The Diurnal Cycle of Precipitation in the Tropical Andes ... Precip Andes Colombia.pdf · Julian oscillation, and a diminished (increased) high-frequency activity in July–October

The Diurnal Cycle of Precipitation in the Tropical Andes of Colombia

GERMÁN POVEDA, OSCAR J. MESA, LUIS F. SALAZAR, PAOLA A. ARIAS, HERNÁN A. MORENO,SARA C. VIEIRA, PAULA A. AGUDELO, VLADIMIR G. TORO, AND J. FELIPE ALVAREZ

Escuela de Geociencias y Medio Ambiente, Universidad Nacional de Colombia, Medellín, Colombia

(Manuscript received 9 October 2003, in final form 7 July 2004)

ABSTRACT

Using hourly records from 51 rain gauges, spanning between 22 and 28 yr, the authors study the diurnalcycle of precipitation over the tropical Andes of Colombia. Analyses are developed for the seasonal marchof the diurnal cycle and its interannual variability during the two phases of El Niño–Southern Oscillation(ENSO). Also, the diurnal cycle is analyzed at intra-annual time scales, associated with the westerly andeasterly phases of the Madden–Julian oscillation, as well as higher-frequency variability (�10 days), mainlyassociated with tropical easterly wave activity during ENSO contrasting years. Five major general patternsare identified: (i) precipitation exhibits clear-cut diurnal (24 h) and semidiurnal (12 h) cycles; (ii) theminimum of daily precipitation is found during the morning hours (0900–1100 LST) regardless of season orlocation; (iii) a predominant afternoon peak is found over northeastern and western Colombia; (iv) over thewestern flank of the central Andes, precipitation maxima occur either near midnight, or during the after-noon, or both; and (v) a maximum of precipitation prevails near midnight amongst stations located on theeastern flank of the central Cordillera. The timing of diurnal maxima is highly variable in space for a fixedtime, although a few coherent regions are found in small groups of rain gauges within the Cauca andMagdalena River valleys. Overall, the identified strong seasonal variability in the timing of rainfall maximaappears to exhibit no relationship with elevation on the Andes. The effects of both phases of ENSO arehighly consistent spatially, as the amplitude of hourly and daily precipitation diminishes (increases) duringEl Niño (La Niña), but the phase remains unaltered for the entire dataset. We also found a generalizedincrease (decrease) in hourly and daily rainfall rates during the westerly (easterly) phase of the Madden–Julian oscillation, and a diminished (increased) high-frequency activity in July–October and February–Aprilduring El Niño (La Niña) years, associated, among others, with lower (higher) tropical easterly wave (4–6day) activity over the Caribbean.

1. Introduction

Colombia is located in the northwestern corner ofSouth America, exhibiting complex geographical, envi-ronmental, and hydroecological features in which theAndes plays a major role. In terms of precipitationamount and distribution, Snow (1976, p. 362) describesthe Andes as “a dry island in a sea of rain,” but theatmospheric dynamics and precipitation variability as-sociated with the Andes are not well understood at awide range of time–space scales. Improving our under-standing of the atmospheric dynamics and precipitationvariability is crucial under current environmentalthreats faced by the tropical Andes, identified as themost critical “hot spot” for biodiversity on earth (Myerset al. 2000).

On seasonal time scales, it is well known that thedisplacement of the intertropical convergence zone(ITCZ) exerts a strong control on the annual cycle ofColombia’s hydroclimatology (Snow 1976; Mejía et al.1999; León et al. 2000; Poveda et al. 2004, submittedmanuscript to J. Hydrol. Eng.). Central and westernColombia experience a bimodal annual cycle of precipi-tation with distinct rainy seasons (April–May and Oc-tober–November), and “dry” (less rainy) seasons (De-cember–February and June–August), which result fromthe double passage of the ITCZ over the region. Aunimodal annual cycle (May–October) is witnessedover the northern Caribbean coast of Colombia and thePacific flank of the southern isthmus, reflecting thenorthernmost position of the ITCZ over the continentand eastern equatorial Pacific, respectively (Hastenrath1990, 2002). A single peak is also evident over the east-ern piedmont of the eastern Andes. Moisture trans-ported from the Amazon basin encounters the oro-graphic barrier of the Andes, thus focusing and enhanc-ing deep convection and rainfall in the eastern flank ofthe Cordillera, with maximum rainfall occurring during

Corresponding author address: Germán Poveda, Escuela deGeociencias y Medio Ambiente, Facultad de Minas, UniversidadNacional de Colombia, Cra 80 x Calle 65, M2-315 Medellín, Co-lombia.E-mail: [email protected]

228 M O N T H L Y W E A T H E R R E V I E W VOLUME 133

© 2005 American Meteorological Society

MWR2853

June–August. The meridional migration of the ITCZ ishighly intertwined with atmospheric circulation fea-tures over the Caribbean Sea, the easternmost PacificOcean, and the Amazon River basin. The presence ofthe three branches of the Andes with elevations sur-passing 5000 m, containing tropical glaciers, and includ-ing diverse intra-Andean valleys in a predominantsouth–north direction introduce strong local orographiceffects that induce atmospheric circulations and deepconvection with heavy rainfall. Over the Pacific coast ofColombia, extreme precipitation rates are witnessed,including one of the rainiest regions on earth (Povedaand Mesa 2000). Other large-scale phenomena associ-ated with the seasonal hydroclimatological cycle overthe region include the low-level “Chorro del OccidenteColombiano” (CHOCO) jet that flows onshore fromthe Pacific Ocean (Poveda and Mesa 2000) and theassociated development of mesoscale convective com-plexes, which in turn exhibit characteristic diurnalcycles (Velasco and Frisch 1987; Poveda and Mesa2000; Mapes et al. 2003a). The interannual variability ofthe diurnal cycle is dominated by the effects of bothphases of El Niño–Southern Oscillation (ENSO). Sucheffects are due to influences of sea surface tempera-tures off the Pacific coast, but also to atmospheric tele-connections, which are enhanced by land surface–atmosphere feedbacks (Poveda and Mesa 1997, 2000;Poveda et al. 2001). At intraseasonal time scales, tropi-cal easterly waves are known to affect precipitation re-gimes over different regions in Colombia in their west-ward propagation during the Northern Hemispheresummer–autumn (Martínez 1993).

