Journal of Hydrology (2008) 361, 159–171

ava i lab le a t www.sc iencedi rec t . com

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

Study of dew water collection in humid tropicalislands

O. Clus a,b, P. Ortega b,c, M. Muselli a,b, I. Milimouk b,d, D. Beysens b,e,f,*

a Universite de Corse and CNRS UMR 6134, Route des Sanguinaires, Ajaccio 20000, Franceb International Organization for Dew Utilization, OPUR, 60, rue Emeriau, Paris 75015, Francec Universite de la Polynesie Francaise, B.P.6570, Faaa 98702, French Polynesiad Equipe du Supercritique pour l’Environnement, les Materiaux et l’Espace, CNRS-ICMCB, Pessac 33608, Francee Equipe du Supercritique pour l’Environnement, les Materiaux et l’Espace, ESPCI-PMMH, 10, rue Vauquelin, Paris 75231,Cedex 05, Francef Equipe du Supercritique pour l’Environnement, les Materiaux et l’Espace, CEA-SBT, Grenoble 38054, France

Received 7 December 2007; received in revised form 23 June 2008; accepted 24 July 2008

00do

*

Pa

KEYWORDSTropical islands;Dew harvesting;Water resource;Sky radiation;Radiative cooling

22-1694/$ - see front mattei:10.1016/j.jhydrol.2008.07

Corresponding author. Addris 75231, Cedex 05, FranceE-mail address: daniel.bey

r ª 200.038

ress: Equ. Tel.: +sens@es

Summary An assessment of the potential for dewwater to serve as a potable water sourceduring a rainless season in a humid tropical climate was carried out in the Pacific islands ofFrench Polynesia. The climate of these islands, in terms of diurnal and seasonal variations,wind and energy balance, is representative of the climate of the tropical Atlantic and Paci-fic oceans. Measurements were obtained at two characteristic sites of this region; a moun-tainous island (Punaauia, Tahiti Island) and an atoll (Tikehau, Tuamotu Archipelago). Dewwas measured daily on a 30� tilted, 1 m2 plane collector equipped with a thermally insu-lated radiative foil. In addition, an electronic balance placed at 1 m above the ground witha horizontal 0.16 m2 condensing plate made of PolyTetraFluoroEthylene (Teflon) was usedin Tahiti. Dew volume data, taken during the dry season from 16/5/2005 to 14/10/2005,were correlated with air temperature and relative humidity, wind speed, cloud coverand visible plus infrared radiometer measurements. The data were also fitted to a model.

Dew formation in such a tropical climate is characterized by high absolute humidity, weaknocturnal temperature drop and strong Tradewinds. Thesewinds prevent dew from formingunless protected e.g. by natural vegetal windbreaks. In protected areas, dew can then formwith winds as large as 7 m/s. Such strong winds also hamper at night the formation near theground of a calm and cold air layer with high relative humidity. As the cooling power is lowerthan in the Mediterranean islands because of the high absolute humidity of the atmosphere,both effects combine to generatemodest dewyields. However, dewevents are frequent andprovide accumulated amounts of water attractive for dew water harvesting. Slight modifi-

8 Elsevier B.V. All rights reserved.

ipe du Supercritique pour l’Environnement, les Materiaux et l’Espace, ESPCI-PMMH, 10, rue Vauquelin,33 140795806; fax: +33 140794523.pci.fr (D. Beysens).

160 O. Clus et al.

cations of existing rain collection devices on roofs can enhance dew formation and collec-tion. Dew harvesting thus appears as an attractive possibility to provide the local populationwith a complementary – but on occasion, essential – water resource.

ª 2008 Elsevier B.V. All rights reserved.

Introduction

Water in islands is mostly provided by the phreatic zonesthat are replenished by precipitation. Although waterrequirements of the local population are adequate in highmountainous islands, the situation is quite different in lowislands (atolls). There, the available water is scarce and orsaline, especially during the dry season. It is precisely dur-ing this season that tourism peaks, thus placing a newstress on the demand for fresh water. In the Pacific atolls,the ground water salinity has increased through over-pumping and new wells have not followed the traditionalrules requiring a minimal distance between wells, ocean,lagoon and waste water release (Dupon, 1987). As a conse-quence, rain water (and expensive imported bottled water)has become the exclusive source of potable water. Thegalvanized iron roofs of dwellings serve as rain collectors.Large water tanks provide about 4 months of water supplyto inhabitants, with a rationed use in the dry season. How-ever, droughts appear periodically and the situation canbecome hazardous.

Literature review

In some regions of the world, dew water – if available – ap-pears to be a simple solution to complement sources of po-table water. Dew water is indeed used by plants and smallanimals where, in arid and semi-arid environment, it is sig-nificant to sustain their activity (Gindel, 1965; Steinbergeret al., 1989). Dew comes from atmospheric humidity thatis transformed into liquid water on a surface that is pas-sively cooled by radiation (Monteith, 1957; Beysens, 2006).In theory, the maximum dew yield that can be expected isabout 0.8 L/m2, being restricted by the available coolingpower (25–100 W m�2) with respect to the latent heat ofcondensation (2500 kJ kg�1).

The question whether dew can be a reliable source ofwater has been posed many times. Recently, systematicinvestigation were performed of adapted condensing archi-tecture using high yield radiative materials with hydrophilicproperties (complete wetting case) that enhance dew waternucleation (see Beysens, 2006) and dew drop gravity recov-ery. The influence of local meteorological parameters suchas wind speed and relative humidity was systematicallystudied (Nilsson, 1996; Vargas et al., 1998; Muselli et al.,2002; Muselli et al., 2006a; Beysens et al., 2003, 2005; Ja-cobs et al., 2008; Sharan et al., 2007). Water quality wasalso investigated (Beysens et al., 2006; Muselli et al.,2006b). For a plane condensing structure an optimum tiltangle from horizontal was found to be 30� (Beysens et al.,2003). The maximum dew yield that the present authorsare aware of is about 0.6 L m�2 measured in Jerusalem(Berkowicz et al., 2004), not far from the expectedmaximum.

