Atmospheric deposition of nitrogen (N) and base cations was measured for 5–7 years on the island of Hawaii and for 1.5 years on Kauai. On Hawaii, mean annual fluxes of K+, Mg2+, and Ca2+ were 15, 17, and 13 kg ha−1 yr−1, respectively. Fog interception was the largest deposition pathway. Sea salt contributed the majority of cations, although biomass burning and Asian dust were significant sources for some years. Total N deposition (inorganic and organic) averaged 17 kg N ha−1 yr−1. Fog interception was also the largest source of N, depositing 16 kg N ha−1 yr−1. Precipitation deposition was 1.0 and 0.2 kg N ha−1 yr−1, respectively on Hawaii and Kauai. Dry deposition on Hawaii was 0.1 kg N ha−1 yr−1. Organic N averaged 16 and 12% of total N in rain and fog, respectively. The δ15N values for NO3−-N are consistent with long-range transport of N from Asia in the spring/summer and from North America in the fall/winter as nonvolcanic sources. Atmospheric deposition on Hawaii may completely account for a previously identified soil N imbalance.
 Atmospheric deposition of plant nutrients is of special interest in Hawaii due to its important role in ecosystem growth limitations, and also due to the way those limitations change through time. As Hawaiian ecosystems age from 0 to 4 Myr, processes shift from limitation by atmosphere-derived compounds, to limitation by compounds present in high concentrations in new volcanic substrates [Chadwick et al., 1999], as atmosphere-derived compounds accumulate in the soil and the substrate-derived compounds weather away. Therefore growth on the young island of Hawaii is limited by atmosphere-derived N [Vitousek et al., 1993], while growth on 4.1 Myr old Kauai is limited by initially substrate-derived phosphorus (P) [Chadwick et al., 1999].
 The atmosphere also becomes an increasingly important source of the base cations, potassium (K+), calcium (Ca2+), and magnesium (Mg2+), through time [Chadwick et al., 1999]. While recent research on base cation deposition has focused on their important role in neutralizing acid deposition [Hedin et al., 1994; Larssen and Carmichael, 2000; Lee et al., 1998], base cations are also important plant nutrients. On young islands, there is a high availability of cations in the soil because of high concentrations present in the relatively unweathered rock substrate. However, in wet, tropical climates, weathering can rapidly lead to very low availability of these materials [Vitousek and Sanford, 1986]. On Kauai, concentrations of cations in the soil are extremely low and atmospheric deposition is likely to be the only significant source of new material [Chadwick et al., 1999].
 We measured atmospheric nutrient deposition to a 300-year-old Hawaiian ecosystem from 1993 to 2000. Our research site was located on the island of Hawaii in the Hawaii Volcanoes National Park (Figure 1). The original motivation for this study was an apparent imbalance in the soil N budget. Crews et al.  found that a 33 kg N ha−1 yr−1 source was needed to account for the amount of N present in the soil. In situ fixation contributed only 1.2 kg N ha−1 yr−1 and preliminary precipitation deposition measurements revealed similarly low values [Vitousek, 1994]. We began measuring precipitation deposition in 1993, adding dry deposition measurements in 1994, and fog interception measurements in 1995. While our original motivation was to determine N fluxes (measured as nitrate (NO3−) and ammonium (NH4+)), analyses for chloride (Cl−), sulfate (SO42−), sodium (Na+), K+, Ca2+, and Mg2+ were performed concurrently.
 Initial deposition results indicated that large amounts of inorganic N are deposited by fog influenced by the nearby, active volcano Kilauea [Heath and Huebert, 1999]. In 1996, over 20 kg of inorganic N ha−1 yr−1 was deposited; the majority occurred during a small number of very concentrated, volcanically influenced fog events. Investigation of the volcanic N source led to the discovery of significant thermal fixation of atmospheric dinitrogen (N2) in air contacting the Kilauean lava flows [Huebert et al., 1999]. It appears that the volcano actually fertilizes ecosystems developing on new lava flows through this mechanism.
 Though the initial focus of this study was the atmospheric deposition of N, we found that the atmosphere was also a significant source of base cations. The majority (68%) of cation input was of marine origin [Coeppicus, 1999], which was expected, given the exposure of the site to the persistent, marine trade winds. However, this left 32% of the total cation deposition with an unidentified source. The transport and deposition of dust from Asia is the most likely source [Parrington et al., 1983; Leinen et al., 1994], but until now this possibility has not been evaluated. Quantifying this source is important for understanding base cation deposition and for evaluating the ways in which that deposition may change through time.
 N also has an important local source (the volcano), but as with cations, we have little information on nonvolcanic N sources. Long-range transport of material from Asia and North America is likely to contribute to Hawaiian N deposition, but there was no strong evidence to support this. The isotopic value of NO3− -N in samples can be useful in distinguishing among different sources of N [Wania et al., 2002; Yeatman et al., 2001]. We analyzed a subset of fog and rain samples in order to determine the isotopic signature of volcanically produced N, and to evaluate long-range transport sources.
 Kilauea Volcano exerts enormous influence on the N deposition to proximate ecosystems, contributing up to 75% of the total N deposition during some years (Heath et al., Volcanically produced nitrogen in Hawaii, submitted to Global Biogeochemical Cycles, 2002). As a result, N deposition measurements made near the volcano are unlikely to represent N deposition on other islands. Any differences between islands, in the deposition of N, as well as of base cations, are significant due to the differing nutrient limitations and availability on each island. In 1999, we began measuring precipitation deposition on the 4.1 Myr old island of Kauai to identify any differences in deposition rates.