Previous studies regarding the diurnal cycle of rain-fall for Colombia and tropical South America arescarce and limited to small datasets. Trojer (1959) iden-tified a distinct effect of altitude on the diurnal cycle ofprecipitation maxima over the Cauca River valley, withan early-morning peak for stations located near the val-ley floor, while those at higher elevations showed af-ternoon maxima. Snow (1976, p. 365) found differentdiurnal-maxima characteristics at three localities lo-cated on each of the branches of the Colombian Andes:Quibdó (on the lowlands of the Pacific coast), showinga late-night and early-morning maximum; Chinchiná(central Andes), displaying an early-morning maxi-mum; and Bogotá (high plateau on the eastern Andes),exhibiting an afternoon maximum. Other studies fornorthern South America rely on precipitation estimatesderived from satellite data (Kousky 1980; Meisner andArkin 1987; Velasco and Frisch 1987; Hendon andWoodberry 1993; Janowiak et al. 1994; Negri et al.1994, 2000; Garreaud and Wallace 1997; Imaoka andSpencer 2000; Lin et al. 2000; Soden 2000; Yang andSlingo 2001; Ricciardulli and Sardeshmukh 2002; So-rooshian et al. 2002; Mapes et al. 2003a). For instance,Negri et al. (2000) identified a strong diurnal cycle inAmazonian rainfall and pointed out the important roleof land–atmosphere feedbacks in controlling the inten-

sity of the diurnal cycle. They also identified a rainfallmaximum offshore in the Gulf of Panama that occursduring the morning hours around 0900 LST and a to-pographically induced afternoon maximum in themountains of Venezuela (near 5°N, 63°W) that occursduring May–September. Over the eastern slope of theAndes, a nighttime maximum of precipitation has beenidentified (Garreaud and Wallace 1997; Negri et al.1994). The diurnal march of convection in a small re-gion of the Amazon was studied during the 1999 Tropi-cal Rainfall Measuring Mission (TRMM) Wet-SeasonAtmospheric Mesoscale Campaign of the Large-ScaleBiosphere–Atmosphere Experiment (WETAMC/LBA) (Machado et al. 2002). The three-part study ofMapes et al. (2003a,b) and Warner et al. (2003) reportsobservational and modeling results for the diurnal cycleof precipitation over western Colombia and the neigh-boring Pacific, using a short-period dataset (July–September 2000) that is based on the rain-rate satelliteestimates of Negri et al. (2002a). Most of the aforemen-tioned studies on the diurnal cycle of precipitation overthe region are based upon short-term datasets derivedfrom satellite information that, though spatially distrib-uted, may induce estimation errors. For instance, Negriet al. (2002b), showed that 3 yr of rainfall data derivedfrom the TRMM satellite are inadequate to describethe diurnal cycle of precipitation over areas smallerthan 12°�12° because of high spatial variability in sam-pling.

We aim to study and characterize the long-term sea-sonal diurnal cycle of precipitation (section 3) and itsrelationship with altitude on the Andes (section 4). Wealso aim to investigate the interannual variability of thediurnal cycle associated with both phases of ENSO(section 5). In section 6, the intra-annual variability ofthe diurnal cycle is studied for the easterly and westerlyphases of the Madden–Julian oscillation (30–60 days).In addition, high-frequency variability (less than 10days) of two contrasting ENSO years is studied, usinghourly data from rain gauges located throughout thetropical Andes of Colombia. Conclusions are summa-rized in section 7.

2. Data and methods

The dataset consists of hourly precipitation rates at51 gauges, with records spanning between 22 and 28 yr.Percentages of missing data are generally less than 5%,which was ignored for estimation purposes. The datasetwas provided in paper logs by the Centro Nacional deInvestigaciones del Café (CENICAFE) and EmpresasPúblicas de Medellín (EPM), from which the digitalfiles were created. The gauges lie between 1°15�N and10°20�N, and 77°29�W and 72°40�W, and are situatedbetween 260 and 2595 m in elevation. Details of thegeographical setting and exact location of the gaugesare shown in Table 1 and Fig. 1.

The Andes are divided into three major branches in

JANUARY 2005 P O V E D A E T A L . 229

Colombia, namely Cordillera Occidental (westernbranch), Cordillera Central, and Cordillera Oriental(eastern branch), which are separated by the two intra-Andean valleys of the Magdalena and Cauca Rivers.The “day,” or 24-h period, over which the hourly rain-fall values are examined is arbitrarily defined as run-ning from 0700 to 0700 LST to better capture the largequantities of precipitation observed during late-night–early-morning hours. Gauging stations are numbered inFig. 1 according to their location along different An-

dean slopes: western slope of the Cordillera Occidental(1 to 3); southern “Macizo Colombiano” (4 to 6); east-ern slope of the Cordillera Occidental belonging to theCauca River valley (7 to 12); the western slope of theCordillera Central (13 to 32); the eastern slope of theCordillera Central, Magdalena River valley (33 to 40);the western flank of the Cordillera Oriental (41 to 48);the eastern flank of the Cordillera Oriental (49 and 50);and a single station at the isolated northern Sierra Ne-vada de Santa Marta (51).

TABLE 1. Location and details of the 51 rain gauges.