Objective

The present paper aims to assess the amount of dew waterthat could be collected in Pacific atolls during the dry sea-son that lasts from May to October (Laurent et al., 2004).In addition, a secondary objective is to compare the yieldswith those of a Mediterranean island (Ajaccio) where dataare available.

Despite a slight difference of the oceanic currents andevapotranspiration, the climate of those atolls in terms ofwind, rainfall or energy balance is very close to the tropicaloceanic climate since they are devoid of relief and have avery small land surface. The atoll selected for the experi-ment is thus representative of a large number of islandsspanning over millions of square kilometers in the oceanictropical area (Pacific, Indian or Atlantic). For the TuamotuArchipelago alone, there are more than 70 atolls. A Fewstudies have been concerned with dew formation in tropicalcontinental areas (Salau and Lawson, 1986; Lhomme andJimenez, 1992; Khedari et al., 2000; Weihong and Gou-driaan, 2000; Barradas and Glez-Medellın, 2000). Thereare, to our knowledge, no studies concerning atolls or trop-ical islands, where dew forms under unusual conditions ofstrong (Trade) winds and high atmospheric humidity, thelatter inducing a large natural green house effect.

In order to characterize dew formation and collection inthis environment, two typical sites were chosen: a low coralatoll in the Tuamotu islands (Tuherahera, Tikehau Island,French Polynesia) and a high, mountainous island (Punaauia,Tahiti Island, French Polynesia) 320 km apart. Data havebeen obtained on each site on inclined plane collectorsand supplemented in Tahiti by continuous dew measure-ments on an electronic balance. Data are correlated withair temperature, air relative humidity, wind speed, winddirection and cloud cover plus infrared radiometer measure-ments. In addition, the model developed by Nikolayev et al.(2001) is used to interpret the data and the results.

Experimental setup

Measurement sites

TikehauThe atoll of Tikehau is an island of the Tuamotu Archipel-ago. It consists of several small islands or ‘‘motus’’. Its totalsurface, including the lagoon, is 448 km2. Two dew siteswere used at Tikehau Island on Motu Tuherahera. The firstsite (labeled # 1) is at the airport (lat 14�07 0 S and long148�14 0 W). It is exposed to the dominant Trade winds.The second site (labeled # 2) is located 300 m west from site# 1 and is protected from the wind by coco trees. Both sitesare 0.5 m above sea level (asl).

The climate is humid tropical with two different seasons:the dry season between May and October and the humid sea-

Figure 1 Data recorded during the night 6–7/9/2005 in Tahition PTFE (time t is UTC-10). Left ordinates: temperatures (�C):PTFE surface temperature Tc, full line; air temperature Ta,small interrupted line; dew point temperature Td, dashed line.Right ordinates: dew mass m (g), crosses; cloud cover (octas),full line; wind speed V (m/s) at 10 m, interrupted line.

Study of dew water collection in humid tropical islands 161

son between November and April (Laurent et al., 2004). Thewind regime is characterized by the permanent presence ofthe Trade winds, with a constant (day and night) dominantENE direction. The average wind speed at 10 m was mea-sured on the Takaroa atoll (300 km east of Tikehau, Tua-motu archipelago) from 1966 to 2002 and was 5–8 m/sduring the night, (Laurent et al., 2004). Wind is strongerduring the dry season.

TahitiThe measurement site is located at Outumaoro on the westside of Tahiti (land area 1042 km2), at latitude 17�34 0 S andlongitude 149�36 0 E at the University of French Polynesia.The elevation is 97 m asl and the face of the hill is orientedwestwards. The climate is also humid tropical with well-contrasted dry and humid seasons.

The site, located on the west coast of Tahiti, is protectedfrom direct ENE Trade winds. Nevertheless, during the dryseason the wind sometimes turns to the SE direction, (‘‘Mar-a’amu wind’’) and becomes strong enough to prevent con-densation on exposed surfaces.

Measurement procedure

Dew volumes were measured on the same reference surfacesin Tikehau and Tahiti: condensers of 1 m · 1 m inclined at a30� angle from horizontal. In Tikehau, the condenser is ex-posed W and remains shaded in the morning (until 08:00, lo-cal time UTC-10H). The condenser in Outumaoro is exposedNE and faces an open sky for radiative cooling. The condens-ers were coated with the same condensing foil, 0.35 mmthick, made of TiO2 and BaSO4 micro spheres embedded inlow density polyethylene with a food surfactant (similar toNilsson, 1996; made by OPUR, 2006). The foil was thermallyinsulated from the condenser frame by a 20 mm thick Styro-foam plate. Increasing the water yield also requires a prop-erly designed condenser (Beysens et al., 2003) to preventheating by the ambient air. This is especially important whena nocturnal wind is present, as is often the case on islands.Dew water was measured daily at 07:30 local time. The totalcondensed volume corresponds to water collected by gravityflow in a bottle and the residual dew drops scraped from thecollecting surface (both amounts were recorded).

TikehauThe following parameters are recorded every day on site at07:30 local time (= UTC-10H): air temperature Ta (�C), windspeed V (m/s) at 6 m elevation, wind direction (degree),cloud cover data N (octas). Rain was excluded from thedataset. Data were collected during the dry season, from21/6/2005 to 7/10/2005.

Recorded meteorological data were available every hourat the nearby automated station at the Rangiroa atoll (NE,63 km): wind speed and wind direction Dir (degree), Ta(�C), relative humidity RH (%) giving the dew point tempera-ture Td (�C). Due to the flatness of the atolls, the absence oftopographical undulations and the close locations of thesetwo places in the axis of the dominant wind, these dataare expected to be also representative of the Tikehau condi-tions. Some deviations, however, occur (as the number ofrainy day, see Fig. 2a) and the correlation of the dew yieldswith those data might then suffer of uncontrolled bias.

