTC) across the IP and the BI throughthe warm-season (May to October) during a 15-yearstudy period (1997–2011); mid-afternoon cloud com-posites were basically represented by isolated convec-tive clouds, but also convection embedded in multilevelcloud cover and some other cloud types associated withlarge-scale synoptic disturbances. The novelty of this studyresides in the stratification of 2094 NOAA-AVHRR over-passes as a function of (1) the automated circulation-typingscheme by Jenkinson and Collison (i.e. 26 weather types)and (2) the prevailing wind field at 850 hPa (i.e. 16 windregimes). The application of this technique led to 42 highspatial resolution (1.1-km) regional cloud composites thatprovide valuable information for identifying the areas mostlikely to receive clouds (hot spots). It is concluded thatlarge-scale atmospheric circulation, land–water thermalcontrast and orography play a role in determining the com-posites. With the use of seasonal and monthly cloud com-posites we identified six highly active convective zonesreferred to ROIs (i.e. main mountainous areas of the IPand the BI) and showed the synoptic patterns and low-levelwind regimes that are more favourable to develop con-vective clouds for each ROIs. Even though operationalforecasters already pay particular attention over theseidentified zones, here we supply additional informationsuch as spatial distribution of mid-afternoon clouds and

where these clouds tend to form under each weather typeor wind regime. Moreover, here we focused our anal-ysis on the six identified ROIs (i.e. zones to be moni-tored closely by weather forecasters), but high-resolutionregional cloud composites presented in this article offerinformation about cloud development across the whole IPand the BI, enabling readers to look at the frequency ofclouds under different synoptic patterns in areas other thanthe described ROIs. The main features of the six ROIs canbe summarized as follows:

(1) The Cantabrian Mountains in the northernmost fringeof the IP (ROIs1) exhibits high cloud amounts for weathertypes and wind regimes associated with Atlantic moistureadvections; e.g. preferably under northerly, northwesterlyand westerly cyclonic, directional or even anticycloniccirculations.(2) The eastern Pyrenees (ROIs2) are mainly influencedby northerly and northwesterly advections. Northeasterlyhumid Mediterranean flows under both anticyclonic andcyclonic circulations also play a role.(3) The Central System Mountains (ROIs3) display a highdependency on southwesterly and westerly flows, mainlyunder cyclonic weather types and associated Atlanticmoisture flux.(4) The eastern Iberian System Mountains (ROIs4) arecharacterized by moderate to high amount of clouds forall weather types and wind regimes. However, cyclonicsituations originated from both the Atlantic and theMediterranean moisture sources tend to increase thecloud amount.(5) The Betic System Mountains (ROIs5) are much influ-enced by hybrid cyclonic weather types from the west,southwest and northeast, with enhanced cloud develop-ment for the Mediterranean wind regimes.(6) Finally, the isle of Mallorca (ROIs6) displays lowto moderate cloud amounts for all weather types, beinggreater for cyclonic patterns from west and north direc-tions, and also by northerly, northeasterly and easterlysynoptic-scale flows.

An interesting feature identified by this study is thatmid-afternoon cloud frequency is greater under strongwinds at 850 hPa. This was observed for all the six ROIswith only a few exceptions, but in contrast we also detectedthat for those unfavourable large-scale flows strong windsresulted in lower cloud frequencies than light to moderatewinds. Therefore, we noted that frequency composites ofmid-afternoon clouds are spatially distributed as a functionof (1) the prevailing direction (and strength) of low-levelwind fields, (2) their corresponding sea-surface latent-heatflux and (3) the frictional effects produced by the complexterrain of this mid-latitude region, which causes conver-gence and enhances lifting processes (i.e. focusing con-vection in slope zones).

