Dew water collector for potable water in Ajaccio
(Corsica Island, France)
, Daniel Beysens
*, Jacques Marcillat
, Torbjo¨rn Nilsson
, Alain Louche
´de Corse, UMR CNRS 6134, Route des Sanguinaires, 20000 Ajaccio France
International Organization for Dew Utilization, 26, rue des Poissonniers, 33600 Pessac France
Equipe du Supercritique pour l’Environnement, les Mate
´riaux et l’Espace, SBT, CEA-Grenoble, France
Laboratoire de Mode
´lisation et de Simulation Nume
´rique en Me
Technopole de Chateau-Gombert, 13451 Marseille Cedex 20 France
Equipe du Supercritique pour l’Environnement, les Mate
´riaux et l’Espace, CNRS-ESEME, ICMCB,
87, Av. du Dr. Schweitzer, 33608 Pessac France
¨gen 3, SE-434 46 Kungsbacka Sweden
Received 30 July 2001; received in revised form 15 February 2002; accepted 25 March 2002
We report on the development of an inexpensive radiative condenser for collecting atmospheric
vapor. Based on the experience gained using a small working model in Grenoble (France), a
prototype of 103m
was established in Ajaccio (Corsica, France). The condensing surface is a
rectangular foil made of TiO
microspheres embedded in polyethylene and has an angle
of 30jwith respect to horizontal. The hollow part of the device, thermally isolated, faces the
direction of the dominant nocturnal wind. Dew measurements were correlated with meteorological
data and compared to dew condensed on a horizontal polymethylmethacrylate (PMMA, Plexiglas)
reference plate. The plate served as a reference standard unit and was located nearby. Between July
22, 2000 and November 11, 2001 (478 days), there were 145 dew days for the reference plate (30%),
but 214 dew days for the condenser (45%). This yield corresponds to 767 l (3.6 l, on average, per
dew day). The maximum yield in the period was 11.4 l/day. Dew mass can be fitted to a simple
model that predicts dew production from simple meteorological data (temperature, humidity, wind
velocity, cloud cover). Chemical analyses of the water collected from the plate were performed from
October 16, 1999 to July 16, 2000 and from the condenser, from July 17, 2000 to March 17, 2001.
The following parameters were investigated: suspended solids, pH, concentration of SO
0169-8095/02/$ - see front matter D2002 Elsevier Science B.V. All rights reserved.
PII: S 0169-8095(02)00100-X
Corresponding author. Commissariat a l’Energie Atomique-Equipe du Supercritique pour l’Environnement,
les Materiaux et l’Espace (CEA-ESEME), ICMCB, 87, Av. du Dr. Schweitzer, 33608 Pessac Cedex France.
Tel.: +33-5568-462-98; fax: +33-5568-427-61.
E-mail address: firstname.lastname@example.org (D. Beysens).
Atmospheric Research 64 (2002) 297 – 312
ions. Only Cl
ions were sometimes found significant. Wind direction analyses
revealed that Cl
is due to the sea spray and SO
to the combustion of fuel by an electrical plant
located in the Ajaccio Gulf. Except for a weak acidity (average pHc6) and high concentration of
suspended solids, dew water fits the requirements for potable water in France with reference to the
D2002 Elsevier Science B.V. All rights reserved.
Keywords: Water production; Chemical analysis; Hydrometeorology; Atmospheric deposition
Collecting dew water from the atmosphere has been the goal of several studies, with
mitigated success (Jumikis, 1965; Nikolayev et al., 1996; Beysens et al., 1998; Alnaser
and Barakat, 2000). In Feodosia (Crimea, Ukraine), there are remnants of a huge dew
condenser erected in 1912 by F.I. Zibold. His efforts were followed by several attempts in
France, in Trans-en-Provence by Knapen (1929) and in Montpellier by Chaptal (1932).
The yield of Zibold’s condenser was reported to have reached 350 l/day (Jukov, 1931),
while the other condensers, more massive, did not exceed a few liters per day. In such
massive condensers, where the (latent) heat brought by condensation does not appreciably
warm up the condensing wall because of its large specific heat, the wall temperature rarely
falls below the dew point temperature. The Zibold condenser, because of its condensing
area open to the sky, could, nevertheless, be cooled by radiative transfer, thus allowing a
better dew yield (see the analysis by Nikolayev et al., 1996).