No. Rain gauge Lat (N) Lon (W) Elev (m) Period of recordMissing

data (%)Elev from

valley floor (m)

1 Sireno 06°23� 75°40� 1210 Mar 1978–Apr 1999 0.02 Santa Bárbara 06°24� 75°43� 2595 Mar 1978–Dec 1999 0.03 Mandé 06°27� 75°08� 495 May 1978–Apr 1999 0.04 Manuel Mejía 02°25� 76°45� 1700 Jan 1972–Dec 1999 2.15 El Sauce 01°37� 77°06� 1610 Jan 1971–Dec 1999 12.46 Ospina Pérez 01°15� 77°29� 1700 Jan 1972–Dec 1999 0.77 Julio Fernández 03°48� 76°32� 1360 Jan 1972–Dec 1999 1.0 4808 Manuel M. Mallarino 04°13� 76°19� 1380 Jan 1973–Dec 1999 2.5 5709 Santiago Gutierrez 04°43� 76°10� 1550 Aug 1971–Dec 1999 5.7 820

10 Alban 04°46� 76°�11 1500 Jun 1973–Dec 1999 1.7 78011 Rafael Escobar 05°28� 75°39� 1320 Oct 1970–Dec 1999 0.112 Miguel Valencia 05°36� 75°12� 1570 Jan 1971–Dec 1999 1.913 Rosario 05°28� 75°40� 1600 Jan 1971–Dec 1999 0.9 108614 Luker 05°12� 76°28� 1420 Mar 1970–Dec 1999 2.7 36015 Agronomía 05°03� 75°29� 2150 Jan 1972–Dec 1999 1.1 148016 Santagueda 05°05� 76°15� 1010 Apr 1972–Dec 1999 1.2 35017 Santa Ana 05°01� 75°40� 1250 Jan 1972–Dec 1999 2.2 58018 Cenicafé 05°00� 75°36� 1310 Jan 1972–Dec 1999 0.4 63019 Naranjal 04°59� 75°39� 1400 Jan 1971–Dec 1999 0.6 72020 Santa Helena 04°57� 75°37� 1525 Nov 1980–Dec 1999 2.1 84021 El Jazmın 04°55� 75°38� 1600 Jan 1972–Dec 1999 2.1 91022 Planta Tratamiento 04°48� 75°40� 1450 Mar 1969–Dec 1999 4.0 74023 La Catalina 04°45� 75°45� 1350 Nov 1976–Dec 1999 1.0 63024 El Cedral 04°42� 75°32� 2120 Jan 1973–Dec 1999 5.8 142025 Arturo Gómez 04°40� 75°47� 1320 Jan 1972–Dec 1999 1.5 59026 Bremen 04°40� 75°37� 2040 Mar 1972–Dec 1999 1.7 130027 Maracay 04°36� 75°46� 1450 Feb 1982–Dec 1999 0.6 71028 El Sena 04°34� 75°39� 1550 Dec 1971–Dec 1999 8.8 80029 La Bella 04°30� 75°40� 1450 Jan 1972–Dec 1999 0.4 69030 Paraguaicito 04°23� 75°44� 1250 Jan 1972–Dec 1999 0.5 47031 La Sirena 04°17� 75°55� 1500 Jan 1971–Dec 1999 1.0 70032 La Selva 03°40� 76°18� 1800 Jan 1979–Dec 1999 3.3 90033 Jorge Villamil 02°22� 75°33� 1500 Jan 1972–Dec 1999 1.134 El Limón 03°40� 75°35� 990 Jan 1971–Dec 1999 7.2 61035 Chapetón 04°27� 75°16� 1300 Jan 1971–Apr 1988 2.2 101036 La Trinidad 04°54� 75°03� 1430 Jun 1972–Dec 1999 0.6 119037 Llanadas 05°05� 75°41� 1020 Jan 1972–Dec 1999 10.0 122038 Inmarco 06°17� 72°48� 260 Aug 1968–Aug 1992 0.0 14039 Bizcocho 06°18� 75°05� 1070 Apr 1971–Dec 1999 0.0 112040 Peñol 06°24� 75°51� 1880 Apr 1960–Feb 1999 0.0 175041 Aguas Blancas 06°50� 73°29� 920 Jan 1972–Dec 1999 8.442 Bertha 05°53� 73°34� 1700 Jan 1972–Dec 1999 5.1 155043 Yacopí 05°28� 74°22� 1340 Jan 1972–Dec 1999 1.544 Santa Inés 04°43� 74°28� 1250 Jan 1972–Dec 1993 1.9 99045 Misiones 04°33� 74°25� 1540 Jan 1977–Dec 1999 3.0 126346 Granja Tibacuy 04°22� 74°26� 1550 Jan 1972–Dec 1999 1.4 125047 Luis Bustamante 03°54� 74°34� 1643 Jan 1972–Dec 1999 18.948 La Montaña 03°33� 74°54� 1260 Jan 1973–Dec 1999 15.949 Blonay 07°34� 72°37� 1235 Jan 1972–Dec 1999 0.750 Franciso Romero 07°46� 72°40� 1000 Jan 1972–Dec 1999 1.151 Pueblo Bello 10°25� 73°34� 1000 Jan 1972–Dec 1999 3.4


To examine interannual variability, we quantify thediurnal cycle of precipitation during both phases ofENSO: El Niño and La Niña. ENSO years were ob-tained from the multivariate ENSO index (MEI), de-veloped by the National Oceanic and Atmospheric Ad-ministration (NOAA), as follows: 1957/58, 1965/66,1972/73, 1982/83, 1986/87, 1991/92, 1994/95, 1997/98,and 2002/03 for El Niño, and 1964/65, 1970/71, 1973/74,1975/76, 1988/89, and 1998/2000 for La Niña (see http://www.cdc.noaa.gov/�kew/MEI/).

