Using eddy covariance and stable isotope mass balance

techniques to estimate fog water contributions to a Costa

Rican cloud forest during the dry season

__________________________________________________________________

S. Schmid, R. Burkard

University of Bern, Bern, Switzerland

K.F.A. Frumau, C. Tobón, L.A. Bruijnzeel

VU University, Amsterdam, The Netherlands

R. Siegwolf

Paul Scherrer Institute, Villingen, Switzerland;

W. Eugster*

University of Bern, Bern, Switzerland and ETH Zurich, Zurich, Switzerland

* Corresponding author:

Werner Eugster, ETH Zurich, Institute of Plant, Animal and Agroecosystem Sciences,

Universitätsstrasse 2, CH–8092 Zurich, Switzerland.

E-mail werner.eugster@agrl.ethz.ch. Fax: +41 44 632 1153.

1

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

1

Abstract

Fog deposition, precipitation, throughfall, and stemflow were measured in a

windward tropical montane cloud forest near Monteverde, Costa Rica, for a 65-day

period during the dry season of 2003. Net fog deposition was measured directly with

the eddy covariance (EC) method and amounted to 1.2 ± 0.1 mm day –1 (mean ±

standard error). Fog water deposition was 5–9% of incident rainfall for the entire

period, which is at the low end of previously reported values. Stable isotope

concentrations ($18O and $ 2H) were determined in a large number of samples of each

water component. Mass balance-based estimates of fog deposition were 1.0 ± 0.3 mm

d–1 and 5.0 ± 2.7 mm d–1 (mean ± SE) when $18O and $ 2H were used as tracer,

respectively. Comparisons between direct fog deposition measurements and the

results of the mass-balance model using $ 18O as a tracer indicated that the latter might

be a good tool to estimate fog deposition in the absence of direct measurement under

many (but not all) conditions. At 506 mm, measured water inputs over the 65 days

(fog plus rain) fell short by 46 mm compared to the canopy output of 552 mm

(throughfall, stemflow, and interception evaporation). The discrepancy is attributed to

underestimation of rainfall during conditions of high wind.

Keywords: Fog, Stable Isotope Tracer, Interception, Tropical Montane Cloud Forest,

Costa Rica

Introduction

Montane cloud forests are widely believed to receive significant extra amounts of

water through the capture of passing low cloud (fog). The quantification of this

additional water input is of great practical importance where upland watersheds with

cloud forests supply water to downstream populations (Zadroga, 1981; Brown et al.,

1996; Rhodes et al., 2010). Fog water inputs are usually quantified indirectly, be it

through comparison of rainfall and crown drip during times with and without fog (e.g.

Harr, 1982; Sigmon et al., 1989; Hafkenscheid et al., 2002; Holder, 2003), modeling

(e.g. Lovett, 1984; Yin and Arp, 1994; Walmsley et al., 1996; Ritter et al., 2008), or,

by simply using some kind of passive “fog” gauge (e.g. Juvik and Ekern, 1978;

Goodman, 1985; cf. Bruijnzeel et al., 2005). The drawback of indirect methods is that

the quantification of fog deposition is dependent on correct rainfall and net

2

24

252627282930313233343536373839404142434445

4647484950515253545556

2

precipitation measurements and, by necessity, represents net amounts because of

unmeasured losses of water evaporated from the wetted canopy. Reliance on fog

gauges introduces the problem of separating contributions by fog and (wind-driven or

inclined) rain to the overall catch (McJannet et al., 2007; cf. Frumau et al., 2010a,b;

Giambelluca et al., 2010; Tanaka et al., 2010). An additional problem concerns the

translation of water amounts collected by a vertical gauge surface to effective

hydrological input onto a horizontal plane, and the impossibility of any type of fog

gauge to represent natural vegetation. The distinction between fog and rain water

inputs is crucial, because the latter are – due to the larger droplet sizes involved –

deposited independently of the surface cover. Fog water, however, is strongly

influenced by turbulence. Because of the rougher and larger surface area of forest

canopies, fog water deposition is much larger over forests than over, for example,

pasture (e.g. Thalmann et al., 2002).

This paper presents the results of direct net fog water deposition

measurements by means of the eddy covariance method (Beswick et al., 1991;

Eugster et al., 2006; Beiderwieden et al., 2008) made above a windward montane

cloud forest in northern Costa Rica. The results are compared with those obtained

with a mass-balance model using the stable isotopes 18O and 2H as tracers (cf. Guswa

et al., 2007; Scholl et al., 2010). To further validate the direct fog deposition

measurements, the remaining components of the wet-canopy water budget equation

(Holwerda et al., 2006a) were measured or calculated and the different approaches

evaluated.

Materials and methods

Site description

Hydrological and micrometeorological measurements were made between 10

February and 13 May 2003 at 1,460 m.a.s.l. in the San Gerardo headwater area within

the Caño Negro drainage basin on the exposed Atlantic slopes of the Cordillera de

Tilarán in northern Costra Rica, ~10 km NE of the town of Santa Elena (10º 21' 33''

N, 84º 48' 5'' W). The meteorological measurements were made on top of a 24 m tall

scaffolding tower situated about 200 m from the crest of a fully forested 10° slope of

N350E exposure. Measurements of net precipitation were made on a nearby slope

with a somewhat different exposure and gradient (slope 30°, N270E). The forest was

3

5758596061626364656667686970717273747576777879

80

81

8283848586878889

3

a windward lower montane cloud forest having its main canopy surface at 21–22 m,

with epiphyte-laden emergent trees extending above this level by several metres.