 Our past N deposition results have represented inorganic N deposition only. In California's San Joaquin Valley, Zhang and Anastasio  measured organic N that was 16% of the total N in fog water, while at a Chilean site that is more comparable to ours, Weathers et al.  found organic N concentrations to be more than four times those of inorganic N. Scattered measurements by this and other groups indicate that organic N may be significant in Hawaii as well, representing 100% of the N present in some precipitation samples [Coeppicus, 1999; Cornell et al., 2001; Vitousek and Walker, 1989]. In 1999, we added analyses for organic N in rain and fog samples.
 Because fog concentrations can vary by more than 10 times between events [Heath and Huebert, 1999], sampling a small percentage of the events in a year may give a skewed annual average. During 1999, the number of fog events sampled was increased from an average of 9 to 48 events per year, so that our measured concentrations would be more representative of the actual annual mean concentration. This more comprehensive data set not only gives us higher confidence in our annual deposition results, but it also allows us to evaluate a critically important problem that is frequently encountered in deposition research: What is the best way to extrapolate intermittent data to longer periods of time?
 This paper presents a more comprehensive view of nutrient deposition to Hawaiian montane ecosystems than was possible in the past. Results from Kauai allow us to assess the variability of chemical deposition between two Hawaiian ecosystems. Organic N analyses make our N deposition results more complete. Our most current results from the Thurston site are presented along with past results that have been recalculated using an updated fog interception methodology. The result is chemical deposition data spanning 7 years: one of the most comprehensive records in the tropical Pacific and certainly the most complete in Hawaii.
2. Description of Sites
2.1. Thurston Site
 Our research site on the island of Hawaii was near the Thurston Lava Tube in Hawaii Volcanoes National Park (Figure 1). It is at 1190 m elevation, on the windward side of the island, and it was frequently influenced by orographic fog and rain. The average annual rainfall was between 2 and 3 m [Giambelluca et al., 1986] while fog interception contributed an additional 1.6 m annually (J. H. Carrillo and B. J. Huebert, Fog interception in Hawaii calculated with a water balance approach: Results and uncertianties, submitted to Journal of Hydrology, 2002). The forest canopy at this site was approximately 13 m high and the vegetation was predominantly Meterosideros polymorpha (Ohia lehua), Cibotium glaucum (tree fern), and Hedychium gardnerianum (Kahili ginger).
 Our research was conducted at two sites at approximately the same elevation. The majority of measurements were made from a 14 m tower in the middle of a 50 m by 30 m clearing. Measurements made at this location include rainfall amount (R), temperature (T), relative humidity (RH), wind speed (WS) and direction (WD), and liquid water content (LWC) of fog. This was also the location of our fog collector and aerosol/gas filters. A precipitation collector that collected samples for chemical analyses was located on the ground about 30 m away from the tower. Samplers on the tower were arranged across the predominant wind direction at varying heights to minimize interferences. From 1995 to the fall of 1998, when the large tower was installed, instruments were mounted on an 8 m tower in the same clearing. Prior to 1995, instruments were mounted on a 4 m telephone pole in the clearing. Throughfall (TF) and stemflow (SF) measurements (for our fog interception calculation described in section 3.4) were made at a second location approximately 1.5 km from the tower.
2.2. Kokee Site
 Our research sites on the island of Kauai were located in the Kokee State Park and in Na Pali Kona Forest Reserve. Both sites are at approximately 1130 m elevation and also received 2–3 m rainfall per year [Giambelluca et al., 1986]. We have no measurements of the Kokee fog interception amount, although because the sites had similar elevation and orientation to the trade winds, the fog water input may be similar to the 1.6 m measured at the Thurston site. Like the Thurston site, the vegetation was M. polymorpha (Ohia lehua) dominated forest, although it was somewhat lower in stature at approximately 10 m.
 The two sampling locations were separated by less than 2 km. The westernmost site was located on the western edge of a ridge, in an approximately 10 by 20 m clearing off of the road. This was the location of our precipitation collector (for chemical samples). It was also the location of the Hawaii Ecosystems Project's R, T, RH, WS, WD, and solar radiation measurements.
 The other site was to the northeast of the precipitation site, on the eastern-facing side of the same ridge. Our instruments were located on a 10 m tower surrounded by forest. The top section of the tower was approximately 1 m above the nearby vegetation. We measured R, T, and RH from this tower. WS and WD were measured from a pole extending about 1 m above the tower top.
3. Measurement Techniques
Table 1 summarizes our measurements at Thurston and Kokee. At Kokee, we have used Hawaii Ecosystems Project data for time periods when our instruments had either not yet been installed or were not functioning properly.
Table 1. Timetable of Measurements and Instruments Installed at Thurston and Kokee
 WS and WD were measured with RM Young propeller-vane anemometers from the top of the Thurston (since December 1993) and Kokee (since October 1999) towers. In April 1999, the RM young anemometer at Thurston was replaced with a Solent, Ltd. sonic anemometer. T and RH were measured from the top of the Thurston (since December 1993) and Kokee (since December 1999) towers using Vaisala T and RH sensors (either model HMP133Y, HMP35C, or HMP45C).