Intra-annual variability of the diurnal cycle is inves-tigated by discriminating according to the phase of the

Madden–Julian oscillation (MJO; Madden and Julian1994; Maloney and Hartmann 2000). Characterizationof easterly and westerly phases of the MJO was madeusing the index suggested by Maloney and Hartmann(2000), defined as the first principal component ofzonal wind pentads over the eastern Pacific at 850 hPa,during the period 1949–1997, estimated with data fromthe National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) 40-Year Climate Reanalysis Project (Kalnay etal. 1996). The westerly (easterly) phase of the MJOcorresponds to those periods in which the index is

FIG. 1. Geographical setting and detailed location of the set of rain gauges. Numbers correspond to rain gaugesdescribed in Table 1.


greater (lesser) than plus (minus) one standard devia-tion of the series. The average diurnal cycle of precipi-tation was estimated for the easterly, westerly, and neu-tral phases of the MJO.

We also characterize the high-frequency variability(less than 256 h or 10 days) of hourly precipitation dur-ing La Niña and El Niño years. The activity of tropicaleasterly waves (4–6 days) is dominant within such fre-quency band during the June–October period. Towardthat end, we perform wavelet analyses of the series ofhourly rainfall rates, using the Morlet mother wavelet(Torrence and Compo 1998).

3. Seasonal variability

Fast Fourier transform analyses of the entire dataset(not shown) revealed significant spectral peaks at 24and/or 12 h, corresponding to distinct diurnal and se-midiurnal cycles. No evidence was found of regionalcoherence in the predominance of the identified peri-odicities. At monthly time scales, most rain gauges ex-hibit a seasonally varying diurnal cycle of precipitation,which is itself extremely variable over space. Figure 2shows results of the seasonal march of the diurnal cyclefrom a subsample of 17 stations selected to reflect thespatial disposition and results of the larger sample. Iso-lines represent the percentage of daily precipitation to-tal, on days with measurable precipitation, fallingthroughout the course of the day and are estimatedthrough interpolation of the original 24 h � 12 monthsmatrix.

Several features are worth noticing: (i) the presenceof differing timings of diurnal maxima at any given sta-tion, including shifts from unimodal to multimodal be-havior, within the year; (ii) the presence of unimodal,bimodal, or multimodal diurnal behavior at differentstations; (iii) the coexistence of different timing of thediurnal maxima at nearby stations during the same sea-son, and in summary; and (iv) an astounding variabilityof the number of daily maxima of the diurnal cycle,even at very small spatial scales.

However, a generalized pattern can be drawn fromthese figures: the minimum percentage of daily precipi-tation is always found during the morning hours of0900–1100 LST regardless of season or location. Also,two broad regional statements can be made. First, thepredominant afternoon maxima found in many stationslocated over northeastern and western Colombia, suchas at Mandé (3, located at the ten o’clock position inFig. 2), Luis Bustamante (47, at three o’clock in Fig. 2),Maracay (27, at seven o’clock in Fig. 2), Pueblo Bello(51), Manuel Mejía (4), El Sauce (5), and Paraguaicito(30), among others, may be explained as the result ofconvective precipitation associated with solar thermalforcing, enhanced by the entrance of low-level, mois-ture-laden winds onshore from the Caribbean and Pa-cific. These winds ascend due to orographic lifting and

reach condensation levels at sunset hours. Therefore,this daytime convection tends to be in the form of smallcloud systems (Mapes et al. 2003a). Second, a commonmaximum of precipitation prevails during the middle ofthe night amongst stations located on the eastern flankof the Central Cordillera, corresponding to the west-ern-edge margin of the wide Magdalena River valley[Bizcocho (39, at 12 o’clock in Fig. 2), La Trinidad (36,at one o’clock in Fig. 2), Inmarco (38), El Limón (34),Chapetón (35), Llanadas (37), and Peñol (40)]. Such alate-night–early-morning maximum is mainly associ-ated with the life cycle of mesoscale convective systems(Velasco and Frisch 1987), with an upscale develop-ment of storm size through the night, supported by di-urnally driven breezes and gravity waves (López andHowell 1967; Mapes et al. 2003a).

The identified diurnal patterns of precipitation seemto lack the type of temporal or spatial structure, orlarge-scale order, that one might expect in a region withsuch strong seasonality resulting from the passage ofthe ITCZ and such distinct, well-delineated, topo-graphic features. For instance, even within the groupingof stations near the Venezuelan border and to the west,which have been identified as sharing a similar unimo-dal behavior, local differences may be detected. Thetime of maximum precipitation at Aguas Blancas (41,located at two o’clock in Fig. 2) shifts during the yearfrom afternoon (July–August) to evening hours (No-vember–March); however, the eastward-facing Francis-co Romero (50, above the previous one in Fig. 2) showsa maximum during early-evening hours (1800 to 2000LST) throughout the year. Likewise, in the west,Mandé (3, at ten o’clock in Fig. 2) has a maximum ofprecipitation from 1700 to 2000 LST throughout theyear, while the timing of the maximum shifts markedlyat Sireno (1, above the previous one in Fig. 2) to 1900–0100 LST during May–October. The following detailedanalysis of the spectral characteristics of precipitationat stations in geographic proximity to one another, inthe Cauca and Magdalena valleys, illustrates that thespatial coherence in the diurnal cycle of rainfall is lim-ited to very small-scale regions within the intra-Andeanvalleys.

a. Cauca River valley

• Eastern flank of the Cordillera Occidental (westernbranch). The diurnal cycle at stations located 480 and700 m above the valley floor exhibit bimodal behav-ior, with a stronger afternoon maximum (1300 to1600) and a second midnight–predawn maximum, asexemplified by Julio Fernández (7) and Mallarino(8). Above 700 m, precipitation is almost exclusivelyconcentrated in the afternoon hours, as at SantiagoGutierrez (9, at four o’clock in Fig. 2) and Alban (10),or near midnight at the northerly Rafael Escobar (11)and Miguel Valencia (12).