More recently, a small condenser made with a special foil 1.21.2 m
, cooled by
radiative heat transfer, was tested in Sweden and Tanzania with some success by coauthor
Nilsson (1996). In the present study, we report on the dew yield and chemical analyses of
water obtained in a much larger (310 m
) collector constructed and tested for 1 year in
the Ajaccio Gulf (Corsica Island, France). Corsica exhibits a climate typical of a
Mediterranean Island. According to our visual observation, we mention that fog never
occurred during the measurement period.
2. Preliminary studies
In the formation of dew (Monteith, 1957; Beysens, 1995), meteorological parameters,
such as air temperature, air humidity (dew temperature), and sky radiation (cloud cover),
cannot be controlled. It is, however, possible to increase the yield of dew harvesting by
modifying (i) the emitting properties of the condensing surface (foil), (ii) the wind velocity
on the foil, (iii) the condensation time, and (iv) the recovery of the water drops. Previous
studies by Nilsson et al. (1994) and Beysens et al. (2001) allowed us to determine the main
characteristics required for a high-yield dew collector, taking into account the above
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312298
The foil is made of TiO
microspheres embedded in a sheet of polyethylene.
This material is discussed in Nilsson et al. (1994) and explains why this foil exhibits
improved emitting properties in the near infrared (to provide radiative cooling of room
temperature surfaces) and efficiently reflects the visible (sun) radiation.
A weak wind (<1 m/s) is necessary to provide sufficient humid air around the
condenser. Strong wind, which increases heat losses, cancels the radiative cooling. From
our initial studies (Beysens et al., 2001), we found that we could minimize wind
influence and recover water drops by gravity using a plane condensing area with an
angle hwith respect to horizontal, thermally isolated from the ground with 2-cm-thick
polystyrene foam. The angle favors the sliding of dewdrops, with obviously a better
yield for angles approaching 90j. However, as the effective radiation surface diminishes
with h, a compromise had to be found. Experiments performed with a model at 1:10
scale, coupled with numerical studies of the air flow perpendicular to the condenser
main axis, revealed that an angle h=30jwas suitable. The thermal gain, as measured by
the reduced parameter
as the condenser temperature, T
the air temperature, and T
the reference plate
temperature), can reach more than 20% when compared with a horizontal surface taken
as a reference. The wind velocity is also much reduced near the surface of the foil if the
latter is directed with the hole facing the nocturnal dominant wind.
3. The condenser
According to these initial studies, we established (Fig. 1a) a10
on the side of a hill. The site is in the Ajaccio Gulf (Corsica Island, France; latitude:
41j55VN; longitude: 8j48VE), 400 m from the sea at 70-m elevation, on the mid-slope of
a hill. The wind regime is characterized by a nocturnal wind with a NE dominant direction
(1.8 m/s average) but with two directions (NW/SW) for the diurnal dominant wind,
characteristic of a Mediterranean Island climate.
The foil is fixed by lateral cables on a light grid attached by cables (Fig. 1b). The cables
are fixed to beams anchored to the ground. Between the foil and the grid are sandwiched
3-cm-thick polystyrene foam plates to provide thermal isolation. Water is gathered by
gutters and a 25-l polyethylene tank.
The hollow part of the device faces the direction of the dominant nocturnal wind
(Fig. 1a,b). The condenser faces SW to remain shaded longer in the morning. It is
precisely in the early morning that the air temperature is the lowest, thus the closest to
the dew point temperature, favoring dew condensation. With such a configuration, we
experienced dew formation even in the daytime, so long as the sun did not directly
irradiate the foil.
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312 299
3.2. Water yield
We compared the data obtained on the condenser, and from a nearby reference plate
located within 40 m on a terrace at 7 m above the ground (see Figs. 2 and 3). The reference
Fig. 1. (a) The 310 m
dew condenser at Ajaccio (Corsica Island, France). F: foil; T: water collection tank. (b)
The condenser. T
are temperature sensors taped on the foil. The arrow indicates the dominant nocturnal wind
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312300
Fig. 2. Dew harvested on the condenser (black bars) compared to the reference plate (white bars).
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312 301
plate was made of a 5 mm thick, 0.40.4 m
plate of Plexiglas (polymethylmethacrylate,
abbreviated here as PMMA) placed on a thermally isolated mount formed of an aluminum
foil in contact with the plate and a 5-mm-thick sheet of polystyrene foam beneath the foil.
Dew point temperature, air temperature, wind velocity (at 10 cm above the plate), and
wind direction (at 3 m above the plate and 10 m above the ground) were measured every
15 min. During the period July 22, 2000–November 11, 2001 (478 days) (date is that of
the morning), we experienced 145 dew days for the reference plate (30%) and 214 dew
days for the condenser (45%). The condenser thus provides a gain of 148% in dew days.