Overall epiphyte biomass was determined at 16.2 t ha-1 (Köhler et al., 2007). Tree

ferns and palms were common in the understory. Long-term rainfall data for the upper

Atlantic slopes are not available, but conventionally measured rainfall input (rain

only) between 1 July 2003 and 30 June 2004 at the site was c. 6,000 mm (K.F.A.

Frumau, unpublished data). A somewhat drier period tends to occur

in the area between February and April, with monthly

rainfalls of generally less than 150 mm, vs. 500 mm or

more during the remainder of the year (Clark et al.,

2000). Due to the prevailing high wind speeds there is a major horizontal

precipitation component in the form of wind-driven rain (mostly drizzle) and fog ( cf.

Clark et al., 1998; Frumau et al., 2010a,b; Häger and Dohrenbusch, 2010). Tobón et

al. (2010) provide further basic information on site climate.

Methods and instrumentation

Net fog water deposition was measured directly by the eddy covariance technique.

Net deposition is defined as the difference between gross deposition – the water that is

actually deposited on the leaves or on other surfaces – and formation of fog droplets

due to condensation. The latter constitutes a counteracting, upward flux which

reduces the measured flux at the height of the eddy covariance set-up (see Eugster et

al. (2006) for details). Eddy covariance flux measurements are directly measuring

deposition rates, and thus the only assumptions that need to be made are that: (i) the

measurements are representative for a larger surface area (homogeneity of upwind

canopy surface and topography), and (ii) turbulent conditions are stationary during the

measurements (e.g. no abrupt changes in wind direction and wind speed). The

measurements were performed with a three-dimensional ultrasonic anemometer

(model 1199 HSE, Gill Instruments Ltd., Lymington , UK) coupled with an active

high-speed FM-100 cloud particle spectrometer (Droplet Measurement Technologies

Inc., Boulder, CO, USA). Fog droplets were continuously measured in 40 size classes

between 2 and 50 µm diameter and recorded together with the 3-dimensional wind

speed information at a frequency of 12.5 times per second. Internal measurements

were carried out 100 times per second, and therefore each of the recorded outputs

4

90919293949596979899100101102103104

105

106107108109110111112113114115116117118119120121122

4

represented the average of eight measurements. For further details on the eddy

covariance measurements the reader is referred to Burkard et al. (2003) and Eugster

et al. (2006). Vertical rainfall was measured by two standard rain gauges (100 cm 2

orifice) that were manually emptied on a daily basis, and one automatic tipping-

bucket rain gauge (200 cm2 orifice, 0.2 mm per tip). The precipitation measurements

were corrected for aerodynamic losses around the gauge due to wind following the

procedure of Førland et al. (1996), and for the effects of sloping ground according to

Sharon (1980). Horizontal precipitation (i.e. fog and wind-driven rain) was measured

using a modification of the cylindrical louvered gauge of Juvik and Ekern (1978)

(Frumau et al., 2010a). Throughfall (TF) was measured with the roving gauge

technique using 60 totalizers with an orifice of 100 cm 2 each (cf. Lloyd and Marques,

1988; Holwerda et al., 2006b). Stemflow (SF) was measured on 30 trees using spiral

gutters and made up < 1% of uncorrected rainfall (K.F.A. Frumau and C. Tobón,

unpublished data). Wet-canopy evaporation was calculated with the wet-canopy

version (i.e. surface resistance rs = 0) of the Penman-Monteith equation (Monteith,

1965).

<Table 1>

In addition to the eddy covariance measurements, fog water deposition

was also estimated with a mixing model as described by Brunel et al. (1995), using

the stable isotopes 18O and 2H as tracers. Samples for the determination of isotope

concentrations were taken on a nominally daily basis from rain, fog, TF and SF. Exact

durations for TF sampling are given in Table 1, whereas more detailed information

can be found in the raw data set (http://doi.pangaea.de/10.1594/PANGAEA. 735614).

Fog water was collected using a modified Caltech Active Strand Cloudwater

Collector (CASCC; Demoz et al., 1996) with a shielded inlet to avoid raindrops (>

100 "m) from being collected (Figure 1). The teflon strands collect all fog droplets

larger than approximately 5–7 "m. CASCC sampling was executed automatically

(including transfer to a closed flask to prevent re-evaporation; cf. Scholl et al., 2010)

whenever visibility was < 500 m. At the end of each sampling period a representative

subsample was taken from the collected fog water for analysis. Rain water was

collected with a sampler built according to the specifications of IAEA (2002) whereas

a locally constructed, rotating gauge with a vertical orifice (RGVO) was used to

sample wind-driven rain (horizontal precipitation) (Figure 2). A representative

subsample of the water collected by each individual TF and SF collector was used for

5

123124125126127128129130131132133134135136137138139

140141142143144145146147148149150151152153154155156

5

isotope analysis. For days on which samples were available from all water types

collected over the same time interval, the proportions of fog and rain water in TF were

calculated using the following mixing model containing two end members:

f = (!TF - !P) / (!F - !P), (1)

where f is the fraction of rain water in TF and the various subscripts of $

denote the respective isotopic ratios of either $18O or $2H in TF, fog (F) and rain (P)

with reference to the Vienna Standard Mean Ocean Water (V-SMOV). Thus, !TF is a

placeholder for the $18OTF or $2HTF tracer. Correspondingly !F and !P represent $18OF

and $2HF or $18OP and $2HP , respectively, in Eq. (1). Cases where !TF was outside the

range given by $F and $P were rejected.