TahitiIn Tahiti, dew yield was also measured every morning on thesame 1 m · 1 m foil collector as described above. Dew (andalso to some extent, rain) mass m was also recorded every15 min on a reference plane surface placed at 1 m abovethe ground. This surface is made of a 400 mm · 400 mm(collecting area Sc = 0.16 m2) and e = 1.05 mm thicknessPolyTetraFluororoEthylene (PTFE, commercial name: Tef-lon) plate. The plate is placed on a 12.5 lm thick aluminumfoil, using a 5 mm thick polystyrene foam for thermal insu-lation. It is placed on an electronic, temperature-compen-sated Mettler Toledo balance connected to a PC. Thebalance is protected from the wind up to the plate level.Zero is arbitrary. Note that the wind induces a force direc-ted upwards due to the Bernoulli pressure.

The following physical parameters were recorded every15 min on a data logger connected to a computer: PTFE platesurface temperature Tc from a Type K thermocouple(±0.1 �C), dew mass m (g), air relative humidity, air temper-ature and dew point temperature, foil temperature Tc. Thewind speed was measured by a cup anemometer (stallingspeed: 0.5 m/s) located within 1.50 m from the plate, 3 mfrom the foil collector and 1.6 m above the ground. Windspeed data have been extrapolated at z = 10 m height byusing the classical logarithmic variation (see e.g. Pal Arya,1988):

VðzÞ ¼ V10lnðz=zcÞ=lnð10=zcÞ; ð1Þ

where zc (taken here to be 0.1 m) is the roughness length.Rain (mm), wind direction (degree) and cloud cover dataN (octas) are measured at the Faaa airport meteorologicalstation situated about 2 km NNE.

A typical recording is shown in Fig. 1. When necessary, therealmass is easily inferred from the eventswherewind is near

162 O. Clus et al.

zero. Typically, as soon as the condenser surface tempera-ture Tc < Td in the evening, the balance detects a slow massincrease with a slope dm/dt < 8 g/h. When in the morningTc > Td, evaporation takes place with a very large negativeslope dm/dt. The time of dew production is determined bythe time (dt) where Tc < Td (Fig. 1). This is a simplificationwhereby water is supposed to completely wet the substrate.The total dewmass is taken as the maximum condensed massm0, when Tc reaches Td in the morning.

From m, it is easy to deduce the equivalent dew waterprecipitation h

hðmmÞ ¼ ½m or m0�ðgÞ=160: ð2Þ

The numerical factor corresponds to the condenser surfacearea.

There are no fog or frost occurrences. Rain events aredetected by visual observation or the high value of m and/or its rate dm/dt, which exceeds 5 · 10�2 mm/h. Rainevents were excluded from the dataset. The dew data wereobtained on the balance during the dry season, between 8/6/2005 and 14/9/2005.

In order to evaluate the night-time radiative balance onthe surface of the condenser, the global sky irradiance wasmeasured by two radiometers installed at 5 m off theground, at a distance of 20 m from the condenser and inan open area. The measurements were performed in theband 0.3–3 lm by a pyranometer (CM3 from Kipp & Zonen)and in the band 5–50 lm by a pyrgeometer (CG3 from Kipp& Zonen, with built-in heater to prevent condensation) withboth sensors placed in the same setup. These data were col-lected during the dry season, from 14/7/2005 to 14/10/2005 (unfortunately, some data were lost between August15 and September 24.)

Dew yields

General dew data

Fig. 2 and Table 1 contain rain and dewfall data for Tikehauand Tahiti. Note that the dry season was exceptionally hu-mid during the study period, with more than 50% of rainevents below 2 mm/night. Dew yields above 0.2 mm/nightrarely occur at the Tikehau and Tahiti sites. It is less thanthe maxima that are collected elsewhere, e.g. in the Medi-terranean basin where 0.42 mm/night was measured inAjaccio, Corsica Island (France) (Muselli et al., 2002) andabout 0.6 mm/night in Jerusalem (Israel) (Berkowiczet al., 2004). The Tahiti measurements show a high dew fre-quency, 53%, as compared to 33% in Ajaccio over one year(Beysens et al., 2005), but with a low mean yield equal to0.068 mm. The dew events are relatively stable in amountand in frequency along the measurement period. The cumu-lated dew volume is high (5.58 mm in Tahiti and 3.5 mm inAjaccio during the summer, dry season, see Beysenset al., 2005). Dew measurements in Tikehau at site # 1 dur-ing the first period (June–July) showed low dew collection.In contrast, site # 2 (used later) provided better results withmore frequent dew events that also have good yields. Thedistribution of dew yields in Tikehau (Fig. 2c) and Tahiti(Fig. 2d) shows that the more frequent dew events do notcorrespond to the smallest yields.

Analysis of Tikehau data

Radiation budget and condenser positionSite # 1 at the airport was chosen because the area is ex-posed, giving a maximum radiative transfer to the sky.However, the presence of Trade winds was found unfavor-able for representative data. Indeed, Fig. 2 shows only afew dew events and with small dew yields. During the21/6/2005–8/8/2005 period, only four yields above0.1 mm were recorded and these occurred on unusuallycalm nights. During these 4 days, the wind speed at 10 melevation was below 2.5 m/s. In contrast, the nightly(20:00–7:00) mean wind speed at 10 m elevation measuredin Rangiroa from 2001 to 2005 during the dry season was5.48 m/s with a standard deviation of 3.2 m/s. The per-centage of time with a wind speed below 2.0 m/s is only9.8%. It is known that wind speeds (at 10 m elevation)greater than 3 m/s significantly decreases collected dewvolumes (Muselli et al., 2002). Consequently, it wasdecided to move the condenser to a garden of a house en-closed by trees (6 and 8 m height). The dimensions of theopen area are 44 m · 43 m, located at the end of the air-port area, protected from direct dominant wind with 12rows of coco trees. This situation is representative of mosthouses in the Tuamotu atolls.