To conclude, the first stratified regional cloud frequencycomposites compiled in this study for both the IP andthe BI represents a valuable information for improvingconvective short-term forecasting, if numeric prediction



of the atmospheric circulation indicators is available. The15-year climatology could also have practical applicationsfor those interested on the climate monitoring of storms, orfor those that make decisions related to, e.g. solar energyuse, hydrology, agriculture, tourisms, among many otherspheres of application. This claims the usefulness in satel-lite meteorology/climatology of stratifying remote sensingdata as a function of different atmospheric classifications;a technique little used so far in the scientific literature sincethe 1960s but with high potential for many regional appli-cations. For instance, satellite data provide a greater spatialrepresentation of cloud frequency than surface-based mea-surements. Finally, future work should also look at com-piling regional cloud composites stratified as a functionof various synoptic regimes through the cold season, i.e.November till April, and spanning the roughly 35 years ofAVHRR imagery.

Acknowledgements

This research was supported by (1) the JCI-2011-10263grant and (2) the projects CGL2011-27574-C02-02 andCGL2011-27536/HID financed by the Spanish Commis-sion of Science and Technology and FEDER and (3)the MEDACC project (LIFE12 ENV/ES/000536). Theauthors would like to thank the INTA’s CREPAD Pro-gram for supplying the AVHRR data (assistance providedby Angel Garcia-Sevilla and Marta Romeo-Gallego); andto Rafael Baena-Calatrava, Imanol Echave-Calvo and Fer-gus Reig-Gracia for the IDL programming assistance andDr Tinghai Ou for the reanalysis data processing. Theauthors wish to acknowledge the two anonymous review-ers for their detailed and helpful comments to the originalmanuscript.

References

Azorin-Molina C, Chen D. 2009. A climatological study of the influenceof synoptic-scale flows on sea breeze evolution in the Bay of Alicante(Spain). Theor. Appl. Climatol. 96: 249–260.

Azorin-Molina C, Connell BH, Baena-Calatrava R. 2009. Sea-breezeconvergence zones from AVHRR over the Iberian Mediterranean areaand the Isle of Mallorca, Spain. J. Appl. Meteorol. Climatol. 48(10):2069–2085.

Azorin-Molina C, Chen D, Tijm S, Baldi M. 2011. A multi-year studyof sea breezes in a Mediterranean coastal site: Alicante (Spain). Int. J.Climatol. 31(3): 468–486.

Azorin-Molina C, Baena-Calatrava R, Echave-Calvo I, Connell BH,Vicente-Serrano SM, López-Moreno JI. 2013. A daytime over landalgorithm for computing AVHRR convective cloud climatologies forthe Iberian Peninsula and the Balearic Islands. Int. J. Climatol. 33(9):2113–2128.

Azorin-Molina C, Tijm S, Ebert EE, Vicente-Serrano SM,Estrela-Navarro MJ. 2014. Sea breeze thunderstorms in the easternIberian Peninsula. Neighborhood verification of HIRLAM andHARMONIE precipitation forecasts. Atmos. Res. 139: 101–115.

Banta RM, Olivier LD, Levinson DH. 1993. Evolution of the MontereyBay sea-breeze layer as observed by pulsed Doppler lidar. J. Atmos.Sci. 50: 3959–3982, DOI: 10.1007/s10546–007-9185-6.

Berg P, Moseley C, Haerter JO. 2013. Strong increase in convectiveprecipitation in response to higher temperatures. Nat. Geosci. 6:181–185.

Blanchard DO, López RE. 1984. Variability of the convective fieldpattern in south Florida and its relationship to the synoptic flow.NOAA Tech. Memo. ERL ESG-4, Boulder, CO, 77 pp.

Cardoso RM, Soares PMM, Miranda PMA, Belo-Pereira M. 2013. WRFhigh resolution simulation of Iberian mean and extreme precipitationclimate. Int. J. Climatol. 33(11): 2591–2608.

Chen D. 2000. A monthly circulation climatology for Swedenand its application to a winter temperature case study. Int. J.Climatol. 20: 1067–1076, DOI: 10.1002/1097-0088(200008)20:10<1067::AID-JOC528>3.0.CO;2-Q.