<Figure 1>

<Figure 2>

Results and discussion

Isotopic signatures of fog, rain, throughfall, and stemflow water

In total, 566 water samples were collected during the 65-day field period. Figure 3

shows the isotopic signatures of RGVO, rain, fog, and throughfall water together with

the global meteoric water line (GMWL), the local meteoric water line as determined

by Rhodes et al. (2006), and a local meteoric water line based on the rain water

samples collected by the present study. Since the scatter was large (as may be

expected from short-term sampling), only rain water samples with $18O < 0 ‰ and

$2H < –20 ‰ were used, yielding the following relation: $ 2H = (7.59 ± 0.23) $18O +

(3.90 ± 1.49) ‰ V-SMOW (r2 = 0.991, n=11). The presently derived mixing water

line for dry season conditions is flatter than the one reported by Rhodes et al. (2006)

for the two full years between June 2003 and April 2005..The difference in slope

between the two local meteoric water lines is consistent with the expectation that

winter (colder temperatures) precipitation should lead to lower slopes. In addition,

enhanced evaporation during dry season conditions also tends to produce a lower

slope (cf. Clark and Fritz, 1997).

6

157158159160

161

162

163164165166167168169

170

171

172

173

174175176177178179180181182183184185186187

6

<Figure 3>

Fog deposition

Average net fog deposition measured by the eddy covariance system for conditions

with visibility below 1,000 m was 0.05 mm h-1 or 14.2 mg · m-2 s-1. This rate is similar

to the fog deposition measured with the same equipment in an elfin cloud forest

setting near Pico del Este in Puerto Rico (0.04 mm h -1 or 10.2 mg · m-2 s-1; Eugster et

al., 2006; Holwerda et al., 2006a) and lies within the range of reported values

measured with this technique outside the tropics (Beswick et al., 1991; Vong and

Kowalski, 1995; Vermeulen et al., 1997; Burkard et al., 2003). The average daily

deposition rate of 1.2 ± 0.1 mm also lies within the range of 0.2–4.0 mm d-1 of

reported cloud water interception rates in tropical montane areas as obtained by

various indirect methods (Bruijnzeel and Proctor, 1995; Bruijnzeel, 2005). Fog

deposition expressed as a percentage of rainfall (5%) was at the low end of the

spectrum reported by Bruijnzeel and Proctor (1995) and Bruijnzeel (2005) (4 –281%

of corresponding rainfall). Whilst the observed daily deposition rates are plausible,

therefore, the contribution by fog water to total water input proved rather small. Based

on the direct measurements, fog water added a mere 19 mm of water to the forest

water budget during the 65-day measurement period during which a possibly

unusually low fog frequency of 26% (i.e. times with horizontal visibility of 1000 m or

less) was observed. The higher than normal position of the cloud base, and the

associated lack of fog, were caused by the occurrence of the “ temporales del pacifico”

weather pattern (Clark et al., 2000). During several days, the field site was

climatically speaking situated on the leeward side of the Continental Divide due to the

occurrence of the associated warm and dry western winds. According to Clark et al.

(2000), this type of weather system tends to occur mostly during the hurricane season

(August–October) and its occurrence in March 2003 may have led to non-

representative conditions. However, Clark et al. (2000) also report that during 20–

25% of all hours during the dry season the upper slopes and ridges along the

Continental Divide in the (leeward) area of nearby Monteverde are immersed in

clouds and are thus affected by fog. These percentages compare well with the

visibility-based observation of 26% fog occurrence during the present field period.

Therefore, the inferred 5% fog contribution during the study period may be typical for

7

188

189

190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219

7

the dry season after all, although longer-term observations would be needed to

confirm this (cf. Guswa et al., 2007; Lawton et al., 2010).

Isotope concentrations of all water types were available for 31 sampling

days (Table I). The fractions of fog water in throughfall ( TF) as calculated with the

mixing model of Eq. (1) were outside the acceptable range of 0–100% on 14 days

(45%) and 21 days (68%) for $18O and $2H, respectively. Most likely, the time series

of available samples for the individual components indicated effects of dynamic

changes in weather conditions. For example, Eugster (2007) reported that at a similar

site in Puerto Rico, $18O in fog water tended to be related more closely to values

found in precipitation falling 36–60 hours prior to the fog event. This may explain

why fog water sometimes exhibits a more negative isotopic signature than

precipitation. Whilst this has no consequences for the application of the type of tracer

mixing model used here, such findings may help to understand why in only 55% of all

cases TF had a $18O signature intermediate between the values of the two end

members (i.e. fog and precipitation). At 32%, this fraction was even lower for $ 2H (cf.

Table 1).

Therefore, $18O was considered to be the more reliable tracer for

determining the fraction of fog water in TF samples. The fractions obtained for the

remaining days using $18O as a tracer are shown in Figure 4. The calculated fractions

of fog water in stemflow (SF) were all outside the acceptable range. Therefore, SF

was assumed to receive the same share of fog water as determined for corresponding

TF. Since SF is a very small component compared to TF, this assumption does not

introduce a large additional uncertainty in estimated amounts of daily fog deposition.