 Since October 1998, a Gerber Scientific, Inc. Particle Volume Monitor (PVM, model 100) was installed on the top section of the tower at Thurston. This measured fog LWC. The sum or average of all measurements was recorded every 10 min (Thurston) or every hour (Kokee) by a Campbell Scientific, Inc. 21X data logger. Radiation was measured by the National Park Service Air Resources Division using a LI-COR, Inc., LI-200SZ radiometer. This instrument was mounted on the roof of our field lab, about 4 m from the tower. Sums were saved on an hourly basis and were available since October 1999.
3.2. Precipitation Deposition
 Precipitation was collected for chemical analyses at Thurston and Kokee using Aerochemetrics, Inc., wet-only precipitation collectors. Precipitation was collected weekly since September 1993 at Thurston and since January 1999 at Kokee. The rainfall amount at both sites was measured with Texas Instruments, Inc. TE-525 tipping bucket rain gauge (0.254 mm per tip). Prior to 1995, the precipitation amount at Thurston was determined by weighing the precipitation bucket.
3.3. Dry Deposition
 Since January 1995 at Thurston, aerosol inorganic ions and nitric acid vapor (HNO3) concentrations were measured using 47 mm Teflon per nylon filter packs (1 μm Zeflour Teflon and Nylasorb nylon, both from Gelman Science). Meteorologically derived deposition velocities [Hicks et al., 1985] were used to calculate HNO3 deposition. To estimate the importance of aerosol deposition, we calculated deposition amounts using both a deposition velocity of 0.1 cm s−1 (used by Duce et al.  for the ocean surface) and 1.0 cm s−1 (calculated by Zhang et al. ) for a vegetated surface). Daytime (6:00 A.M. to 6:00 P.M.), nighttime (6:00 P.M. to 6:00 A.M.), and control (no airflow) filter packs were collected weekly. To reduce contamination by R or fog (F), the filters faced downward in an enclosure that was open only on the bottom. The filter pump was programmed to shut off when the precipitation collector opened or if the RH was over 95%. In November 1998 when the PVM was installed, the pump was set to shut off if the precipitation collector was open or if the LWC was above 0.005 g m−3. Filter measurements are not available for Thurston from the fall of 1996 to the fall of 1998. Dry deposition was not measured at Kokee.
3.4. Fog Interception
 Fog water was collected at Thurston since July 1995, using an active Teflon string collector [Daube et al., 1987]. Samples were collected intermittently, at random times throughout the year, from 1995 to 1998 and 2000. To better characterize the chemistry of Thurston fog, sampling was intensified for the calendar year 1999.
 The fog interception amount was calculated using a water balance approach shown in the following equation [Juvik and Nullet, 1993]
Here the water inputs, rain (R) and F, are equal to the sum of the measurable or calculable parameters TF, SF, canopy storage (CS), and evaporation (E). If we also measure R, we can solve for F as a residual term.
 TF was measured using four 6.4 m pieces of aluminum angle, arranged to create collection troughs, which drained into tipping bucket rain gauges. Aluminum or galvanized steel collars were used to collect SF from eight representative trees. CS was determined by examining “rain-only” (no fog) periods, using the relationship described by Juvik and Nullet . Finally, E was calculated using the Penman-Monteith equation [Monteith, 1965]. The details of these measurements, the fog interception calculation, and its uncertainty are presented in a related paper (J. H. Carrillo and B. J. Huebert, Fog interception in Hawaii calculated with a water balance approach: Results and uncertianties, submitted to Journal of Hydrology, 2002). It should be noted that, because fog interception is calculated as the difference between two often large terms, the uncertainty of this value can be substantial, with a mean annual deposition uncertainty of +39%/−30%. For fog chemical deposition, the fog interception amount dominates the uncertainties.
3.5. Sample Handling
 Details of cleaning and sample handling procedures are reported by Heath  and Coeppicus . Only new or altered procedures are described here.
 At the Thurston site, all fog samples analyzed for organic N were collected in Teflon bottles, while R samples were collected in a polyethylene bucket. After R and F samples were collected they were separated into aliquots that were analyzed for inorganic ions, organic N, P, isotopic N, and pH (in that order if there was not enough sample for all analyses). P results are presented in a related paper (Benitez-Nelson et al., Volcanic phosphorus deposition to the Hawaiian Island Chain, submitted to Biogeochemistry, 2002). Organic N samples were stored in cleanroom bags and isotopic N samples were stored in Nalgene bottles. Both were frozen until analyses.
 At Kokee, when the precipitation was collected, all but approximately 2 l of water was discarded. Chloroform (3–4 ml) was added to inhibit biological activity, and the sample was shipped to either the Thurston laboratory for processing, then to Honolulu or directly to the Honolulu laboratory. These samples remained unrefrigerated for up to 2 weeks. At the laboratory, aliquots were taken for inorganic ion and P analysis. Samples were refrigerated in Nalgene bottles. Since all Kokee samples had chloroform added, they were not analyzed for organic N or pH.
3.6. Sample Analyses
 Organic N samples were thawed and photolyzed for 3 hours with ultraviolet (UV) light using a Metrohm Model 705 UV Digester at 75°–85°C. The UV light oxidized organically bound N to NO3−, NO2−, or NH4+, which we analyzed with ion chromatography. Results were compared with unphotolyzed samples and the difference in total N was assumed to be organic N. In a comparison study, samples photolyzed for longer times yielded no additional N. No oxidant was added because this has been shown to increase contamination but not oxidation of sample N in precipitation [Cornell and Jickells, 1999; Scudlark et al., 1998].