• At stations across the valley, on the western flank of


FIG. 2. Seasonal march of the diurnal cycle of rainfall at 17 selected stations in the tropical Andes of Colombia. Thediurnal cycle is defined from 0700 to 0700 LST, and interpolated isolines indicate percent of total daily rainfall, with thecolor scale shown at the bottom. Boundaries of the neighboring (left) Cauca and (right) Magdalena River valleys areshown in white. The inset at the bottom left shows details of rain gauges located in the western flank of the CentralCordillera.


Fig 2 live 4/C

the central Andes, the diurnal cycle exhibits maximaeither at or near midnight, or during the afternoon, orboth. Examples of the first kind are Luker (14), San-tagueda (16, at four o’clock in Fig. 2), Santa Ana (17),Cenicafé (18), Naranjal (19, at nine o’clock in Fig. 2),and Santa Helena (20). Examples of stations withafternoon maximum of precipitation are Agronomia(15), El Cedral (24), Bremen (26, at five o’clock inFig. 2), Maracay (27), El Sena (28), and combinedafternoon and near-midnight peaks appear at ElJazmin (21), Planta de Tratamiento (22), La Catalina(23), Arturo Gomez (25, at six o’clock in Fig. 2), andLa Bella (29). The relationship with elevation ishighly variable. Stations located between 350 and400 m above the valley floor exhibit a single midnightpeak, as at Santagueda (16, at four o’clock in Fig. 2)and Luker (14). Between 400 and 900 m, patternsbecome more bimodal, with peaks during eveninghours (1700 to 1900), and a secondary midnight peak,as illustrated at Paraguaicito (30), Arturo Gómez (25,at six o’clock in Fig. 2), La Catalina (23), La Bella(29), La Sirena (31), Maracay (27), Planta Trata-miento (22), El Sena (28), and La Selva (32). At sta-tions located from 900 to 1500 m above the valleyfloor there is a strong afternoon peak, though a muchsmaller midnight peak appears during August–December, such as at Bremen (26), El Cedral (24),and Agronomía (15). At Rosario (13) and El Jazmín(21) the diurnal cycle exhibits multimodality betweenafternoon and predawn hours.

b. Magdalena River valley

• At all stations on the eastern flank of the CordilleraCentral, the diurnal cycle exhibits a precipitationmaximum occurring near midnight (2300 to 0300 hLST), with very little precipitation before 1700 hLST. This is particularly clear on the eastern flanks,such as at Bizcocho (39, at twelve o’clock in Fig. 2),and Inmarco (38). But the peak appears earlier andless pronounced at Bizcocho during the period No-vember–March. No such seasonal changes are appar-ent at Inmarco.

• Across the valley, on the western flank of the Cor-dillera Oriental, with the exception of Bertha (42,afternoon peak), Luis Bustamante (47, at threeo’clock in Fig. 2, with midday peak), and Tibacuy (46,bimodal, afternoon, and midnight), all stations ex-hibit one of three forms of seasonally varying behav-ior:

(i) shifting unimodal, from a near-midnight precipi-tation maximum during October–March to anafternoon maximum in May–September, as atAguas Blancas (41, two o’clock in Fig. 2), LaMontaña (below the previous one in Fig. 2), andYacopí (43, below the two previous ones);

(ii) shifting from unimodal with a maximum in the

afternoon throughout the year to bimodal (a sec-ondary peak appears in the afternoon and nearmidnight) from November to March, as at SantaInés (44);

(iii) shifting unimodal, from a nighttime maximum(1900 to 2300 LST) during November–March toan afternoon maximum during May–September,as at Misiones (45).

4. The role of altitude and the position ofthe ITCZ

In an attempt to explain the observed behavior of thediurnal maxima of rainfall, potential relationships withaltitude and with the seasonal meridional migration ofthe ITCZ are examined. Figure 3 shows no clear rela-tionship between the timing of diurnal maxima and astation’s altitude. An apparent relationship with the al-titude is evidenced only in small groups of rain gauges,as in the previous discussion of the Magdalena andCauca River valleys. Similar analyses (not shown) areperformed according to the location of the rain gaugeson each branch and aspect of the Andes, but no generalspatial pattern emerges. Despite the generality of thebimodal pattern witnessed in average monthly precipi-tation over the Andes, the relationship between theannual and diurnal cycles is highly variable in space.Our analysis (not shown) indicates that the seasonallydry and wet epochs discussed in the introduction areassociated with seasonal shifts in diurnal maxima in 24of the 51 rain gauges. Geographically, this association ismost pronounced over the western flank of the Cordil-lera Oriental (eastern branch, rain gauges 41 to 48).The possible forcing of the ITCZ position on the diur-nal cycle may be related to the predominant southeast-erly (boreal summer) or northeasterly (austral summer)surface trade winds, and also to lower surface atmo-spheric pressures that favor low-level moisture conver-gence and deep convection.

Topography and aspect appear to play an importantrole in controlling the observed diurnal cycle; however,relief alone does not explain the observed seasonalvariability of diurnal maxima displayed by many raingauges. In tropical mountainous regions with suchrough terrain as the Andes, the marked spatial variabil-ity of diurnal precipitation maxima is strongly deter-mined by very local topographic and environmentalcharacteristics that interact with coexisting large-scaleforcing mechanisms, such as the aforementioned night-life cycle of mesoscale convective systems (Velasco andFrisch 1987; Mapes et al 2003a; Zuluaga and Poveda2004), and the position of the ITCZ, to generate deepconvection. We suggest that such combination of fac-tors may contribute to explain the strong space–timevariability identified here. The limitations of the avail-able data impede a full explanation of the complex pat-


tern of alternating unimodal and bimodal diurnal cyclesof rainfall identified at many stations.