In order to evaluate the radiative budget correspondingto the new condenser position, a specific FORTRAN codewas developed (Clus et al., 2006) for the conditions pre-sented in Fig. 3a. The integration of the radiative budgeton elementary solid angles of the hemispherical sky wasperformed in steps of 1 · 1 degrees. The incoming angularsky long wave radiation dIX (W/m2) in an elementary solidangle dX is modeled following Eqs. (3) and (4) (Berger andBathiebo, 2003):

dIX ¼ �h � r� T4a � dX=p: ð3Þ

Here,

�h ¼ 1� ð1� �totalÞ1=ðb cos hÞ; ð4Þ

where r is the Stefan-Boltzmann constant, Ta is the ambienttemperature measured at the ground level (fixed at 288 Kfor the simulations) and etotal is the relative sky emissivity,approximated here to 0.8, a reasonable value for an air with80% RH (Nikolayev et al., 2001). The angular emissivity eh isgiven by the Eq. (4) where h is the angle relative to the ze-nith (vertical) direction and b = 1.66 (Berger and Bathiebo,2003). The angular emissivity is assumed to be eh = 1 if anobstacle (tree, house, relief, etc.) is shading the sky forthe considered dX. The condenser is taken as a gray bodywith emissivity 0.94 (Nikolayev et al., 2001) for the calcula-tion of the angular long wave radiation emitted in the solidangle dX. Both incoming and dissipated power on the sur-face of the condenser are corrected by the tilt angle ofthe condenser when calculating the radiative budget foreach dX.

As shown in Fig. 3a, the condenser is placed in an openarea bordered with 6 m high trees approximately along theS, W, N directions and with 8 m trees along the E direction.The house shades the same band of sky as the trees and canbe ignored. The radiative budget is integrated for differentpositions (X) of the condenser along an axis situated 22 m

Figure 2 (a) Dew yield evolution (top half) collected at the Tikehau (Tuherahera) sites # 1 and # 2. The 55 rain data (bottom half)are from the Rangiroa airport. The 44 rain events were observed in Tikehau. (b) Dew yield evolution from the Tahiti site(Outumaoro). Rain data (bottom half) are from the Faaa airport. (c) Distribution of dew yields at Tikehau. Site # 1: black column;site # 2: gray column. (d) Distribution of dew yields at Tahiti.

Study of dew water collection in humid tropical islands 163

within the S side (Fig. 3b). The position X = 13.5 m corre-sponds to the maximal heat dissipation, �77.8 W/m2. Thisvalue changes only slightly between 10 and 30 m. The same

integration carried out for an open field such as the airportarea gives exactly the same value. There is thus no detect-able energy loss due to the presence of trees.

Table 1 Statistics of dew occurrence

Tikehau sites 1 and 2(21/6/2005–7/10/2005, 109 days)

Tahiti (16/5/2005–14/10/2005,151 days)

Rainfall (cumulative, mm) 247.8 176.2Dewfall (cumulative, mm) 2.64 5.58Dew/rain (%) 1.09 3.17

Minimum dew yield (mm) 0.013 0.004Maximum dew yield (mm) 0.23 0.22Average dew yield (mm/dew event) 0.102 0.068% Dew events (all days) 23.9 53.5% Dew events (excluding rainy days) 38.8 69.8

Figure 3 (a) Site # 2 at Tikehau (schematic). C: 30� tilted condenser with the hollow back side facing the Trade winds. X (m):distance condenser – trees. Grey circles: 8 m high trees. Grey ellipses: 6 m high trees. (b) Simulation of the radiative budget as afunction of X.

Figure 4 Cumulated dew yields at Tikehau with respect towind speed (0.5 m/s steps). Site # 1: black data; site # 2: graydata.

164 O. Clus et al.

Wind speedFig. 4 contains dew amounts collected in both sites withrespect to wind speed. The mean wind speed in Rangiroaduring the dry season nights is <V>10m = 5.5 m/s with astandard deviation of 3 m/s. Thus most of the wind speedsin Rangiroa are above 4 m/s, which is the limiting velocityfor dew formation (Beysens et al., 2005). Fig. 4 stressesthe difficulty to obtain significant dew yields in the freeopen area # 1, even with the condenser back side exposedto the incoming wind. In this configuration, only a few dewevents correspond to V < 4 m/s. In contrast, dew is morefrequent and occurs for V up to 7 m/s in the protected site# 2.

Cloud coverDew was observed at site # 1 for the lowest cloud coveramounts (N 6 3 octas, Fig. 5). During such strong wind con-ditions dew can form only when the cooling power is large;this condition is met only when the sky is clear or N small.

Note that clear sky events are exceptional due to thepresence of high absolute humidity in this tropicalatmosphere.

Figure 5 Cumulated dew yields collected at Tikehau withrespect to cloud cover (octas) averaged during condensationtime. Site # 1: black; Site # 2: gray.

Study of dew water collection in humid tropical islands 165

Analysis of Tahiti data

Wind speedThe dew yield per night on the foil is shown in Fig. 6 withrespect to wind speed for the period 08/06/2005–14/10/2005. Here, the wind speeds have been averaged over thedew duration time (temperature PTFE Tc < Td). The situa-tion of the experimental site, protected from the Tradewinds, explains the low values of wind speeds (79% of thedata have a mean wind speed < 1 m/s).

The largest dew yields occur for wind speeds around0.7 m/s, which also correspond to the most frequent windspeed. The high frequency of dew events (69.8% of thenon-rainy days) as compared to those observed on Tikehau(38.8%, see Table 1) is due to the lower wind speeds asencountered in Tahiti.

Cloud coverIn Tahiti, the largest dew yields (Fig. 7) occur for N 6 2octas. The cloud cover was averaged over the dew forma-

Figure 6 Cumulateddewyields collected atTahiti onOPUR foilwith respect to wind speed measured on the site (corrected for10 m elevation) and averaged during the condensation period.

tion time. The distribution of mean cloud cover (notshown), averaged between 20:00 and 8:00 during all nightsof the same time period, is quite similar to Tikehau andshows many nights with N > 2. A detailed discussionusing the radiometer data is given in Section ‘‘Skyradiation’’.