 All the liquid samples and filter extracts were analyzed for inorganic ions (NO3−, NO2−, SO42−, Cl−, Na+, NH4+, and K+) using a Dionex 300 Series ion chromatograph. Analyses for Mg2+ and Ca2+ began in January 1997. Anions were analyzed on an OmniPac Pax-500 column with a 25 mM H2SO4 autoregenerant and a 1–35 mM NaOH per 5% methanol eluant solution. Cations were analyzed on an IonPac CS-12 column with a 20 mM HCl per 2mM MSA eluant and a self-regenerating cation suppressor.
3.7. Blank and Control Samples
 For the aerosol and gas filters, a weekly control filter pack was mounted on the sampling tower alongside the weekly sample filter packs. It was treated identically to the sample filters, except that no air was pulled through. Concentrations measured on the control filters were subtracted from the corresponding sample filter concentration for that week.
 For precipitation at both Thurston and Kokee, bucket blanks were taken for each weekly sample. Prior to installation in the precipitation collector, each collection bucket was soaked with deionized water for 1 week. An aliquot of the soaking water was saved and was processed identically to a precipitation sample. The concentration of the blank was subtracted from the measured precipitation concentration from that bucket.
 Prior to fog collection, the collector strings were sprayed with deionized water using a hand-held spray bottle. An aliquot of the rinsate was saved and processed identically to a fog sample. After rinsing, the collector was covered with plastic. If an event did not occur within 48 hours, the strings were rerinsed and a new blank was taken. Concentrations in the blank were subtracted from the first 4-hour sample of a fog event. We assumed that any contamination would have been collected in this first sample.
3.8. Nitrogen Isotopic Analyses of Nitrate
 The nitrogen isotopic composition of nitrate was measured in 26 fog samples and nine precipitation samples (six from Thurston, three from Kokee) collected during 1999 and 2000. An attempt was made to select samples collected under a variety of meteorological conditions. Measurements were made by the quantitative conversion of nitrate to N2O by bacterial denitrification, followed by isotopic analysis of the product N2O by continuous flow isotope ratio mass spectrometry [Sigman et al., 2001]. Individual analyses are referenced to injections of N2O from a pure gas cylinder and then standardized using an internationally accepted nitrate isotopic reference material, IAEA-N3 [Bohlke and Coplen, 1995]. Despite the high sensitivity of the denitrifier method, which requires ∼10–20 nmol N per analysis, sufficient sample volume was available for only a few of the precipitation samples, explaining the smaller number of isotopic analyses on precipitation relative to fog.
4. Data Analysis
4.1. Computing Fog Deposition
 Chemical deposition by fog was computed by taking the product of chemical concentrations in fog water and the fog interception amount. Because fog water was collected for chemical analysis intermittently, there are many events for which we have a water flux (measured continuously), but no chemical data. Additionally, due to instrument failure, there are a few periods of time for which we have chemical concentrations, but no water flux data, and periods of time with no data at all. In order to compute annual N fluxes, we must determine the best way to extrapolate our data to annual values.
Heath  used the average of three different approaches, since she had no way to determine which was the most accurate. For the first approach, an average N flux per fog interception event was computed for events with both chemical and water flux data. To extrapolate to an annual value, this average deposition was multiplied by the number of fog interception events in a year. For the second approach, rather than scaling the measured deposition by the number of events, an average N deposition per cm of fog interception was calculated (concentration may vary with size of the event) and this was then normalized to the cm of fog interception for the year. For the third approach, the average N concentration of all the collected fog water for a year was multiplied by the annual fog interception amount.
 The relatively large number of chemical data that we have for 1999 has allowed us to evaluate for the first time, the success of each of the three techniques at correctly estimating the deposition. There are 48 events for which we have both chemical and water flux data for 1999, compared to an average of six events per year for other years. For this analysis, we defined the total deposition as the measured deposition for the 48 events. We then randomly selected six events of the 48 and used the three approaches previously described to extrapolate to the total value. This was repeated with 12 different combinations of events and results were compared with the total measured deposition.
 We found that Methods 1 and 2 consistently overestimated the deposition amount, by as much as 10 times for N. Method 3 (the product of average concentration and the total water input) consistently produced results that were by far the closest to the measured values. For N, the deposition amount estimated using Method 3 was within an average of 8% of the actual deposition. For sea salt compounds, the estimated values were often within 10% of the actual values. Besides its higher degree of accuracy, an advantage of this approach is that it allows us to use all of our data, not just data for events with both chemical and water flux data.
5. Results and Discussion
5.1. Cation Deposition
 Cation deposition at Thurston shows the same trend as the N deposition [Heath and Huebert, 1999]: dry deposition is very small (less than 0.2 kg ha−1 yr−1 for cations); precipitation deposition is slightly larger (averaging 2–5 kg ha−1 yr−1 for K+, Ca2+, and Mg2+); and fog interception is by far the largest chemical input (averaging 10–15 kg ha−1 yr−1 for K+, Ca2+, and Mg2+; Tables 2–7). For each deposition mechanism, Tables 2–4 list species' concentrations while Tables 5–7 tabulate deposition fluxes. The deposition amounts and chemical concentrations in precipitation were within the range of values measured at other tropical locations [Cavelier et al., 1997; Vitousek and Sanford, 1986]. Few data exist on tropical cation deposition by fog. Our results are higher, but within the same range as those of Asbury et al.  in Puerto Rico. Our deposition results are higher than reported by Clark et al.  for a Costa Rican forest, in part because our fog interception amount is nearly twice as great.