5. Interannual (ENSO) variability

Both El Niño and La Niña strongly affect precipita-tion amounts at seasonal and annual time scales in Co-lombia. El Niño is associated with negative anomaliesin rainfall, river flows, soil moisture, and vegetationactivity, particularly during September–February, whileLa Niña is generally associated with positive rainfallanomalies (Poveda 1994; Poveda and Mesa 1997; Po-veda and Jaramillo 2000; Poveda et al. 2001; Waylenand Poveda 2002). Are these anomalies the result ofchanges in the amplitude and/or phase of the diurnalcycle as a consequence of changes in the number, du-ration, and/or intensity of storms? To answer thesequestions, the average diurnal cycles associated withboth ENSO phases are estimated and compared tothose of “normal” years. The hydrological year is de-fined from June (year 0, during El Niño onset) throughMay (year �1). Figure 4 shows the average diurnalcycle during both phases of ENSO and normal years,for the stations shown in Fig. 2.

Without exception, hourly rainfall increases (dimin-ishes) throughout the diurnal cycle during La Niña (ElNiño). During La Niña, average daily total rainfall in-creases between 5.7% and 40%, as compared to “nor-mal” years, with average increase of 20.8% among allstations. Equivalent percentages during El Niño dimin-ish between 6.9% and 56% for an average decrease of22.2%. Overall total daily precipitation is 48% higher

during La Niña years than during El Niño years. Ap-plication of a two-sample Kolmogorov–Smirnov non-parametric test (Press et al. 1986) of differences be-tween two probability distribution functions (PDFs) ofhourly records during El Niño and La Niña revealedthat, at more than 85% of the stations, the null hypoth-esis of similar PDFs during both phases of ENSO wasrejected. In spite of the effects of both phases of ENSOon the amplitude of diurnal precipitation, the phase ofthe diurnal cycle remains unaltered.

6. Intra-annual variability

a. Phase of the MJO

During the westerly (easterly) phase of the MJO thediurnal cycle of precipitation shows greater (lesser) am-plitude than the long-term average diurnal cycle. Figure5 displays the results for the selected stations and indi-cates that the phase of the diurnal cycle remains unal-tered. A detailed analysis shows that during the west-erly phase there is an increase in daily precipitation at85% of the stations. This increase averages 37% (14%to 76.9%), compared with the long-term diurnal pre-cipitation. On the other hand, the easterly phase is as-sociated with a decrease in daily precipitation at 88%of the stations, the anomaly being of 72% (0% to206.7%). No change in phase was detected. Interest-ingly, during the westerly phase there is greater tropicalstorm activity over the Caribbean (Maloney and Hart-mann 2000), which in turn may disrupt atmosphericpatterns, leading to heavy rainfall events over northernSouth America.

FIG. 3. Seasonal distribution of the relationship between altitude above sea level and thetiming of rainfall diurnal maximum, for the entire dataset.


FIG. 4. Average diurnal cycle of precipitation during El Niño (red), La Niña (blue), and normal (black) years, for theset of rain gauges depicted in Fig. 2. Hourly precipitation rates are shown in mm h�1 (right axis), and cumulativeprecipitation is in mm (left axis).


Fig 4 live 4/C

FIG. 5. Average diurnal cycle of precipitation rates (mm h�1) during the westerly (blue), normal (black), and easterly(red) phases of the MJO, for the set of rain gauges depicted in Fig. 2.


Fig 5 live 4/C

b. High-frequency variability (2 to 256 h)

Figure 6 shows the wavelet spectra at Paraguaicito(No. 30 in Table 1), corresponding to the June–Mayperiod of 2 yr of differing ENSO phase: La Niña 1988/89 (left) and El Niño 1982/83 (center). The distributionof the squared absolute value of the correspondingwavelet coefficients is shown in the first two graphs ofFig. 6. Each wavelet spectrum is scaled so that the sumof the squared absolute value of the coefficients is equalto the total variance. Thus, regions shaded with a colorcorresponding to 90% indicate that those coefficientsexplain more variance than 90% of all wavelet trans-form coefficients. In a sense, they represent those re-gions where the variance is concentrated. The thirdgraph of Fig. 6 shows the global time-integrated wave-let power spectrum (Torrence and Compo 1998).

Results show that the largest proportion of the vari-ance is concentrated in the frequency bands associatedwith 32 and 256 h (1.3 and 10.7 days), whose origindeserves further investigation. Also, differences appearin the high-frequency bands between the two phases ofENSO. During El Niño, high-frequency variability de-creases throughout the entire June–May period, asevidenced by the time-integrated power spectrum,especially during July–October (year 0, of El Niñoonset), which can be partly explained by the observeddiminished activity of 4–6-day (96–144 h) tropical east-erly waves, over the Caribbean and northern SouthAmerica, during El Niño (Gu and Zhang 2001). Itis important to note that the ability of the easterlywaves to perturb Colombian weather depends on thespecific path followed by their westward propagation.This might explain why the variance in the 4–6-daybands is highly intermittent. A decrease in high-frequency variability is also evidenced in El Niño dur-ing mid-February to mid-April (year �1), at 4–256 h(0.17 to 10.6 days). The explanation of this variabilitydeserves further investigation. Figure 6 also shows that

intraseasonal variability (30–60 day or 720–1440 h) ap-pears to be stronger during June–August during bothENSO phases, but also during October–December ofEl Niño. The generality of these latter observations re-quires further investigation.