<Figure 4>

The eddy covariance-based fog water fluxes (EC) and the deposition rates

estimated with the mass balance approach for $ 18O were in the same order of

magnitude as indicated by the 1:1 line, except for two events with unrealistically high

deposition rates in view of the prevailing wind speeds and fog liquid water content

(see inset in Figure 5). Because the correlations were higher when $ 18O was used as a

tracer (r = 0.72, p = 0.009) instead of $2H (r = 0.54, p = 0.045), the numbers given

below pertain to the results obtained with $ 18O. Overall, fog deposition as calculated

with the mixing model may be considered to represent the gross fog water flux at the

canopy level whereas the EC-based flux represents the net fog water flux. On average,

8

220221222223224225226227228229230231232233234235236

237238239240241242243244245246247248249250251252253

8

the gross fog water flux was 1.8 times the directly measured amount (excluding the

two events that gave unrealistic results; Figure 5, inset). The corresponding values

were 0.37 ± 0.06 mm d–1 for the EC (net flux) and 0.66 ± 0.13 mm d–1 for the mixing

model (gross flux; n=14 events) or 5% and 9% of incident rainfall, respectively.

Eugster et al. (2006) reported the gross fog water flux at the canopy level for an elfin

cloud forest site in Puerto Rico to be 1.74 times the net flux measured by eddy

covariance, a value that is consistent with the presently derived factor. In view of this

similarity in the difference between gross and net fog water fluxes, a factor of 1.8 may

also be representative for other tropical montane cloud forest sites and illustrates the

importance of fog droplet formation as the air moves upslope. As stated earlier, this

condensation counteracts the deposition of fog droplets onto the vegetation surface.

<Figure 5>

Wet-canopy water balance

For the computation of the wet-canopy water balance, a period of 65 days with

uninterrupted measurements of all components was selected (9 March–13 May 2003).

For this period, the following amounts (all in mm) were determined:

487(corrected rainfall) + 19(direct fog) # 497(TF) + 50(wet-canopy evaporation) + 5(SF) (2)

For these 65 days, the average daily difference between canopy inputs (rain and fog)

and outputs (TF, SF, evaporation) was not statistically significant (paired Wilcoxon

rank sum test, 95% confidence level, p = 0.30). However, during a two-day storm

event with high wind speeds and inclined rainfall, the difference between inputs and

outputs became quite large (see Figure 6: days 91 and 92, or 1 and 2 April 2003).

If it is assumed that the uncertainty in the TF measurements (which made up 97% of

the total output during these two days) was small due to the use of a large number of

collectors (cf. Lloyd and Marques, 1988; Holwerda et al., 2006b), then it is most

likely that the “missing” amounts during these two days reflect an underestimation of

the inputs rather than an overestimation of the output components.

< Figure 6>

9

254255256257258259260261262263264265266

267

268

269270271272273

274

275276277278279280281282283284285

9

If the unaccounted water would have been due to underestimated fog

deposition alone, then a daily deposition rate of more than 18 mm would have been

required, which would seem unrealistically high in view of the published range

reported by Bruijnzeel (2005). In addition, the mixing model-based fraction of fog

water in TF was very small for large TF events (Figure 4), which is indicative of the

fact that fog inputs were only significant at times when precipitation rates were

comparatively low. For example, a daily TF total of 10 mm would yield a contribution

of ~5% by fog according to the mixing model (Eq. (1)). Measured daily TF totals for

1 and 2 April 2003 were very much larger (165 and 134 mm, respectively, with

corresponding measured rainfall amounts of 134 and 44 mm, respectively). Whilst no

isotopic information was available for these extreme events , based on Figure 4 it

would be expected that the associated fraction of fog water in TF for these two events

would be well below 5%. This suggests that total fog deposition must have been less

than 15 mm during these two days (note that the eddy covariance system measured a

total amount of 5.5 mm). Therefore, it is concluded that the bulk of the unexplained

gap of 46 mm in the overall wet-canopy water balance (Eq. (2)) was due to

underestimation of rainfall, despite the application of corrections for aerodynamic

losses around the gauge and for slope effects. This conclusion is supported by

measurements made with the louvered cylindrical gauge (MJU) and the rotating

vertical orifice gauge (RVOG), both of which collected large volumes of water during

these two days. The MJU recorded 459 mm and 100 mm of horizontal precipitation

for 1 and 2 April, respectively. Furthermore, isotopic values of the water collected

with the RVOG showed a clear contrast to those of fog water (as collected by the

modified CASCC) throughout the measuring campaign (Figure 7), whereas no

significant difference was found with the $-values in rain water, particularly for $ 2H

(Figure 7b). It is most likely, therefore, that the water collected by the rotating gauge

during the storm events of 1 and 2 April was mostly wind-driven rain.

<Figure 7>

Conclusions

The components of the wet-canopy water balance of a windward montane cloud forest

in north-western Costa Rica were measured or calculated for a period of 65 days

during the dry season of 2003. There was no statistically significant difference

10

286287288289290291292293294295296297298299300301302303304305306307308309310311312313

314

315

316317318

10

between totals of canopy inputs (fog and rain) and outputs (throughfall, stemflow and

wet-canopy evaporation) over this period if expressed in daily terms. Fog inputs

ranged between 5% (eddy covariance-based net flux measurements) and 9% (tracer-

based mixing model using $18O) of incident rainfall. Rainfall was corrected for the

effects of sloping ground and aerodynamic losses around the gauge. On average, the

corrected amounts were 27% higher. However, for an extreme composite storm event

with strong winds and therefore substantial inclined or near-horizontal rainfall, the

applied corrections were insufficient and rainfall was still underestimated. During the

same event, throughfall greatly exceeded incoming rainfall and a cylindrical louvered

screen gauge collected large amounts of water, confirming the occurrence of near-

horizontal precipitation.

The mixing model for $18O was found to be a good tool to assess the

fraction of fog water in throughfall during 16 out of 31 precipitation events.