Relative humidity and dew point temperatureThe relative humidity is a key parameter for dew formation.A high RH corresponds to a small Ta–Td and thus less coolingis needed for dew condensation. Fig. 8 shows the dew yielddata correlated with Td–Ta, or alternatively with ln(RH) asthese quantities are nearly proportional in the studiedrange. PTFE and foil dew data are included. Also shownfor the sake of comparison are foil data obtained with thesame 1 m2 condenser in a Mediterranean Island (Ajaccio)from 31/8/2003 to 8/7/2004. The maxima, selected withina temperature step of 1 K, are also reported in Fig. 8. Nearlyall the data lie below a line that fits these maxima accordingto

h ¼ h0

DT0½DT0 � ðTd � TaÞ�: ð5Þ

DT0 is the maximum cooling temperature and h 0 the maxi-mum dew yield. The data below this line simply means thatTa–Td (or RH) is the main parameter that limits the dewyield. Of course other factors such as wind speed and cloudcover also matter but they decrease the dew yield from thisline of maxima.

Tahiti data and Tikehau data (not shown) give nearly thesame results when fitted to Eq. (5). Concerning DT0, the val-ues for PTFE (�4.3 K) and foil (�4.7 K) correspond to a rel-ative humidity RH � 75%. The yield h 0 = 0.26 mm (PTFE) orh 0 = 0.30 mm (foil) are similar. In contrast, the Ajaccio re-sults appear markedly different as DT0 = �10.3 K(RH � 51%) and h 0 = 0.36 mm. Cooling power and meandew rate are then larger in Ajaccio than in Tahiti. Note thatno dew events were found in Tahiti for 0 < Ta–Td < 1.5 K or100 > RH > 91%. (in contrast to Ajaccio). This observation isexplained below.

Figure 7 Cumulated dew yields collected at Tahiti withrespect to cloud cover averaged during the condensation time(08/06/2005–14/10/2005).

Figure 8 Dew yield data with respect to Td–Ta (lower abscissa in K) or RH (upper abscissa, log scale, in %). (a) Tahiti PTFE data, (b)Tahiti foil data, (c) Ajaccio foil data. The maximum yields in Td–Ta steps of 1 K are bound by lines. The best fits of these data areshown by straight lines.

166 O. Clus et al.

Discussion

Two features of the humid tropical climate in the study re-gion are strong Trade winds and high absolute humidity.These characteristics are at the origin of the particularitiesof dew formation and collection.

Wind speed and humidity

The dew yield study at the two Tikehau sites demonstratesthat protecting the condenser from direct wind is very effi-

cient for improving dew yield. However, wind has other ef-fects on condensation.

Firstly, wind enhances the heat exchange with air (theNusselt number increases). At the first site (airport), a meanwind speed as large as 5.48 m/s was too high to permit suf-ficient cooling and frequent condensation (Fig. 4). However,the second site, protected by coco trees, allows higher dewyields to be observed. The influence of coco trees on thewind profile can be described as (i) a permeable windbreakwith a highly porous base from 0 to 5 m and (ii) an imperme-able dense leaf layer from 5 to 8 m. For a windbreak com-

Study of dew water collection in humid tropical islands 167

posed of a permeable basis (e.g. tree trunk) with an imper-meable superior level (e.g. tree foliage), Guyot (1999) no-ticed a decline of more than 50% in wind speed up to adistance of nine times the windbreak height, a result inde-pendent of the wind speed.

Secondly, wind speed has an influence on the low atmo-spheric layers and on local relative humidity. Dew can formonly when the condenser surface is cooled below the dewpoint temperature. What thus really matters is the gapTa–Td � ln(RH). The tropical oceanic atmosphere is very hu-mid with an average total atmospheric water vapor columnclose to 50 mm (American Geophysical Union, 2006). Strongwind (5.5 m/s on average) ensures that the surface atmo-spheric layer (close to the ground) is well-mixed and thatthe temperature profile close to the ground (in the first15 m layer) is adiabatic (Guyot, 1999). Under such condi-tions, RH is quite independent of V. As a result, significantdew yields can be found even with high wind speed if thecondenser is properly protected from convective heating.Fig. 9a displays RH data for years 2001–2005 with respectto mean wind speed during clear nights only (mean cloudcover N 6 2) at the Faaa (Tahiti) meteorological station.Selecting the clear nights excludes in a simple way rainand storm events. It shows that significant RH (>70%) canbe maintained for V up to 7.5 m/s (10 m elevation).

Such strong mixing of the surface atmospheric layer,however, smoothens the day and night RH variation and thuslimits RH to values of the order of 90%, as outlined above insection Analysis of Tahiti data. For example, 5 years of mea-surements (2001–2005) at the Rangiroa atoll during the dryseason (June–October, Laurent et al., 2004) shows a night(18:00–06:00) average RH = 77% and a day (07:00–19:00)RH = 70%.

For the sake of comparison, Fig. 9b shows the RH varia-tion with mean wind speed for a typical Mediterranean site(Ajaccio). Here, RH > 70% only for V < 4.5 m/s. This findingagrees with dew yields declining sharply for V > 3 m/s andthe near absence of dew for V > 4 m/s (Muselli et al.,2002). In such climates (Mediterranean, semi-arid), dryerthan the tropical climates, large differences in temperatureare observed between day and night due to the cooling of

Figure 9 Mean relative humidity (%) correlated with mean wind sto 8:00) and mean cloud cover N 6 2. (a) Faaa airport (Tahiti, FreFrance), 2004 data.

the planetary limit atmospheric layer in contact with theradiative cooling of the soil. The variations are importantin a 500 m thick layer, with a thermal inversion in the sur-face layer at 15 m above the ground (Guyot, 1999). Theinversion temperature profile can reach 5–10 �C in the sur-face layer. In such climates, clear night temperature profileis not adiabatic. The important radiative cooling of soil in-duces high RH enhancement at the ground level that canreach nearly 100% when wind speed is low enough, as ob-served above in Fig. 8c. In contrast, increasing wind speedis an important cause of decreasing RH in mixing the surfacelayer and homogenizing the temperature profile and ex-plains why dew hardly forms even in protected areas.