Combs CL, Stuart NA, DeMaria M, Vonder Haar TH. 2001. Windregime GOES cloud cover composites for the Wakefield, VA countywarning area. Preprints, In 11th Conference on Satellite Meteorologyand Oceanography, American Meteorological Society, Madison, WI,165–167.

Connell BH, Gould KJ, Purdom JFW. 2001. High-resolution GOES-8visible and infrared cloud frequency composites over NorthernFlorida during the summers 1996–1999. Weather Forecast. 16:713–724, DOI: 10.1175/1520-0434(2001)016<0713:HRGVAI>2.0.CO;2.

Estoque MA. 1962. The sea breeze as function of the prevail-ing synoptic situation. J. Atmos. Sci. 19: 244–250, DOI:10.1175/1520-0469(1962)019<0244:TSBAAF>2.0.CO;2.

Gimeno L, Nieto R, Trigo RM, Vicente-Serrano SM, Lopez-Moreno JI.2010. Where does the Iberian Peninsula Moisture come from? Ananswer based on a Largrangian approach. J. Hydrometeorol. 11(2):421–436, DOI: 10.1175/2009JHM1182.1.

Goodess C. 2000. The Construction of Daily Rainfall Scenarios forMediterranean Sites Using a Circulation Type Approach to Downscal-ing, PhD thesis, University of East Anglia, Norwich, UK.

Goodess CM, Palutikof JP. 1998. Development of daily rainfall scenariosfor southeast Spain using a circulation-type approach todownscaling. Int. J. Climatol. 10: 1051–1083, DOI: 10.1002/(SICI)1097-0088(199808)18:10<1051::AID-JOC304>3.0.CO;2-1.

Gould KJ, Fuelberg HE. 1996. The use of GOES-8 imagery and RAMS-DIS to develop a sea breeze climatology over the Florida Pan-handle. Preprints, In Eighth Conference on Satellite Meteorologyand Oceanography, American Meteorological Society, Atlanta, GA,100–104.

Grawoski WW, Betchtold P, Chen A, Forbes R, Halliwell C, Khairoutdi-nov M, Lang S, Nasuno T, Petch J, Tao WK, Wong R. 2006. Daytimeconvective development over land: a model intercomparison based onLBA observations. Q. J. R. Meteorol. Soc. 132: 317–344.

Grimalt M, Tomàs M, Alomar G, Martin-Vide J, Moreno-García MC.2013. Determination of the Jenkinson and Collison ́s weather typesfor the western Mediterranean basin over the 1948–2009 period.Temporal analysis. Atmosfera 26: 75–94.

Helmis CG, Papadopoulos KH, Kalogiros JA, Soilemes AT, Asi-makopoulos DN. 1995. The influence of the background flow onthe evolution of the Saronikos Gulf sea breeze. Atmos. Environ. 29:3689–3701, DOI: 10.1016/1352 - 2310(95)00008 - M.

Hoinka KP, de Castro M. 2003. The Iberian thermal low. Q. J. R.Meteorol. Soc. 129(590): 1491–1511.

Jenkinson AF, Collison BP. 1977. An initial climatology of gales overthe North Sea. Synop. Climatol. Branch Memorandum, No. 62, Mete-orological Office, Bracknell, UK.

Jones PD, Hulme M, Briffa KR. 1993. A comparison of Lamb circulationtypes with an objective classification scheme. Int. J. Climatol. 13:655–663, DOI: 10.1002/joc.3370130606.

Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, GandinL, Iredell M, Saha S, White G, Woollen J, Zhu Y, Leet-maa A, Reynolds R, Chelliah M, Ebisuzaki W, Higgins W,Janowiak J, Mo KC, Ropelewski C, Wang J, Jenne R, JosephD. 1996. The NCEP/NCAR 40-year reanalysis project. Bull.Am. Meteorol. Soc. 77: 437–471, DOI: 10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.