However, the method needs further refinement as several relevant questions remain

unsolved, including: (i) Why did the method not work well for stemflow samples? (ii)

Why does $18O appear to be a useful tracer for this purpose whereas $ 2H yielded

completely different and arguably less reliable results? (iii) Does isotopic

fractionation occur when fog water is sampled with an Active Strand Cloud Water

Collector? And (iv) how large is the effect of evaporation on isotopic concentrations

in throughfall water?

Longer-term measurements of both directly measured net fog deposition

and isotopic composition of fog, rain and throughfall water are needed to unravel the

existence of relationships between isotopic signatures in fog and rain, and the time

lags between the two (Eugster, 2007). This is expected to increase confidence in the

idea that the isotope mixing approach could usefully replace the expensive and

sophisticated eddy covariance instrumentation. However, it would first be necessary

to understand why the results obtained with 2H as a tracer differed so much from those

based on 18O (see also Scholl et al., 2010) at this wind-exposed location, when Rhodes

et al. (2006) found isotopic values in the precipitation at nearby (leeward)

Monteverde to follow the global meteoric water line.

Acknowledgements

11

319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349

350

11

This study was supported by the Swiss National Science Foundation (grant no. 2100-

068051) and the Forestry Research Programme of the UK Department for

International Development (project R7991). The views expressed here are not

necessarily those of DFID. We thank Professor Jeff Collett (Colorado State

University) for the loan of two CASCC collectors, and Jürg Schenk (University of

Bern) for technical assistance with the eddy covariance set-up.

References

Beiderwieden E, Wolff V, Hsia YJ, and Klemm O. 2008. It goes both ways: Measurements of simultaneous evapotranspiration and fog droplet deposition at a montane cloud forest. Hydrological Processes 22: 4181–4189.

Beswick KM, Hargreaves KJ, Gallagher MW, Choularton T, Fowler D. 1991. Size-resolved measurements of cloud droplet deposition velocity to a forest canopy using an eddy correlation technique. Quarterly Journal of the Royal Meteorological Society 117: 623–645.

Brown MB, de la Roca I, Vallejo A, Ford G, Casey J, Aguilar B, Haacker R. 1996. A valuation analysis of the role of cloud forests in watershed protection. Sierra de las Minas Biosphere Reserve, Guatemala and Cusuco N.P., Honduras . Philadelphia, USA: RARE Center for Tropical Conservation.

Bruijnzeel LA. 2005. Tropical montane cloud forests: A unique hydrological case. In Forests, water and people in the humid tropics , Bonell M, Bruijnzeel LA. (eds). Cambridge University Press: Cambridge, UK; 462-483.

Bruijnzeel LA, Proctor J. 1995. Hydrology and biogeochemistry of tropical montane cloud forests: what do we really know? In Tropical Montane Cloud Forests , Hamilton LS, Juvik JO, Scatena FN. (eds). Ecological Studies 110. Springer Verlag: New York; 38–78.

Bruijnzeel LA, Eugster W, Burkard R. 2005. Fog as a hydrologic input. In Encyclopaedia of hydrological sciences , Anderson MG, McDonnell JJ. (eds). John Wiley: Chichester, UK; 559–582.

Brunel, J.-P., G.K. Walker, and A.K. Kenneth-Smith. 1995. Field validation of isotopic procedures for determining sources of water used by plants in a semi-arid environment. Journal of Hydrology 167: 351–368.

Burkard R, Bützberger P, Eugster W. 2003. Vertical fogwater flux measurement above an elevated forest canopy at the Lägeren research site, Switzerland. Atmospheric Environment 37: 2979–2990.

Clark ID, Fritz P. 1997. Environmental Isotopes in Hydrogeology . Lewis Publishers, Boca Raton, 328 pp.

12

351352353354355356357

358

359360361

362363364365

366367368369

370371372

373374375376

377378379

380381382

383384385

386387

12

Clark KL, Nadkarni NM, Schaeffer D, and Gholz HL. 1998. Atmospheric deposition and net retention of ions by the canopy in a tropical montane forest, Monteverde, Costa Rica. Journal of Tropical Ecology 14: 27-45.

Clark KL, Lawton RO, Butler PR. 2000. The physical environment. In Monteverde. Ecology and Conservation of a Tropical Cloud Forest , Nadkarni NM, Wheelwright NT. (eds). Oxford University Press: Oxford, UK; 15–34.

Demoz BB, Collett JL, Daube BC. 1996. On the Caltech Active Strand Cloudwater Collectors. Atmospheric Research 41: 47–62.

Eugster W, Burkard R, Holwerda F, Scatena FN, Bruijnzeel LA . 2006. Characteristics of fog and fog-water fluxes in a Puerto Rican elfin cloud forest. Agricultural and Forest Meteorology 139: 288-306.

Eugster W. 2007. The relevance of fog for the vegetation: is it the water or the nutrients that matter? In Proceedings of the Fourth International Conference on Fog, Fog Collection and Dew, Biggs A, Cereceda P. (eds). La Serena, Chile, 22--27 July 2007, pp. 359–362.

Førland EJ, Allerup P, Dahlström B, Elomaa E, Jonsson T, Madsen H, Perälä J, Rissanen P, Vedin H, Vejen F. 1996. Manual for operational correction of Nordic precipitation data. Norwegian Meteorological Institute: Oslo, Norway.

Frumau KFA, Bruijnzeel LA, Tobón C. 2010a. A practical approach to the measurement of inclined rainfall in windy montane terrain. Hydrological Processes, this issue.