Table 2. Precipitation Concentrations and Rainfall Amounts at Thurston and Kokee
 Weathering of the rock substrate at Thurston contributes almost 150 kg Ca2+ ha−1 yr−1 to the ecosystem, but this weathering input decreases rapidly with substrate age and is more than 8 orders of magnitude lower at Kokee [Chadwick et al., 1999]. While atmospheric deposition is not a negligible input at Thurston (an average total deposition of 15 kg Ca2+ ha−1 yr−1) it is an order of magnitude smaller than the weathering source. If we assume that fog interception contributes a similar Ca2+ input at Kokee, then the atmospheric source must sustain the Ca2+ needs of this ecosystem, as suggested by Chadwick et al. .
5.2. Biomass Burning
 A striking feature of our cation measurement results is the large interannual variability in the K+ concentrations and deposition amounts (Tables 2–7). The K+ values at Thurston for 1995–1997 were higher than other years for precipitation, while for fog, 1996 had higher values. The majority of this K+ deposition is contributed by a handful of weeks (for precipitation) or events (for fog) with very high K+ concentrations.
 It is possible that the elevated K+ we occasionally measure may result either directly or indirectly from Kilauea Volcano. Concentrations of K+ in fumarolic vapor condensate can be extremely high (28,000 μmol l−1 in condensate from Galeras Volcano; Alfaro and Zapata ). Though these high K+ samples contained other volcanic indicators, such as SO42− and Cl−, in elevated concentrations, this is common for samples from Thurston. However, the majority of other samples containing volcanic indicators do not have elevated K+ concentrations. Nonetheless, because of the variety of volcanic activity on Hawaii, it is possible that some of this elevated K+ may be a direct emission from the Pu’u ’O’o Vent.
 Another explanation is that these samples were indirectly influenced by Kilauea Volcano through biomass burning. Occasionally, when new lava outbreaks occur the flows intercept vegetation causing it to ignite. Elevated K+ concentrations in biomass burning plumes are ubiquitous [Andreae, 1983; Maenhaut et al., 1996; Pereira et al., 1996], so if our site was impacted by such a plume it is likely that it would result in elevated K+ concentrations in precipitation and fog [Lacaux et al., 1992]. In order to evaluate this possibility, we searched archived weekly reports of the Hawaiian Volcano Observatory (HVO). We found that on several occasions when we observed an elevated K+ concentration in precipitation, there was reported biomass burning due to new flow outbreaks onto forested land during our sampling period (Figure 2) [HVO, 1995, 1996, 1997]. While there are samples with elevated K+ concentrations for which we could not find reports of biomass burning, this does not preclude the possibility that there were unreported fires.
 The high K+ deposition that seems to result from biomass burning caused by volcanic activity represents another way in which the volcano actively contributes nutrients to proximate ecosystems. In this case, the K+ deposited to nearby ecosystems is not truly new material, the way the volcanically fixed N is, because it was originally part of the local plant biomass. However, in the absence of this atmospheric transport and deposition, the K+ in plant material covered by lava flows would have become unavailable as it was buried underneath layers of fresh lava. In this sense, this mechanism represents a pathway for material in condemned ecosystems to be recycled rapidly.
5.3. Asian Dust Deposition
 If we assume all of the Na+ found in our samples is of marine origin, we may compute the amount of non–sea salt (NSS) derived K+, Mg2+, and Ca2+ by comparing their ratios to Na+ in the sample to those in sea salt (SS). Figure 3 shows SS and NSS deposition amounts of the base cations in precipitation at Thurston and Kokee for 1999 through June of 2000, the period of time for which the sampling was done at both sites. The left side of Figure 3 shows the SS components of cation deposition. For these marine-derived elements, there is both a wintertime and springtime peak, resulting from increased storm activity during these times of year. Higher wind speeds result in the generation of more sea-spray. It seems likely that with additional years of data, these two peaks might merge into one, lasting from November until May.
 The NSS-derived base cation deposition is shown on the right side of Figure 3. For K+ and Mg2+, there is very little material that is not associated with SS. However, Ca2+ shows very clear springtime peaks at both Thurston and Kokee, most likely due to the transport and deposition of dust from Asia.
 Springtime dust storms over central Asia suspend tremendous amounts of material in the air. This material may be transported great distances in the free troposphere, frequently reaching Hawaii and occasionally even the continental United States [Braaten and Cahill, 1986; Betzer et al., 1988; Parrington et al., 1983; Uematsu et al., 1983]. In the measurements of springtime Asian dust taken by our group off the coast of China during the ACE-Asia field project, the dust we measured was characterized by extremely high concentrations of Ca2+ (as high as 730 nmol m−3), with only slight enhancements of K+ and Mg2+ (14:1:1, molar ratio of Ca2+:K+:Mg2+).
 These enhanced aerosol Ca2+ concentrations, with smaller peaks in aerosol K+ and Mg2+, are consistent with what we observe in Hawaii, both at our Thurston and Kokee sites as well as at the Mauna Loa Observatory (MLO), where our group has been making aerosol measurements since 1987 (methodologies described by Lee et al. ). The nightly Teflon/nylon filter packs sample predominantly free tropospheric air [Lee et al., 1993]. Monthly concentrations of aerosol base cations at Thurston and MLO are plotted in Figure 4. The Ca2+ peak that we see at Thurston appears as a more pronounced feature at MLO. The smaller Mg2+ peak appears in the Thurston aerosol as well as in Thurston and Kokee precipitation. While the K+ data are less clear, we should note that our K+ measurements are often very near our detection limit. Note that the apparent negative NSS concentrations for K+ and Mg2+ are possible since NSS is calculated by difference and SS concentrations are typically many times NSS concentrations in Thurston aerosol.