Karlsson K-G. 2003. A 10 year cloud climatology over Scandinaviaderived from NOAA Advanced Very High Resolution Radiometerimagery. Int. J. Climatol. 23(9): 1023–1044.

Klitch MA, Weaver JF, Kelly FP, Vonder HT. 1985. Convective cloudclimatologies constructed from satellite imagery. Mon. Weather Rev.113(3): 326–337.

Linderson M. 2001. Objective classification of atmospheric circula-tion over southern Scandinavia. Int. J. Climatol. 21: 155–169, DOI:10.1002/joc.604.

Llasat MC, Rigo T, Barriendos M. 2003. The “Montserrat-2000”flash-flood event: a comparison with the floods that have occurred inthe northeastern Iberian Peninsula since the 14th century. Int. J. Cli-matol. 23: 453–469.

Llasat-Botija M, Llasat MC, López L. 2007. Natural hazards and thepress in the western Mediterranean region. Adv. Geosci. 12: 81–85.



Martin-Vide J. 2001. Limitations of an objective weather-typing systemfor the Iberian Peninsula. Weather 56: 248–250.

Martin-Vide J, Olcina-Cantos J. 2001. Climas y Tiempos de España.Alianza Editorial: Madrid.

Mazarakis N, Kotroni V, Lagouvardos K, Argiriou AA. 2009. The sensi-tivity of numerical forecasts to convective parameterization during thewarm period and the use of lightning data as an indicator for convectiveoccurrence. Atmos. Res. 94: 704–714.

Millan MM, Estrela MJ, Miro J. 2005. Rainfall components: variabilityand spatial distribution in a Mediterranean area (Valencia region). J.Clim. 18(14): 2682–2705.

Pascual R. 1999. Summer convection in Catalonia: an analysisusing Meteosat IR imagery (in Spanish). In Proceedings of IVNational Symposium of Forecasting, Spanish Meteorological Agency,Madrid.

Pascual R, Callado A, Berenguer M. 2004. Convective storm initiationin central Catalonia. In Proceedings of Third European Conferenceon Radar Meteorology (ERAD), Visby, Island of Gotland, Sweden,464–468.

Ramis C, Alonso S. 1988. Sea-breeze convergence line in Majorca. Asatellite observation. Weather 43: 288–293.

Romero R, Doswell CA, Riosalido R. 2001. Observations and fine-gridsimulations of a convective outbreak in Northeastern Spain: impor-tance of diurnal forcing and convective cold pools. Mon. Weather Rev.129(9): 2157–2182.

Sánchez JL, Fraile R, de la Fuente MT, Marcos JL. 1998. Discriminantanalysis applied to the forecasting of thunderstorms. Meteorol. Atmos.Phys. 68(3–4): 187–195.

Spellman G. 2000. The application of an objective weather-typing sys-tem to the Iberian Peninsula. Weather 55: 375–385.

Trigo RM, DaCamara CC. 2000. Circulation weather types andtheir influence on the precipitation regime in Portugal. Int. J. Climatol.20: 1559–1581, DOI: 10.1002/1097-0088(20001115)20:13<1559::AID-JOC555>3.0.CO;2–5.

Vicente-Serrano SM. 2004. Evolución Espacio-Temporal de las Sequíasen el Sector Central del Valle del EBRO: Causas y ConsecuenciasAmbientales, PhD thesis, Department of Geography and CountryPlanning, University of Zaragoza, Zaragoza, Spain.

Wilson PH. 2008. Warm-Season Lake-/Sea-Breeze Severe Weather inthe Northeast. Master thesis, Department of Earth and AtmosphericSciences, University at Albany, State University of New York, Albany,NY, 115 pp.


AVHRR warm-season cloud climatologies under various synoptic regimes across the Iberian Peninsula...

Documents

Transcript of AVHRR warm-season cloud climatologies under various synoptic regimes across the Iberian Peninsula...