Frumau KFA, Burkard R, Schmid S, Bruijnzeel LA, Tobón C, Calvo-Alvarado JC. 2010b. Fog gage performance under conditions of fog and wind-driven rain. Hydrological Processes , this issue.

Giambelluca TW, DeLay JK, Nullet MA, Scholl MA. 2010. Interpreting canopy water balance of fog screen observations: Separating cloud water from wind-blown rainfall at two contrasting forest sites in Hawai'i. In Tropical Montane Cloud Forests: Science for Conservation and Management, Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge University Press: Cambridge, UK; xxx-yyy.

Goodman J. 1985. The collection of fog drip. Water Resources Research 21: 392-394.

Guswa AJ, Rhodes AL, and Newell SE. 2007. Importance of orographic precipitation to the water resources of Monteverde, Costa Rica. Advances in Water Resources 30: 2098–2112.

Hafkenscheid RLLJ, Bruijnzeel LA, de Jeu RAM, Bink NJ. 2002. Water budgets of two upper montane rain forests of contrasting stature in the Blue Mountains, Jamaica. In Proceedings of the Second International Colloquium on Hydrology and Water Management in the Humid Tropics . Technical Documents in Hydrology no. 52, Gladwell JS (ed). IHP-UNESCO: Paris and CATHALAC: Panama City; 399–424.

13

388389390

391392393

394395

396397398

399400401402

403404405

406407408

409410411

412413414415416417

418

419420421

422423424425426427

13

Häger A, Dohrenbusch A. 2010. Structure and dynamics of tropical montane cloud forests under contrasting biophysical conditions in north-western Costa Rica. Hydrological Processes , this issue.

Harr RD. 1982. Fog drip in the Bull Run municipal watershed, Oregon. Water Resources Bulletin 18: 785–789.

Holder CD. 2003. Fog precipitation in the Sierra de las Minas Biosphere Reserve, Guatemala. Hydrological Processes 17: 2001-2010.

Holwerda F, Burkard R, Eugster W, Scatena FN, Meesters AGCA, Bruijnzeel LA. 2006a. Estimating fog deposition at a Puerto Rican elfin cloud forest site: comparison of the water-budget and eddy covariance methods. Hydrological Processes 20: 2669–2692.

Holwerda F, Scatena FN, and Bruijnzeel LA. 2006b. Throughfall in a Puerto Rican lower montane rain forest: a comparison of sampling strategies. Journal of Hydrology 327: 592–602.

IAEA. 2002. Special issue on the global network of isotopes in precipitation. Water & Environment Newsletter 16: 5.

Juvik JO, Ekern PC. 1978. A climatology of mountain fog on Mauna Loa, Hawaii Island. Technical report no. 118. Hawaii Water Resources Research Center: Honolulu, U.S.A.

Köhler L, C. Tobón C, Frumau KFA, Bruijnzeel LA. 2007. Biomass and water storage of epiphytes in old-growth and secondary montane rain forest stands in Costa Rica. Plant Ecology 193: 171–184.

Lawton RO, Nair US, others. 2010. Quantitative measures of immersion in cloud and the biogeography of cloud forests. In Tropical Montane Cloud Forests: Science for Conservation and Management, Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge University Press: Cambridge, UK; xxx-yyy.

Lloyd CR, Marques AO. 1988. Spatial variability of throughfall and stemflow measurements in Amazonian rain forest. Agricultural and Forest Meteorology 43: 277–294.

Lovett GM. 1984. Rates and mechanisms of cloud water deposition to a subalpine balsam fir forest. Atmospheric Environment 18: 361–371.

McJannet D, Wallace JS, Reddell P. 2007. Precipitation interception in Australian tropical rainforests: II. Altitudinal gradients of cloud interception, stemflow, throughfall and interception. Hydrological Processes 21: 1703–1718.

Monteith JL. 1965. Evaporation and the environment. Symposium of the Society for Experimental Biology 19: 205–234.

14

428429430

431

432433

434435

436437438439

440441442

443444

445446447

448449450

451452453454

455456457

458459

460461462

463464

14

Rhodes AL, Guswa AJ, Newell SE. 2006. Seasonal variation in the stable isotopic composition of precipitation in the tropical montane forests of Monteverde, Costa Rica. Water Resources Research 42: W11402. DOI:10.1029/2005WR004535.

Rhodes AL, Guswa AJ, Newell SE. 2010. Using stable isotopes to identify orographic precipitation events at Monteverde, Costa Rica. In Tropical Montane Cloud Forests: Science for Conservation and Management, Bruijnzeel LA, Scatena FN, Hamilton LS. (eds). Cambridge University Press: Cambridge, UK; xxx-yyy.

Ritter A, Regalado CM, Aschan G. 2008. Fog water collection in a subtropical elfin laurel forest of the Garajonay National Park (Canary Islands): A combined approach using artificial fog catchers and a physically-based impaction model. Journal of Hydrometeorology 9: 920–935.

Scholl M, Eugster W, Burkard R. 2010. Understanding the role of fog in forest hydrology: stable isotopes as tools for determining input and partitioning of cloud water in montane forests. Hydrological Processes , this issue.

Sharon D. 1980. The distribution of hydrologically effective rainfall incident on sloping ground. Journal of Hydrology 46: 165–188.

Sigmon JT, Gilliam FG, Partin ME. 1989. Precipitation and throughfall chemistry for a montane hardwood ecosystem: Potential contributions from cloud water. Canadian Journal of Forest Research 19: 1240–1247.