Sky radiation

The absolute power Psky received on the condenser due to IRsky radiation cannot give precise elements for comparisonwith other sites as it depends strongly on local temperatureand air absolute humidity. What really matters in dew for-mation is the cooling power

Pr ¼ Pc � Psky; ð6Þ

that is, the difference between the condenser power Pc andthe sky radiative power Psky. The latter is measured with thetwo radiometers (bands 0.3–3 lm and 5–50 lm). The con-denser emission power can be estimated by the Stephan–Boltzmann law such as

Pc ¼ Sc�crðTc þ 273Þ4; ð7Þ

where Sc is the condenser surface, r is the Stephan–Boltz-mann constant and ec is the emissivity of the condenser foil(=0.94 from Nikolayev et al., 2001).

In Fig. 10a, the mean cooling power <Pr> is comparedwith the dew yields in Tahiti and the mean relative humidity<RH> averaged during the condensation period (the periodwhere Tc < Td). For comparison, the same quantities areplotted in Fig. 10b for a Mediterranean climate (Ajaccio).Strikingly, the individual dew yields are on average threetimes lower in Tahiti than in Ajaccio.

peed (10 m elevation) during clear night conditions (from 20:00nch Polynesia) from 2001 to 2005. (b) Ajaccio (Corsica Island,

Figure 10 Dew yield on foil (black dots, left ordinate) and mean relative humidity <RH> (open circles, right ordinate) with respectto the mean cooling power (<cooling Power> as measured during the condensation time. (a) Tahiti site, from 14/7/2005 to 14/10/2005. (b) Ajaccio site, Corsica Island (France)), from 20/9/2004 to 3/2/2005.

168 O. Clus et al.

A first explanation can be found in the lower coolingpower, which is 9% lower at Tahiti on average. This lowerpower is due to the larger content of humidity in the atmo-sphere. The total vapor equivalent water column is indeed50 mm for the Tahiti area and 16 mm in the Ajaccio area(American Geophysical Union, 2006). However, the mostimportant effect is due to RH, which, when dew forms, islower on average in Tahiti. In addition, strong winds preventlarge values to be reached at night as outlined in Section‘‘Wind speed above’’.

In contrast, the smaller amount of vapor in Ajaccio is dis-tributed closer to the ground, permitting ground radiativecooling and larger RH in the lowest atmospheric layer (Guyot,1999). As a result, RH is larger on average by 4% in Ajaccio. Onthe Mollier condensation diagram that relates Td to RH, thisdifference inRH corresponds to aTd lower by almost 1 K in Ta-hiti. This lower Td necessitates further cooling and the corre-sponding energy is lost for condensation (see Fig. 8 where thedew yield decreases very rapidly with increasing Ta–Td).

This lower RH and the lowest cooling energy availabledue to enhanced green house effects explain why dew wateryields are less in these tropical climates although the air,paradoxically, contains more humidity.

Modeling and comparison with other sites

In order to further compare dew formation in tropical is-lands with other places (Mediterranean islands and Euro-pean continental locations), data were fitted to thenumerical model developed by Nikolayev et al. (2001). Thismodel is based on the model by Pedro and Gillespie (1982)and Nikolayev et al. (1996) and is valid only during the nightperiod (no evaporation).

It is based on the heat balance equation

dTc

dtðMcc þmcwÞ ¼ Pr þ Phe þ Rcond: ð8Þ

Here, M and m are the condenser and condensed watermass, respectively, cc and cw are the specific heat of thecondenser material and of the water, and t represents time.

The variables in the right side of the equation represent thedifferent thermal processes involved in the heat transfer atthe condenser level: Pr for radiative transfer, Phe for heatconvective exchange with the ambient air and Rcond forthe energy gain due to the condensation latent heat Lc byunity of mass. Thus,

Rcond ¼ Lcdm

dtð9Þ

and for Phe

Phe ¼ Sc aðTa � TcÞ ð10Þ

where a is a thermal transfer coefficient. The parameter a iscorrelated with the boundary layer thickness and dependson the wind speed in the case of laminar flows (the mostcurrent situation in this study) by:

a ¼ kfffiffiffiffiffiffiffiffiffiV=D

p: ð11Þ

Here, the parameter f is empirical; f = 4 W K�1 m�2 s1/2

for a flat plane with the dimension D ¼ffiffiffiffiScp

(Pedro and Gil-lepsie, 1982). A correction coefficient k, of order unity, hasbeen introduced to account for the relative position of thecondenser in relation to the anemometer.

The equation representing the massm is described by thecondensation rate

dm

dt¼

Scb½psatðTdÞ � pcðTcÞ� if positive

0 if negative

�ð12Þ

where psat(T) is the saturation pressure at temperature T,pc(Tc) is the vapor pressure on the condenser dependingon the humidification state of the substrate before conden-sation, b is the mass transfer coefficient which is propor-tional to the parameter a

b ¼ 0:656 ga=ðpcaÞ: ð13Þ

Here, p is the atmospheric pressure (considered as a con-stant during a simulation), ca is the air specific heat, g is anadjustable parameter of order unity, added for the simula-tion, which takes into account the particular conditions ofthe air flow around the condenser.

Study of dew water collection in humid tropical islands 169

In the model, the sky irradiation energy (long-wave radi-ation term) is estimated by neglecting the effects on skytransmittance of aerosols, green house gases, etc. Thisapproximation is justified by the fact that the more impor-tant green house gas is water, whose presence is alreadyconsidered in the model by the cloud coverage N (see belowEq. (15)). Thus the equation is (Pedro and Gillespie, 1982and Campbell, 1977)

Rl ¼ Sc � �c � �s � rðTc þ 273Þ4: ð14Þ

Here, es is the emissivity of the sky that depends on theambient temperature Ta and on the cloud cover N (in octas)

es ¼ es0 þ N½1� es0 � 8=ðTa þ 273Þ�; ð15Þ

where es0 = 0.72+0.005Ta. As N is usually obtained on a 3-hourly observation basis, a linear interpolation is performedto calculate it for smaller (15 min) time intervals.