 This same annual pattern of cation deposition is apparent in the fog deposition at Thurston as well. Figure 5 is similar to Figure 3, but plotted for Thurston fog since 1995. As we have noted previously, we collect fog intermittently so that we do not have as evenly distributed data as for precipitation and dry deposition (although in the figure, values have been extrapolated to monthly values). This fact makes the very similar pattern that we see in fog at Thurston even more remarkable. Like precipitation, there is very little deposition of NSS K+ or Mg2+, while NSS Ca2+ deposition is substantial in the spring and early summer. Additionally, the spring peak in SS is apparent. This suggests that the spring peak in precipitation SS deposition is not solely due to increased rainfall, but to increased atmospheric concentrations as well.
5.4. Precipitation N Deposition
 Our longest and most reliable N deposition results are for precipitation deposition at Thurston. The mean rainfall total N deposition is 1 kg N ha−1 yr−1 (Table 3), which is within the range of values found for nearby sites [Harding and Miller, 1982; Vitousek et al., 1993]. The deposition is rather consistent from year to year. Lower measured rainfall and inorganic N deposition during 1995 and 2000 are likely a result of sampling primarily during the first, typically drier, half of the year. NO3− and NH4+ are roughly equivalent on average, although in a given year (or week) one or the other may dominate significantly. This variability may be a result of changes in volcanic activity. While oxidized N is formed in air near the lava flows, reduced N should be primarily emitted at Pu’u O’o Vent. Which of these two related but different sources influences our site will depend on the nature of the volcanic activity and the wind direction at the time (J. H. Carrillo et al., Volcanically produced nitrogen in Hawaii, submitted to Global Biogeochemical Cycles, 2002).
 There is seasonality in rainfall, with relatively more rainfall during the winter months from November to January; the amount of N deposition follows a similar trend as the amount of rainfall from month to month (Figure 6). This is not generally true for the interannual deposition amount. The frequency of rain events is consistent from year to year, with a rain event occurring on average less than every 2 days (J. H. Carrillo and B. J. Huebert, Fog interception in Hawaii calculated with a water balance approach: Results and uncertianties, submitted to Journal of Hydrology, 2002). Higher rainfall years result primarily from heavier, not more frequent, rainfall. Intensive rain sampling has shown that most N is scavenged during the beginning of a rain event, so longer or heavier events should not necessarily result in more N deposition. The higher N deposition during months with more rainfall may result from the typically higher event frequency during those months, rather than from larger rainfall amounts.
5.5. Dry N Deposition
 Dry deposition of HNO3 and aerosol NO3− and NH4+ is the smallest atmospheric N source at Thurston, depositing on average less than 0.2 kg N ha−1 yr−1 (Table 5). In recent years, our estimated dry deposition rose significantly due to changes in our instrumentation and sampling procedures that have affected our Vd calculations (note that concentrations have not changed significantly; Table 4). Figure 7 shows monthly dry deposition from September 1998 to July 2000, all measured using the new procedures. During this time, the filter packs and anemometer were mounted well above the forest canopy.
 In November 1998, we began using the PVM data rather than RH to determine when “dry” (no rain or fog) conditions were met, resulting in more measured dry time (Table 5). Indeed, a comparison has shown that our RH sensor significantly overestimates the amount of fog (J. H. Carrillo and B. J. Huebert, Fog interception in Hawaii calculated with a water balance approach: Results and uncertianties, submitted to Journal of Hydrology, 2002). As a result, data since November 1998 are more likely to represent the actual dry deposition amounts. HNO3 deposition, the largest dry input, is highest during the summer months, which typically receive less rainfall. This was exaggerated during the spring of 2000, when rainfall amounts were significantly below climatological values.
 Our calculated HNO3 deposition velocities are higher than most reported values [Duce et al., 1991; Huebert and Robert, 1985], but are probably realistic. The forest canopy at Thurston is directly exposed to the trade winds and 10 min averaged wind speeds are near 5 m s−1, with gusts that are much faster. Additionally, the rough topography of the forest canopy results in more air turbulence than over grassland or ocean. The relatively high calculated HNO3Vd suggests that particulate deposition may also be relatively high. Although we used a range of published values in our calculations, a 1 cm s−1Vd for particles may be more realistic for our location than 0.1 cm s−1. Since we have not measured or calculated a Vd for particles, our estimates contain large uncertainty. However, even using this large Vd, the very low estimated aerosol dry deposition indicates that we have not overlooked a major inorganic N deposition source.
5.6. Fog N Deposition
 N deposition by intercepted fog is consistently the largest N source to the Thurston site, depositing an average of 16 kg N ha−1 yr−1 (Table 7). Past reports of intercepted fog at Thurston have suffered from relatively sparse chemical data and from instrumentation failures that resulted in missed water input sampling days. During 1999, we sampled many more events. This, in addition to an improved procedure for estimating fog interception amounts, gives us greater confidence in the accuracy of our results.
 The fog interception amount (water deposition only) does not exhibit the same interseasonal variability that precipitation does (Figure 8), but is fairly consistent throughout the year. However, fog N deposition has a peak in the spring-summer and is lowest during the fall months, because of interseasonal differences in the N concentration in fog. Because the trade winds are weakened during the late spring and summer, there is a higher probability of volcanically influenced (and often N-rich) air reaching our site.