The condenser yield corresponds to 767 l, i.e., an average of 3.6 l (0.12 mm) per dew day.
The maximum yield in the period was 11.4 l (0.38 mm).
The comparison of the histograms of dew volumes obtained on the reference plate and
the condenser (Fig. 3) shows that dew gained by the condenser corresponds mainly to the
average dew yield, of the order of 0.12 mm.
3.3. Water drop recovery
Most of the water was recovered in the tank by gravity. In the morning, we systematically
wiped the drops remaining on the foil and compared the volume of water before and after
wiping. The results are reported in Fig. 4. The average scraped water is 1.2 l/day. In other
words, without wiping, the sun evaporation would decrease the yield by the same factor.
3.4. Wind influence
In Fig. 5, we report on the dew yield vs. wind speed in the morning (period July 22,
2000–July 23, 2001). Wind speed, as measured at 10 m off the ground within 40 m from
Fig. 3. Histogram of dew yields corresponding to the data in Fig. 2. (condenser: black bars; reference plate: white
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312302
Fig. 4. Volume of water remaining on the foil in the morning and wiped off. The average value is 1.24 l (black line).
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312 303
the condenser, is averaged between 6:00 and 7:00 a.m. (local time). Nearly all the data
correspond to V<4.5 m/s, this value corresponding to a relatively sharp transition. This
speed of 4.5 m/s is apparently a crossover between dew and zero or rare dew. Producing
dew water till such relatively large wind speeds is the result of the special design of our
3.5. Influence of cloud cover
We correlated our data with cloud cover (N, in octas) as measured within 10 km at the
Ajaccio airport meteorological station (period July 22, 2000 – November 30, 2000). It is
clear from Fig. 6 that the maximum yield decreases strongly with N, approximately as
with hin mm. A correlation between hand (8N) is expected because (8N) corresponds
to the scattering solid angle. One should note that the maximum yield corresponds to about
Nc1 and not N=0. This is presumably because such zero cover corresponds to a drier
3.6. Fit of the dew data
Nikolayev et al. (1996) proposed a method to determine the amount of condensed dew
from the measurement of only a few parameters: surface and ambient air temperature, air
humidity, cloud cover, and wind velocity. When fitting the data, it appears that the ideal
heat and mass transfer coefficient do not represent the transfer in the real situation. In order
Fig. 5. Dew yield (mm) correlated with the wind velocity between July 22, 2000 and July 23, 2001.
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312304
to account of these particular situations, Nikolayev et al. (2001) introduced two new
numerical factors of order unity that correct the transfers (kfor the heat transfer and gfor
the mass transfer). Fitting the surface temperature data gives k, fitting the mass data gives
g. These numerical factors are constant for a given condenser and are determined
We show in Fig. 7a the result of a fit of the condenser surface temperature (k=1.12) and
dew mass ( g=0.74) on the PMMA plate, and the fit obtained on the foil condenser
(k=1.51, g=1.15). For the latter, we consider only two data points for the dew mass
corresponding to the beginning of dew condensation, and the time of dew collection.
3.7. Chemical properties
Sampling of dew or rain was performed each day on (i) a PMMA plate strictly identical
to the reference condensation plate located within 2 m from it and (ii) the condenser.
During the period October 16, 1999 – July 16, 2000, we analyzed on the PMMA plate 31
samples of dew and 43 of rain. During the period July 17, 2000 – March 17, 2001, we
analyzed from the condenser 99 samples of dew and 41 samples of rain.
The chemicals in both the PMMA plate and the condenser were expected to give the
same results and was confirmed by sampling of both sites (see data between September 2
and 14, 2000 in Fig. 8).
Fig. 6. Dew yield (mm) vs. cloud cover N(octas) between July 22, 2000 and November 31, 2000. The dotted line
corresponds to Eq. (2) (see text).
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312 305
The samples were stored at 4 jC and analyzed between 1 week and 3 months. The
following parameters were investigated: pH, nonfilterable residue (NFR, also called
suspended solids), concentration of SO
ions. Although the pH may
Fig. 7. Example of data fit with the program of Nikolayev et al. (2001) concerning the dew mass collected during
the night of December 13 –14, 2000. The wind velocity is measured at 10 m from the ground. T
condensing surface temperature. (a) PMMA reference plate, with k=1.12 and g= 0.74. The data points are
recorded every 15 min. (b) Foil condenser, with k=1.510 and g=1.150. The data points are recorded every 15
min, except for the dew mass (+), measured at the beginning and the end of condensation.