Slinn WGN, 1982. Predictions for particle deposition to vegetative canopies. Atmospheric Environment 16: 1785–1794.

Tanaka M, Kuraji K, Tantasirin C, Takizawa H, Tagtham N, Suzuki M. 2010. Relationships between rainfall, fog and throughfall at a hill evergreen forest site in northern Thailand. Hydrological Processes , this issue.

Thalmann E, Burkard R, Wrzesinsky T, Eugster W, Klemm O. 2002. Ion fluxes from fog and rain to an agricultural and a forest ecosystem in Europe. Atmospheric Research 64: 147–158.

Tobón C, Köhler L, Frumau KFA, Bruijnzeel LA, Burkard R, Schmid S. 2010. Water dynamics of epiphytic vegetation in a lower montane cloud forest: fog interception, storage and evaporation. In Tropical Montane Cloud Forests: Science for Conservation and Management, Bruijnzeel LA, Scatena FN, Hamilton LS (eds). Cambridge University Press: Cambridge, UK; in press.

Vermeulen AT, Wyers GP, Römer FG, Leeuwen NFMV, Draaijers GP, Erisman JW. 1997. Fog deposition on a coniferous forest in the Netherlands. Atmospheric Environment 31: 375–386.

Vong RJ, Kowalski AS. 1995. Eddy correlation measurements of size dependent cloud droplet turbulent fluxes to complex terrain. Tellus 47B: 331–352.

15

465466467468

469470471472473

474475476477

478479480

481482

483484485

486487

488489490

491492493

494495496497498

499500501

502503

15

Walmsley JL, Schemenauer RS, Bridgman HA. 1996. A method for estimating the hydrologic input from fog in mountainous terrain. Journal of Applied Meteorology 35: 2237–2249.

Yin XW, Arp PA. 1994. Fog contributions to the water budget of forested watersheds in the Canadian Maritime Provinces: A generalized algorithm for low elevations. Atmosphere-Ocean 32: 553–566.

Zadroga F, 1981. The hydrological importance of a montane cloud forest area of Costa Rica. In Tropical Agricultural Hydrology, Lal R, Russell EW (eds). John Wiley: New York; 59–73.

16

504505506

507508509

510511512513

16

Table Caption

Table I: Events with concurrent fog, precipitation, and measurable throughfall ( TF) to

allow for application of the isotope tracer model according to Eq. (1). Estimates of fog

water contributions in sampled TF given as percentages and absolute amounts

(mm d–1) for both $18O and $2H, although only the values obtained with $18O as the

tracer appear to provide realistic results. Eddy covariance (EC) fog water fluxes

relate to the 24 hour period prior to the time when TF samples were collected.

Figure Captions

Figure 1: Schematic representation of the CASCC fog collector.

Figure 2: Locally-built rotating gauge with vertical orifice (RGVO) to capture

horizontal precipitation.

Figure 3: Isotopic ratios in water collected with the rotating gauge with vertical

orifice (RGVO), and in rain, fog and throughfall water. The global mixing water line

(GMWL) of $2H = 12 + 8.5 $18O, and two local mixing water lines are added for

comparison (see text for details).

Figure 4: Measured daily throughfall amounts and corresponding fractions of fog

water in throughfall as calculated with the mixing model of Eq. (1) using $ 18O as the

tracer at the San Gerardo cloud forest site.

Figure 5: Directly measured (eddy covariance-based) daily net fog water deposition

rates vs. rates calculated with the mixing model of Eq. (1) using 18O as the tracer (r2 =

0.52, p = 0.009) at the San Gerardo cloud forest site.

Figure 6: The daily wet-canopy water balance at the San Gerardo forest site between

19 March and 13 May 2003. Inputs are displayed by positive values and dark bars,

outputs are negative with grey bars, whereas the deviation from a closed budget is

indicated by the bold line.

17

515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548

17

Figure 7: Box plots of isotopic signatures of source waters: (a) oxygen isotopic ratios

($18O), (b) hydrogen isotopic ratios ($ 2H). Each box shows the inter-quartile range

with its median (horizontal line). Whiskers show the data range up to 1.5 times the

interquartile range. Values outside this distribution are denoted by open circles.

Identical letters denote similarity (Wilcoxon rank sum test, p ! 0.05), different letters

indicate significant differences (p < 0.05). RGVO denotes rotating gauge with vertical

orifice.

18

549550551552553554555556

18

Schmid et al. Table 1

TF Duration Fogwater Throughfall Precipitation Tracer Model Fogwater in TF EC FluxEvent Date (mL) (hours) !18O !18O !18O !18O !18O (%) (mm/day) !2H (%) (mm/day) (mm/day)

1 2003-02-27 188 25.5 0.81 9.55 -1.22 -8.67 -0.70 2.66 no no – – – – 0.2142 2003-03-01 159 23.5 -1.16 7.41 -0.78 3.51 1.09 7.57 yes no 83.1 0.2 – – 0.1543 2003-03-02 1176 24.0 -0.92 9.32 -0.37 8.30 0.51 5.09 yes yes 61.5 1.2 75.9 1.5 0.6864 2003-03-03 122 25.5 -0.27 12.08 -1.38 -1.50 -0.38 3.33 no no – – – – -0.0615 2003-03-05 1050 26.0 0.44 12.23 0.47 6.78 1.04 7.61 yes no 95.0 1.5 – – 0.6236 2003-03-06 1432 22.5 -0.33 7.63 -0.38 10.77 0.27 9.24 no no – – – – 0.6377 2003-03-07 154 24.0 -0.56 8.12 -0.84 3.97 -1.81 -2.42 yes yes 77.6 0.2 60.6 0.2 0.3888 2003-03-08 665 25.0 -0.64 6.67 -0.97 7.01 1.21 4.12 no no – – – – 0.9839 2003-03-09 1250 20.5 0.70 5.85 -1.22 6.84 – 4.79 no no – – – – 1.477