 The original motivation for this study was to explain an apparent imbalance in the soil N budget at the Thurston site. Our estimated annual fog deposition of 16 kg N ha−1 yr−1 accounts for much of Crews' 33 kg N ha−1 yr−1 of missing N source [Crews et al., 1995]. Given the large interannual variability for fog deposition and the uncertainty associated with our estimate of this term, it seems likely that N deposition by fog interception is the source of virtually all of the previously unaccounted for N in the soil.
 Fog deposition on other Hawaiian islands may be significantly lower than at Thurston. In an analysis of the volcanic source of inorganic N, it was determined that between 22 (NH4+) and 38% (for NO3−) of the deposited N at Thurston was volcanic in origin during 1999 (J. H. Carrillo et al., Volcanically produced nitrogen in Hawaii, submitted to Global Biogeochemical Cycles, 2002). Though we have not yet determined the amount of volcanic N that reaches the other islands in fog, it is likely to be less than Thurston. On all islands, the amount of volcanic N deposition also should be highly variable, depending on the nature of volcanic activity and on meteorological conditions that transport volcanic N.
5.7. Organic N Deposition
 The average annual deposition of organic N by precipitation was 0.1 kg N ha−1 yr−1 and by fog was 2 kg N ha−1 yr−1. Organic N was 16% of the total N in precipitation and 12% of the total N in fog. The greater number of sampled fog events in 1999 suggests that the 17% measured this year is more representative of long-term averages (Table 7), and indeed it is consistent with the precipitation value. The percentage of organic N we saw in fog and precipitation was less than the 70% measured by Vitousek and Walker at a nearby site over a 19-week period of time [Vitousek and Walker, 1989]. These results may not be completely inconsistent, however. Within our yearlong average are the periods of time when organic N dominates, although the largest fraction of organic N over any 19-week period during our measurements was 45%.
 It is possible that we underestimated organic N in precipitation due to evaporation and losses on the polyethylene buckets [Scudlark and Church, 1988]. Additionally, we may overestimate it at times due to conversion of inorganic N to organic N by biological activity over our weeklong collection period, particularly during the summer months. Because fog is sampled on an event basis and Teflon bottles are used for collection, many of the potential artifacts of our precipitation sampling are avoided. The fact that the fraction of organic N in rain and fog is similar suggests that either the positive and negative artifacts in precipitation roughly balance on average, or that they are small.
5.8. Precipitation N Deposition at Thurston and Kokee
 Since drought conditions persisted at Kokee over much of the measurement period, it is unclear how representative the chemical fluxes are. Nonetheless, precipitation inorganic N deposition at Kokee was disproportionately lower than at Thurston (Figure 9) most likely due to the proximity of Kilauea Volcano to the Thurston site (J. H. Carrillo et al., Volcanically produced nitrogen in Hawaii, submitted to Global Biogeochemical Cycles, 2002). Precipitation inorganic N deposition at Kokee should be more representative of inorganic N deposition on islands without active volcanism.
5.9. N Isotopic Composition of Nitrate in Rain and Fog
 Stable isotopic composition can sometimes aid in the identification of different sources or reaction pathways for nitrogenous species. The 15N to 14N ratio of a sample is expressed relative to atmospheric N2, as follows
 A broad range in δ15N was measured for NO3− in both our fog and rain samples. For fog, values ranged from −3.0 to +5.9‰, while for rain the range was from −3.8 to +2.4‰ (Table 8). The majority of the samples that we believe contain significant volcanogenic N (J. H. Carillo et al., Volcanically produced nitrogen in Hawaii, submitted to Global Biogeochemical Cycles, 2002) have δ15N values between 0 and +2‰. Since these values fall within the range of our background variability, it is difficult to use the isotopic results to distinguish volcanically produced N. However, the δ15N results may be helpful in indicating likely sources for the background (i.e., nonvolcanic) N. In this context, there is an apparent shift in δ15N measured in both fog and precipitation, from positive δ15N values from about April through September to negative δ15N values for the rest of the year (Figure 10), which may be due to a change in the N source.
Table 8. Isotopic N Values of NO3−-N in Fog and Precipitation Samples From 1999
δ15N of NO3−
−3.0 to +5.9%
−2.9 to +2.4%
−0.4 to −3.8%
 It is likely that much of the N we measure during trade wind weather arrives by long-range transport from continents. In a comparison of model predictions and measurements of SO42− at MLO, Huebert et al.  report that North American sources are most important from July through September, whereas Asian sources dominate from October through June. Lee et al.  found the similar pattern for total nitrate at MLO using the back trajectory analyses summarized in Figure 11 (redrawn from Lee et al. ). On this map, each number represents a monthly average starting point for 10-day air mass trajectories that arrive at MLO. From November to May, these starting points lie west of Hawaii. This is consistent with the peaks in Ca2+, both at MLO and in our boundary layer sites, presumably from Asian dust, between March and June (Figures 3–5).