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312306
have been affected because of the storage period, we did not find any correlation between
the pH values and the storage time. We also did not notice any correlation with the
collected water volume.
We also checked our procedure by analyzing distilled water, as if it were dew or rain
water, by pouring it on the PMMA plate. The measurement gave a pH=6. That the pH was
not 7, as expected for pure water, is not surprising; pH can be influenced by even a very
Fig. 8. Comparison of the pH values and some ions concentration as measured on the foil collector and on the
PMMA reference plate. Y-axis labels are the ratios pH (foil)/pH (PMMA) and concentration (foil)/concentration
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312 307
small amount of ions in solution. The concentration and measured parameters obtained
from dew and rainwater are drawn in Figs. 9 and 10 and summarized in Table 1.
Note (Fig. 9) that the pH of rainwater is more widely distributed than that of dew water.
It is also, on average, more acidic (rain: 5.6; dew: 5.85). Dew is contaminated only by
local ions, in contrast to rainwater that can be contaminated in regions located quite far
from the sampling location. That dew is less acidic than water is a quite peculiar situation,
Fig. 9. Time variation of the pH of dew (o) and rain (.). The curves correspond to the averaged data: dew (- - -),
rain (—). The same measurement procedure with distilled water gave a pHc6.
Some time-averaged physicochemical properties of dew and rainwater collected in Ajaccio (October 16, 1999 –
March 17, 2001) and comparison with potable water data in France
Rain (average) 5.6F0.1 8F1.3 8.6F1.7 84F51 1.2F0.3 2.3F0.5
Rain (min/max) 3.8/7 0/78 0/64 0/4250 0/13 0/14
Dew (average) 5.85F0.05 119F52.1 28.2F6.1 95F25.6 2.7F0.5 4.7F1.6
Dew (min/max) 4.4/7.3 0/6725 0/375 0/2275 0/30 0/90
Potable water 6.5 – 9 2 250 250 12 –
NFR: non-filterable residue (suspended solids).
130 samples of dew and 84 samples of rain.
61 samples of dew and 45 samples of rain.
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312308
Fig. 10. (a) SO
concentration (mg/l) vs. wind direction (degree) (dew: o; rain: .): (1) electrical fuel plant
direction; (2) nocturnal dominant wind direction. (b) Cl
concentration (mg/l) vs. wind direction (degree) (dew:
o; rain: .).
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312 309
opposite to what is most currently observed (see Okochi et al., 1996). It is also interesting
to note that the time variation of rain pH (and dew to a lower extent) shows a minimum
during the winter season, when fossil energy is widely used.
ions were found significantly higher for only a very few events.
Also, dew samples present ion concentrations, on average, that are more important than
rain samples; the concentration ratios dew/rain lie between 1.13 (Cl
) and 3.3 (SO
Dew can reveal the presence of local pollutants in the atmosphere and thus can be used as
pollution indicator. We investigated the correlation of the pH with SO
Fig. 11. Correlation between the collected water volume and (a) SO
concentration in mg/l and (b) Cl
concentration in mg/l.
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312310
dew and rain did not show any obvious correlation. However, the chemistry of dew and
rain is complex (see Okochi et al., 1996) and its investigation is far beyond the scope of
the present paper.
It is also interesting to correlate the ion concentration with the wind direction. The wind
direction was calculated from data averaged during 6 h before the sampling. Fig. 10a,b
illustrate the correlation of SO
concentration with wind direction. Sulfate is
detected mostly when the wind comes from NE. Ten kilometers from the laboratory and in
the same direction (about 65j), an electrical plant is located in the Ajaccio Gulf and
represents the only known source of SO
due to fuel combustion. The presence of Cl
due to the vicinity of the sea and is correlated with the wind coming from this direction
The correlation of Cl
ions with the amount of dew collected shows that the
ion concentration is higher for smaller water volumes (Fig. 11). Therefore, the presence of
such ions in solution is related to the double constraint of a precise wind direction and low
dew yield, which explains why the events are so rare.
The results were compared (Table 1) to the maximum ion concentration permitted in
France for potable water. Except for a very few events, dew water appeared potable, at
least with respect to the above ions. The pH, although weakly acidic, shows characteristics
comparable to many brands of bottled mineral spring water. However, it has to be noted
that the concentration of nonfilterable residues (suspended solids) is more important. This
is a point that should be examined in a further study.