7. Conclusions

The diurnal cycle of precipitation in the tropicalAndes of Colombia is characterized using long-termobservations. Five main general conclusions can bedrawn: (i) the observations indicate the existence ofseasonally varying diurnal (24 h) and semidiurnal (12 h)cycles. However, these cycles are highly variable inspace; (ii) daily precipitation minima are found pre-dominantly during the late-morning hours of (0900–1100 LST) regardless of season and location; (iii) overnortheastern and western Colombia, precipitationmaxima occur predominantly in the afternoon; (iv)over the western flank of the central Andes, precipita-tion maxima occur either around midnight, or duringthe afternoon, or both; and (v) over the eastern flank ofthe Central Cordillera, precipitation maxima occuraround midnight.

There is no evidence in our dataset of a relationshipbetween the timing of diurnal maximum rainfall withelevation, the branch of the Andes, or aspect, exceptfor some small groupings of stations within the intra-Andean Magdalena and Cauca River valleys. These re-sults indicate that the highly variable topography playsa major role in determining the observed diurnal cycles,yet the identified seasonal variations of diurnal precipi-tation maxima cannot be explained by relief alone. Themigration of the ITCZ may be a controlling factor toexplain the seasonal changes in diurnal precipitationmaxima in about half of the set of rain gauges. Weconjecture that the identified strong time–space vari-ability in the diurnal cycle of precipitation in the tropi-

FIG. 6. Wavelet transform spectra of hourly precipitation series at Paraguaicito (No. 30 in Table 1), during the Jun–May period oftwo contrasting ENSO years: (left) La Niña 1988/89, (center) El Niño 1982/83, and (right) the time-integrated global wavelet powerspectrum. Cross-hatched regions on either end of each plot indicate the “cone of influence,” where edge effects become important.


cal Andes of Colombia results from a combination oflocal- and large-scale environmental conditions, whichdeserve further investigation.

Observations indicate a spatially coherent responseof the diurnal cycle to both phases of ENSO, as hourly(and daily) rainfall diminishes during El Niño and in-creases during La Niña. High-frequency variability ismostly concentrated at 32- and 256-h cycles, duringboth ENSO phases, but high-frequency variability isdiminished (increased) by the occurrence of El Niño(La Niña), in particular during the 4–6-day period ofhighest activity of tropical easterly waves (June–November of the ENSO year 0) and during February–April of year �1. In addition, the phase of the diurnalcycle remains unchanged either by ENSO or by MJO.Analysis according to the phase of the Madden–Julianoscillation indicates an overall increase (decrease) inthe amplitude of diurnal cycle during the westerly (east-erly) phase of the MJO, whereas the phase of the di-urnal cycle remains unaltered.

These findings need to be merged with diagnosticstudies of the diurnal cycle in tropical precipitation de-rived from satellite estimates, and also from generalcirculation (GCM) or mesoscale models. For instance,understanding the diurnal march of precipitation overthe tropical Andes of Colombia may contribute to testits purported feedback to oceanic precipitation over theeastern tropical Pacific, as suggested by Mapes et al.(2003b) and Warner et al. (2003). They can also be usedto improve modeling and forecasting of tropical con-vection over land, but also serve as ground truth forsatellite-estimated rainfall and elucidate the space–timedynamics of tropical precipitation over complex terrain.

Acknowledgments. We would like to thank A. Jara-millo, O. Guzmán, and A. Cadena from CENICAFE,and Empresas Públicas de Medellín for providing ac-cess to the data. Thanks to P. Waylen, C. Penland, T.Warner, J. Ramírez, and two anonymous reviewers fortheir thoughtful comments, which led to an improve-ment of the manuscript. D. Hartmann kindly providedthe MJO series. This work was supported by DirecciónNacional de Investigaciones (DINAIN), of UniversidadNacional de Colombia.

REFERENCES

Garreaud, R. D., and J. M. Wallace, 1997: The diurnal march ofconvective cloudiness over the Americas. Mon. Wea. Rev.,125, 3157–3171.

Gu, G., and C. Zhang, 2001: A spectrum analysis of synoptic-scaledisturbances in the ITCZ. J. Climate, 14, 2725–2739.

Hastenrath, S., 1990: Diagnostics and prediction of anomalousriver discharge in northern South America. J. Climate, 3,1080–1096.

——, 2002: The intertropical convergence zone of the easternPacific revisited. Int. J. Climatol., 22, 347–356.

Hendon, H., and K. Woodberry, 1993: The diurnal cycle of tropi-cal convection. J. Geophys. Res., 98, 16 623–16 637.

Imaoka, K., and R. W. Spencer, 2000: Diurnal variation of pre-

cipitation over the tropical oceans observed by TRMM/TMIcombined with SSM/I. J. Climate, 13, 4149–4158.

Janowiak, J. E., P. A. Arkin, and M. Morrissey, 1994: An exami-nation of the diurnal cycle in oceanic tropical rainfall usingsatellite and in situ data. Mon. Wea. Rev., 122, 2296–2311.

Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.

Kousky, V., 1980: Diurnal rainfall variations in northeast Brazil.Mon. Wea. Rev., 108, 488–498.

León, G. E., J. A. Zea, and J. A. Eslava, 2000: General circulationand the intertropical convergence zone in Colombia (in Span-ish). Meteor. Colomb., 1, 31–38.

Lin, X., D. A. Randall, and L. D. Fowler, 2000: Diurnal variabilityof the hydrologic cycle and radiative fluxes: Comparisonsbetween observations and a GCM. J. Climate, 13, 4159–4179.

López, M. E., and W. E. Howell, 1967: Katabatic winds in theequatorial Andes. J. Atmos. Sci., 24, 29–35.

Machado, L. A. T., H. Laurent, and A. A. Lima, 2002: Diurnalmarch of the convection observed during TRMM-WETAMC/LBA. J. Geophys. Res., 107, 8064, doi:10.1029/2001JD000338.

Madden, R. A., and P. A. Julian, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev., 122,814–837.