Input data of the model are t, V, N, Ta, RH, Tc and mexp,the experimental mass of the condensed water. The modeldetermines the simulated temperature of the substrate Tc,fitwith the adjustable parameter, k, by minimizing the differ-ence between simulated and experimental data. Firstly, theparameter g is fixed as it has a negligible influence on thetemperature Tc and k is determined. Then, with the givenvalue of k, the difference between mexp and mfit is mini-mized by adjusting the parameter g. Here, the fit is con-

Figure 11 The heat (k) and mass (g) transfer coefficientsobtained from fitting the dew data according to the model ofSection ‘‘Modelling and comparison with other sites’’.

Table 2 Dimensionless heat and mass exchange coefficients as owith other sites’’

Site Date Heat exc

Tahiti 20/06/2005–2/10/2005 3.9 ± 0.2Ajaccio 10/09/1999–28/11/2002 3.57 ± 0.0Bordeaux 14/10/2001–14/01/2003 2.54 ± 0.0Grenoble 4/25/2000–6/7/2001 2.52 ± 0.1

The uncertainty corresponds to one standard deviation.

cerned with the 1 m2 condenser whose nightly cumulatedyield is measured in the morning (one data of dew water vol-ume available mexp_total). The final simulated dew mass ismodified by the following equation that enables the valueof g to be determined:

mfitðg; time ¼ timefinalÞ ¼ mexp total ð16Þ

The results of the fits in terms of both parameters k andg are reported in Fig. 11 with respect to date. The valueskeep roughly constant, as expected. The mean values arelisted in Table 2 and compared with values obtained withsites at nearly the same latitude in Europe (42�–45� N)for Ajaccio and two continental locations: Bordeaux (Atlan-tic coastal area, France) and Grenoble (Alpine valley,France). The values of the coefficients k and g in Tahitiare in reasonably good agreement with those from Ajaccioin spite of the large differences in dew yields and climate.This means that the radiative power and the wind effectshave been properly taken into account in the model. Theheat and mass transfer parameters for both islands are,however, systematically larger than for the continentallocations. The reason can be found in the high wind speedin islands, in contrast to weak wind speed currently foundin continental locations that correspond to free convec-tion. This situation, which is characterized by weak, turbu-lent air convection, is indeed not completely satisfactorilytaken into account in Eq. (11) of the model (Beysens et al.,2005). Its complete description is out of the scope of thepresent study.

Concluding remarks

This study was motivated by the evaluation of dew water asa supplementary water resource in water-scarce low islands(atolls). In such islands, the need for fresh water is high,especially during the dry season. For the 2005 dry seasonin French Polynesia, the dew water resource was found tobe around 5.6 mm. This amount (3% of rain water) remainslow when compared to precipitations during the particularlyrainy season that was studied but is nearly twice what is cur-rently found in Mediterranean Islands during the summer dryseason. It corresponds to a significant amount of water(560 L on a modest 100 m2 roof).

Dew formation in a tropical humid climate shows inter-esting and paradoxical features when compared to other cli-mates. The presence of strong Trade winds and the highabsolute humidity of the atmosphere do not allow dew ratesin tropical climates as large as those from some other re-gimes (e.g. the Mediterranean climate). In low islands or

btained from the model of Section ‘‘Modeling and comparison

hange coefficient (k) Mass exchange coefficient (g)

0.23 ± 0.023 0.37 ± 0.0256 0.12 ± 0.0233 0.13 ± 0.04

170 O. Clus et al.

wind-facing sides of high islands, the condensing surfaceshould be protected from the strong direct Trade winds,e.g. by a natural vegetal windbreak.

The mean yield per dew event is about 0.1 mm in Tahiti(tropical climate), to be compared with a mean of about0.17 mm in Ajaccio (Mediterranean climate). In contrast toMediterranean climate, the most important yields occurwhen the air is the driest such as to provide the highestcooling power.

Even though the mean yields are modest, dew eventsare very frequent because of the continuous high relativehumidity level. This makes the cumulated volume impor-tant and dew harvesting by radiative cooling becomes use-ful as a complementary water resource. This is especiallytrue for the periodic drought conditions where no signifi-cant rain occurs from April to September and where thepopulation is 100% dependant on external water supplies.Simple adjustments of rain collection areas (roofs) couldbe made to harvest dew in addition to rain as already beencarried out in Croatia (Beysens et al., 2007) or India (Sha-ran et al., 2007). Radiative and hydrophilic coatings orpaints (see OPUR, 2006) have been used there, with theadditional benefit of providing efficient cooling under sun-ny conditions.

Dew harvesting thus appears as an attractive possibilityto provide a complementary water resource and could be-come quite essential for extreme events such as droughts,cyclones and tsunamis.

Acknowledgments

Funding support is gratefully acknowledged from the‘‘Ministere Francais de l’Outre-Mer’’ and ‘‘CollectiviteTerritoriale de Corse’’. We are indebted to S. Berkowiczfor his critical reading, kind remarks and relevant sugges-tions. We are also thankful to the Natua family for havingkindly welcomed us. We also thank R. Matehau and D. Teivafor their help in the data collection. We express gratitude tothe French Polynesia branch of Meteo France for providingthe meteorological data.

References

American Geophysical Union, 2006. <http://www.agu.org/sci_soc/mockler.html>.

Barradas, V.L., Glez-Medellın, M.G., 2000. Dew and its effect ontwo heliophile understorey species of a tropical dry deciduousforest in Mexico. Agric. Forest Meteorol. 104, 303–313.