Well-designed radiative dew condensers, such as the condenser presented here, could
become a useful source of water in places where fog or rains are lacking. With a 45%
yearly occurrence of dew in Ajaccio, the condenser harvested about 770 l in 16 months of
functioning. A first chemical study of dew water permits one to conclude that such
harvested dew water is potable (concerning the ions investigated), without further
purification although the average pH is weakly acidic (pHc6) and the suspended solid
concentration is high. Although we did not perform any bacteriological analyses of such
dew water, it seems unlikely that dew can be contaminated by bacteria other than those
found in the ambient air, which are very few and harmless. This is, however, a point that
deserves further investigation—in addition to checking for other chemicals and suspended
In November 11, 2001, during a storm with wind velocity larger than 150 km/h, a crack
formed in the foil. The foil appeared to be aged, and it was removed.
We can evaluate the cost of such a water collection device. Assuming a 50-year lifetime
for the structure and 16 months for the foil (structure cost: o1800; foil cost: o30; foil
sewing: o150), we find o0. 30/l. This cost can be decreased by a large factor by saving
money on the infrastructure and the foil sewing, whose cost is nearly entirely due to
manpower. Accounting for only the foil manufacturing, we find a cost of order o0.04/l.
We are now testing a cheaper condenser in our laboratory site and another one in an arid
environment (the Negev desert, Israel).
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312 311
We thank G. Poli and J. Fieschi for their help with the measurements and S. Berkowicz
for a critical reading of the manuscript.
Alnaser, W.E., Barakat, A., 2000. Use of condensed water vapour from the atmosphere for irrigation in Bahrain.
Appl. Energy 65, 3– 18.
Beysens, D., 1995. The formation of dew. Atmos. Res. 39, 215– 237.
Beysens, D., Milimouk, I., Nikolayev, V., 1998. Dew recovery: old dreams and actual results. Proceedings of the
First International Conference on Fog and Fog Collection, Vancouver (Canada), pp. 269 – 272.
Beysens, D., Milimouk, I., Nikolayev, V., Muselli, M., Marcillat, J., 2001. Using Radiation Cooling to Condense
Atmospheric Vapor: A Study to Improve the Water Yield. J. Hydrol., in press.
Chaptal, L., 1932. La lutte contre la se
´cheresse—la captation de la vapeur d’eau atmosphe
´rique. Nature (Paris) 60,
449 – 454.
Jukov, N.N., 1931. O drevnih gydrotehnicheskih soorujeniyah goroda Fodosii v sviazi s voprosom ob izuchenii
principov dobychi vlagi iz atmosfery. Feodosijskoie otdelenie obshchestva po izucheniu Kryma i arheologi-
cheskij muzej, Sbornik statej po ekonomike, bytu i istorii Feodosijskogo rajona, Vol. 1, Feodosia, (in Rus-
sian); translated in French by Milimouk I., Beysens D., 1995. Recueil d’articles sur la condensation de l’eau
´rique. Rapport CEA-Saclay, DIST 95002495, pp. 68– 74.
Jumikis, A.R., 1965. Aerial wells: secondary sources of water. Soil Sci. 100, 83 – 95.
Knapen, M.A., 1929. Dispositif inte
´rieur du puits ae
´rien Knapen. Extrait des me
´moires de la socie
ieurs civils de France (Bull. Jan. – Feb. 1929). Imprimerie Chaix, Paris.
Monteith, J.L., 1957. Dew. Q. J. R. Meteorol. Soc. 83, 322– 341.
Nikolayev, V.S., 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 model for assessing dew/frost surface deposition.
Proceedings of the 2nd International Conference on Fog and Fog Collection, St John’s (Canada), 16 –20 July,
2001. The developed interactive computer applications is available from http://www.opur.u-bordeaux.fr.
Nilsson, T., 1996. Initial experiments on dew collection in Sweden and Tanzania. Sol. Energy Mater. Sol. Cells
40, 23 – 32.
Nilsson, T., Vargas, W.E., Niklasson, G.A., Granqvist, C.G., 1994. Condensation of water by radiative cooling.
Applied Sol. Energy 5 (f), 310– 317.
Okochi, H., Kajimoto, T., Arai, Y., Igawa, M., 1996. Effect of acid deposition on urban dew chemistry in
Yokohama, Japan. Bull. Chem. Soc. Jpn. 69, 3355– 3365.
M. Muselli et al. / Atmospheric Research 64 (2002) 297–312312