10 2003-03-10 502 25.0 -1.15 8.52 0.55 5.09 5.64 6.55 yes no 75.0 0.6 – – 0.59911 2003-03-13 489 23.5 0.71 8.28 -0.14 6.26 -0.82 7.04 yes no 44.4 0.4 – – 0.10112 2003-03-14 406 15.5 -1.22 2.76 -1.26 1.27 -3.41 3.25 yes no 98.2 1.0 – – 0.47313 2003-03-27 9505 25.5 -1.06 0.50 -1.55 1.32 -1.73 -2.79 yes no 26.9 4.0 – – 0.00214 2003-04-03 12415 25.6 -1.74 0.80 -2.05 6.74 -1.07 0.48 no no – – – – 0.91915 2003-04-08 254 25.7 -1.24 7.45 -0.61 -0.09 -0.48 0.46 yes no 17.1 0.1 – – 0.08916 2003-04-11 687 24.2 -0.17 6.54 -0.88 6.17 -6.70 3.04 yes yes 89.1 1.0 89.4 1.0 0.28317 2003-04-12 – 24.1 -0.88 4.12 -1.97 3.73 -2.46 -1.35 yes yes 31.0 – 92.9 – 0.83918 2003-04-15 562 25.7 -1.59 3.07 -1.61 3.41 -1.38 -1.21 no no – – – – 0.65619 2003-04-24 1135 24.2 -7.02 0.27 -1.39 -4.17 -2.01 -3.47 no no – – – – 0.00020 2003-04-29 235 22.2 -2.25 -19.00 -2.87 -13.28 -3.39 -21.25 yes no 45.6 0.2 – – 0.00921 2003-05-01 12505 24.8 -7.97 -61.94 -9.57 -65.18 -9.18 -67.27 no yes – 39.2 7.9 0.02322 2003-05-06 4783 24.7 -0.96 -6.98 -2.36 -15.56 -2.55 -16.75 yes yes 11.9 0.9 12.2 0.9 0.73023 2003-05-07 3335 21.7 -0.78 -4.28 -1.73 -2.55 -1.53 -7.03 no no – – – – 0.51824 2003-05-08 601 26.0 -2.84 -16.57 -2.08 -8.47 -1.83 -9.44 yes no 24.8 0.2 – – 0.24825 2003-05-09 5714 24.0 -4.25 -24.57 -5.53 -35.69 -5.60 -38.44 yes yes 5.2 0.5 19.8 1.9 0.34826 2003-05-10 18852 24.0 -5.50 -48.73 -8.91 -54.55 -7.53 -52.86 no no – – – – 0.88827 2003-05-11 1382 24.0 -2.75 -18.72 -5.30 -21.44 -3.74 -21.99 no yes – – 16.8 0.4 0.57928 2003-05-12 12555 24.3 -8.82 -69.12 -6.78 -35.24 -6.35 -44.67 yes no 17.4 3.6 – – 0.61729 2003-05-13 11366 26.4 -10.52 -76.69 -12.46 -83.25 -11.92 -86.62 no yes – – 33.9 5.8 0.09930 2003-05-14 23600 20.8 -4.17 -37.46 -7.05 -41.28 -6.82 -46.30 no yes – – 56.8 25.8 0.95431 2003-05-15 988 23.8 -6.30 -45.25 -5.93 -31.26 -5.22 -33.26 yes no 65.7 1.1 – – 0.418

mean 4269 23.9 -2.40 -9.94 -2.84 -11.00 -2.63 -12.64 51.2 1.0 49.8 5.0 0.468SE 1141 0.4 0.54 4.63 0.58 4.30 0.65 4.44 7.6 0.3 9.4 2.7 0.066

N 30 31 31 31 31 31 30 31 17 16 10 9 31

!2H !2H !2H !2H

!!

!

!

!

−15 −10 −5 0 5

−100

−80

−60

−40

−20

020

●●

●● ●

●

●●●●

●●

●

●●●●

●●● ●● ●●

●●

●

●

●

●●●

●●

●●

●●●●

●●

●

●●●

●●●●

●

●

●

●●

●●

●●

●●

●

●

●

●

●

●

●

●

●

●

●

●

RGVOrainfogthroughfallδ2H = 7.6 δ18O + 3.9‰ GMWLRhodes et al. (2006)

δ18O [‰] in H 2O

δ2 H [‰

] in

H

2O

Costa Rica Monte Verde

0 5 10 15 20

020

4060

8010

0

TF amount [mm day−1]

Shar

e of

fog

wate

r in

TF [%

]

0.0 0.5 1.0 1.5

0.0

0.5

1.0

1.5

Calculated (mm day−1)

Mea

sure

d (m

m d

ay−1

) 1:1

0 1 2 3 4

0.0

0.4

0.8

!!!

!

!

!

●

●

●

●

●

●

●●●

●

●

●

●●

●

●●

●

●

●

RGVO rain fog throughfall

−10

−50

5 a ab b a

δ18O

[per

mil]

in H

2O(a)

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

RGVO rain fog throughfall

−100

−80

−60

−40

−20

020 a a b a

δ2 H [p

erm

il] in

H2O

(b)