Berger, X., Bathiebo, J., 2003. Directional spectral emissivities ofclear skies. Renew. Energ. 28 (12), 1925–1933.

Berkowicz, S., Beysens, D., Milimouk, I., Heusinkveld, B.G., Muselli,M., Wakshal, E., Jacobs, A.F.G., 2004. Urban dew collectionunder semi-arid conditions: Jerusalem. In: Proceedings: ThirdInternational Conference on Fog, Fog Collection and Dew, CapeTown, South Africa, October 11–15, 2004.

Beysens, D., 2006. Dew nucleation and growth. C.R. Physique 7,1082–1100.

Beysens, D., Milimouk, I., Nikolayev, V., Muselli, M., Marcillat, J.,2003. Using radiative cooling to condense atmospheric vapor: astudy to improve water yield. J. Hydrol. 276, 1–11.

Beysens, D., Muselli, M., Nikolayev, V., Narhe, R., Milimouk, I.,2005. Measurement and modeling of dew in island, coastal andalpine areas. Atmos. Res. 73, 1–22.

Beysens, D., Ohayon, C., Muselli, M., Clus, O., 2006. Chemical andbacterial characteristics of dew and rain water in an urbancoastal area (Bordeaux, France). Atmos. Environ. 40 (20), 3710–3723.

Beysens, D., Clus, O., Mileta, M., Milimouk, I., Nikolayev, V., 2007.Collecting dew to improve water resources: the D.E.W. projectin Bisevo (Croatia). Energy 32, 1032–1037.

Campbell, G.S., 1977. An Introduction to Environmental Biophysics.Springer Verlag, New York.

Clus, O., Muselli, M., Beysens, D., Nikolayev, V., Ouazzani, J., 2006.Computational fluid dynamic (CFD) applied to radiative cooleddew condensers. Proceedings of 5th IEEE International Sympo-sium on Environment Identities and Mediterranean area, Corte,France, July 2006 (accepted, 2007).

Dupon, J.F., 1987. Les atolls et le risque cyclonique, le cas desTuamotu. Cahier des Sciences Humaines 23 (3–4), 567–599.

Gindel, I., 1965. Irrigation of plants with atmospheric water withinthe desert. Nature 5002, 1173–1175.

Guyot, G., 1999. Climatologie de l’Environnement, second ed.Dunod, France (pp. 507).

Jacobs, A.F.G., Heusinkveld, B.G., Berkowicz, S.M., 2008. Passivedew collection in a grassland area, The Netherlands. Atmos. Res.87, 377–385.

Khedari, J., Waewsak, J., Thepa, S., Hirunlabh, J., 2000. Fieldinvestigation of night radiation cooling under tropical climate.Renew. Energ. 20 (2), 183–193.

Laurent, V., Maamaatuaiahutapu, K., Maiau, J., Varney, P., 2004.Atlas Climatologique de la Polynesie Francaise. Meteo, France(pp. 201).

Lhomme, J.-P., Jimenez, O.F., 1992. Estimating dew duration onbanana and plantation leaves from standard meteorologicalobservations. Agric. Forest Meteorol. 62 (3-4), 263–274.

Monteith, J.L., 1957. Dew. Quart. J.R. Meteorol. Soc. 83, 322–341.Muselli, M., Beysens, D., Marcillat, J., Milimouk, I., Nilsson, T.,

Louche, A., 2002. Dew water collector for potable water inAjaccio (Corsica Island, France). Atmos. Res. 64, 297–312.

Muselli, M., Beysens, D., Milimouk, I., 2006a. A comparative studyof two large radiative dew water condensers. J. Arid Environ. 64,54–76.

Muselli, M., Soyeux, E., Beysens, D., Clus, O., 2006b. Is dewpotable? Study of the physical, chemical and biological dewcharacteristics in Ajaccio (France). J. Environ. Qual. 35, 1812–1817.

Nikolayev, V., Beysens, D., Gioda, A., Milimouk, I., Katiushin, E.,Morel, J.P., 1996. Water recovery from dew. J. Hydrol. 182, 19–35.

Nikolayev, V.S., Beysens, D., Muselli, M., 2001. A computer modelfor assessing dew/frost surface deposition. In: R.S. Shemenauer,H. Puxbaum (Eds.), Proceedings of Second International Confer-ence on Fog and Fog Collection, IRDC, St. John’s, Canada, pp.333–336.

Nilsson, T., 1996. Initial experiments on dew collection in Swedenand Tanzania. Sol. Energ. Mater. Sol. Cells 40, 23–32.

OPUR: ‘‘International Organization for Dew Utilization’’, 2006.<www.opur.u-bordeaux.fr>.

Pal Arya, S., 1988. Introduction to Micrometeorology. AcademicPress, San Diego.

Pedro, M.J., Gillespie, T.J., 1982. Estimating dew duration. I.Utilizing micrometeorological data. Agric. Meteorol. 25, 283–296.

Salau, O.A., Lawson, T.L., 1986. Dewfall features of a tropicalstation: the case of Onne (Port Hartcourt), Nigeria. Theor. Appl.Climatol. 37 (4), 233–240.

Study of dew water collection in humid tropical islands 171

Sharan, G., Beysens, D., Milimouk-Melnytchouk, I., 2007. A study ofdew water yields on galvanized Iron roof in Kothara (North-WestIndia). J. Arid Environ. 69, 259–269.

Steinberger, Y., Loboda, I., Garner, W., 1989. The influence ofautumn dewfall on spatial and temporal distribution of nema-todes in the desert ecosystem. J. Arid Environ. 16, 177–183.

Vargas, W.E., Lushiku, E.M., Niklasson, G.A., Nilsson, T.M.J., 1998.Light scattering coatings: theory and solar applications. Sol.Energ. Mater. Sol. Cells 54, 343–350.

Weihong, L., Goudriaan, J., 2000. Dew formation on rice undervarying durations of nocturnal radiative loss. Agric. ForestMeteorol. (4), 303–313.