 What δ15N would we expect for nitrate from Asia versus North America? According to Galloway , N exported from Asia is due to the roughly equal sources of fertilizer use and fossil fuel burning. Though we know of no isotopic measurements specifically for Asian fertilizer or fossil fuel, in other locations, both fertilizer NO3− (−2 to +5‰; Heaton ) and emissions from coal-fired power stations (+6 to +13‰; Heaton ) tend to be relatively high in δ15N. Alternately, a major source of oxidized N from North America is vehicle exhaust [Logan, 1983], which tends to be relatively low in δ15N (−2 to −13‰; Heaton ). Therefore the high δ15N we observe from April to August at Thurston is consistent with an Asian source and the low δ15N we observe from August to April is consistent with a North American source. These N isotope results for our marine boundary layer sites at Thurston and Kokee seem to be consistent with the long-range transport sources determined for the free troposphere at MLO.
 In an analysis of the vegetation at the Thurston site, Vitousek et al.  found that leaves, soils, and N fixers were all depleted in 15N, with δ15N values ranging from −7.7 to −1.2‰. The authors proposed that atmospheric deposition may contribute 15N depleted N, resulting in the low δ15N values they observed. If atmospheric N is responsible for the low δ15N values, it must be in the form of either NH4+-N or organic N, since the volume weighted mean δ15N values of NO3−-N were close to zero, ranging from +0.8‰ for Thurston fog to −0.03‰ and −2.9‰ for Thurston and Kokee precipitation (Table 8). It is in fact quite likely that negative δ15N values in Thurston vegetation are because of NH4+-N deposition. NH4+-N can represent as much as 75% of the total N deposition during some years (Tables 3, 5, and 7), and measurements of the δ15N of atmospheric NH4+ at other locations show that it is typically lower than that of oxidized N at the same location [Freyer, 1978; Garten, 1992; Heaton, 1987]. If the NH4+ ions deposited at Thurston were depleted in 15N, the result would be a net low-δ15N source, since the δ15N of NO3− is near 0‰. Values for δ15N of atmospheric organic N span a broad range (−7.3 ± 7.3‰, Cornell et al. ), but tended to be lower in more remote regions. This implies that organic N may also be a low-δ15N source at our site.
 Atmospheric deposition is an important source of N and base cations to Hawaiian ecosystems and the major pathway for deposition is via fog interception. Though many researchers have observed concentrations in fog that are higher than precipitation for a site [e.g., Jordan et al., 2000; Vong et al., 1997], results from this study indicate that in Hawaii, high fog water fluxes combine with several strong, sporadic chemical sources to result in higher deposition rates of several species than might be expected for a remote, marine environment. High deposition rates for chemical species in fog have been observed in many polluted environments and are a concern with regard to acid and metal deposition [e.g., Dollard et al., 1983; Harvey and McArthur, 1989; Herckes et al., 2002; Igawa et al., 1998]. Our study area is unusual in that the chemical sources determined for our sites are largely natural.
 Measured annual fluxes of K+, Mg2+, and Ca2+ averaged 15, 17, and 13 kg ha−1 yr−1, respectively and fog interception was by far the largest deposition pathway. While sea salt contributed the majority of cations, local biomass burning and Asian dust were also significant sources for some years. Though sporadic in nature, the influence of Asian dust, and owing to the volcanic nature of the Hawaiian Islands, biomass burning, both are sources that are likely to have been important historically.
 Fog interception is also the largest source of N to Hawaiian ecosystems, with a measured deposition rate of 16 kg N ha−1 yr−1 at the Thurston Lava Tube site. Precipitation and dry deposition contribute an average of 1.0 and 0.3 kg N ha−1 yr−1 ,respectively. Precipitation deposition on Kauai contributed 0.2 kg N ha−1 yr−1, although drought conditions persisted at Kokee during much of our measurement period. Organic N was on average 16 and 12% of the N in rain and fog, indicating that it is an important N source. For both rain and fog, there were samples for which 100% of the measured N was organically bound. In fact, during much of the summer of 1999, all of the precipitation N was organic.
 Values for NO3−-δ15N in fog ranged from −3.0 to 5.9‰ and in rain from −3.8 to 2.4‰. While δ15N values for volcanically produced NO3−-N fell within the range of background values, the apparent seasonal isotopic variations in NO3−-δ15N may be because of different nonvolcanic N sources. δ15N tended to be positive during the spring and summer, consistent with long-range transport of material from Asia, and negative during the rest of the year, indicating long-range transport of N from North America. This result is consistent with our measured Ca2+ deposition seasonality, and with free tropospheric sources determined for MLO.
 Volume weighted mean δ15N values were near zero, indicating that the strongly negative δ15N values measured by Vitousek et al.  in soils and leaves at the Thurston site are not a direct result of atmospheric NO3− deposition. NH4+ or organic N deposition may provide the light source, however, since other studies indicate that the δ15N of NH4+ tends to be lower than that of NO3− in atmospheric samples and that organic N tends to be low in δ15N in remote regions.
 Our measurements at Thurston indicate that atmospheric deposition may completely account for the apparent soil N imbalance [Crews et al., 1995]. The 17 kg total N ha−1 yr−1 we measure has large interannual variability. Since much of the deposited N is volcanic in origin, during the years of enhanced volcanic activity, deposition rates are likely to be higher.
 We are grateful to Liangzhang Zhuang for performing most of the chemical analyses, to Claudia Benitez-Nelson for her help in editing the text, and to Peter Vitousek for his encouragement and insightful suggestions. Many samples were collected by Sirit Coeppicus, Karin Schlappa, and David Alexander. This work was supported by the Andrew W. Mellon Foundation, NSF grants ATM-9816637 and ATM-9807631 to B.J.H., and NSF grant OCE-9981479 to D.M.S.