Maloney, E. D., and D. L. Hartmann, 2000: Modulation of hur-ricane activity in the Gulf of Mexico by the Madden–Julianoscillation. Science, 287, 2002–2003.

Mapes, B. E., T. T. Warner, M. Xu, and A. J. Negri, 2003a: Di-urnal patterns of rainfall in northwestern South America.Part I: Observations and context. Mon. Wea. Rev., 131, 799–812.

——, ——, and ——, 2003b: Diurnal patterns of rainfall in north-western South America. Part III: Diurnal gravity waves andnocturnal convection offshore. Mon. Wea. Rev., 131, 830–844.

Martínez, M. T., 1993: Influence on weather patterns of majorsynoptic scale systems in Colombia (in Spanish). Atmósfera,16, 1–10.

Meisner, B. N., and P. A. Arkin, 1987: Spatial and annual varia-tions in the diurnal cycle of large-scale tropical convectiveclouds and precipitation. Mon. Wea. Rev., 115, 2009–2032.

Mejía, J. F., and Coauthors, 1999: Spatial distribution, annual andsemi-annual cycles of precipitation in Colombia (in Spanish).DYNA, 127, 7–26.

Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. B. daFonseca, and J. Kent, 2000: Biodiversity hotspots for conser-vation priorities. Nature, 403, 853–858.

Negri, A. J., R. F. Adler, E. J. Nelkin, and G. J. Huffman, 1994:Regional rainfall climatologies derived from Special SensorMicrowave Imager (SSM/I) data. Bull. Amer. Meteor. Soc.,75, 1165–1182.

——, E. N. Anagnostou, and R. F. Adler, 2000: A 10-year clima-tology of Amazonian rainfall derived from passive micro-wave satellite observations. J. Appl. Meteor., 39, 42–56.

——, T. L. Bell, and L. Xu, 2002a: Sampling of the diurnal cycleof precipitation using TRMM. J. Atmos. Oceanic Technol.,19, 1333–1344.

——, L. Xu, and R. F. Adler, 2002b: A TRMM-calibrated infraredrainfall algorithm, applied over Brazil. J. Geophys. Res., 107,8048, doi:10.1029/2000JD000265.

Poveda, G., 1994: Empirical orthogonal functions in the relation-ship between river streamflows and sea surface temperaturesin the Pacific and Atlantic Oceans. Proc. XVI Latin AmericanHydraulics and Hydrology Meeting (in Spanish), Santiago,Chile, IAHR, 131–144.

——, and O. J. Mesa, 1997: Feedbacks between hydrological pro-cesses in tropical South America and large-scale oceanic–atmospheric phenomena. J. Climate, 10, 2690–2702.

——, and A. Jaramillo, 2000: ENSO-related variability of riverdischarges and soil moisture in Colombia. Biospheric Aspectsof the Hydrologic Cycle, No. 8, IGBP, 3–6.


——, and O. J. Mesa, 2000: On the existence of Lloró (the rainiestlocality on earth): Enhanced ocean–atmosphere–land inter-action by a low-level jet. Geophys. Res. Lett., 27, 1675–1678.

——, A. Jaramillo, M. M. Gil, N. Quiceno, and R. I. Mantilla,2001: Seasonality in ENSO related precipitation, river dis-charges, soil moisture, and vegetation index (NDVI) in Co-lombia. Water Resour. Res., 37, 2169–2178.

Press, W. H., B. P. Flannery, P. Brian, S. Teukolsky, and W. T.Vetterling, 1986: Numerical Recipes: The Art of ScientificComputing. Cambridge University Press, 818 pp.

Ricciardulli, L., and P. D. Sardeshmukh, 2002: Local time- andspace scales of organized tropical deep convection. J. Cli-mate, 15, 2775–2790.

Snow, J. W., 1976: The climate of northern South America. Cli-mates of Central and South America, W. Schwerdtfeger, Ed.,Elsevier, 295–403.

Soden, B., 2000: The diurnal cycle of convection, clouds and watervapor in the tropical upper troposphere. Geophys. Res. Lett.,27, 2173–2176.

Sorooshian, S., X. Gao, K. Hsu, R. A. Maddox, Y. Hong, H. V.Gupta, and B. Imam, 2002: Diurnal variability of tropical

rainfall retrieved from combined GOES and TRMM satelliteinformation. J. Climate, 15, 983–1001.

Torrence, C., and G. P. Compo, 1998: A practical guide to waveletanalysis. Bull. Amer. Meteor. Soc., 79, 61–78.

Trojer, H., 1959: Fundamentos para una zonificación meteo-rológica y climatológica del trópico y especialmente de Co-lombia. Rev. Cenicafé, 10, 288–373.

Velasco, I., and M. Frisch, 1987: Mesoscale convective complexesin the Americas. J. Geophys. Res., 92 (D8), 9591–9613.

Warner, T. T., B. E. Mapes, and M. Xu, 2003: Diurnal patterns ofrainfall in northwestern South America. Part II: Model simu-lations. Mon. Wea. Rev., 131, 813–829.

Waylen, P. R., and G. Poveda, 2002: El Niño–Southern Oscilla-tion and aspects of western South America hydro-climatol-ogy. Hydrol. Processes, 16, 1247–1260.

Yang, G. Y., and J. Slingo, 2001: The diurnal cycle in the Tropics.Mon. Wea. Rev., 129, 784–801.

Zulauga, M. D., and G. Poveda, 2004: Diagnostics of mesoscaleconvective systems in Colombia and the eastern Pacific dur-ing 1998–2002 using TRMM data (in Spanish). Av. Recur.Hidraul., in press.


Documents

The Diurnal Cycle of Precipitation in the Tropical Andes ... Precip Andes Colombia.pdf · Julian oscillation, and a diminished (increased) high-frequency activity in July–October