Field studies of watershed carbon fluxes and budgets are critical for understanding the carbon cycle, but the role of deep regional groundwater is poorly known and field examples are lacking. Here we show that discharge of regional groundwater into a lowland Costa Rican rainforest has a major influence on ecosystem carbon fluxes. This influence is observable through chemical, isotopic, and flux signals in groundwater, surface water, and air. Not addressing the influence of regional groundwater in the field measurement program and data analysis would give a misleading impression of the overall carbon source or sink status of the rainforest. In quantifying a carbon budget with the traditional “small watershed” mass balance approach, it would be critical at this site and likely many others to consider watershed inputs or losses associated with exchange between the ecosystem and the deeper hydrogeological system on which it sits.
 Quantitative understanding of carbon cycling in ecosystems is a topic of ongoing interest to geochemists, ecologists, and hydrologists, with important links to climate change [Battin et al., 2008; Cole et al., 2007]. The fundamental questions of whether some ecosystems operate as net sources or sinks of CO2 to the atmosphere, and whether or when they may flip from sink to source upon warming, remain the focus of active inquiry and sometimes conflicting results. These questions are significant because ecosystems operating as net sources represent a positive feedback on warming.
 A watershed is often used as a convenient subset of an ecosystem for organizing measurements and analysis of the carbon budget. Watersheds have long been used as practical field units for determination of water and solute fluxes and budgets [e.g., Likens and Bormann, 1995], and doing so for carbon is in keeping with growing awareness of the critical connections between terrestrial water and carbon fluxes [Cole et al., 2007]. A potentially large but relatively unstudied factor in ecosystem carbon fluxes is the discharge of regional groundwater that is often high in dissolved carbon.
 Interbasin groundwater flow (IGF), groundwater flow beneath surface topographic divides from one basin or watershed to another, is the natural hydrogeological process responsible for long-distance movement of regional groundwater from upland recharge areas to streams and wetlands in lowland watersheds [Tóth, 2009]. Schaller and Fan  argued for the importance of IGF to climate modeling efforts on the basis of the water and heat energy transported. Here we focus on the carbon transported by IGF and its role in the watershed carbon budget. The fundamental motivating questions include the following: What field data are needed to know whether a rainforest (or other ecosystem) is a net source or sink of carbon, can regional groundwater be important, what measurable ecosystem signals (chemical, isotopic, or flux) are available to help decide, what are the implications for carbon fluxes in streams, and what are the potential errors if regional groundwater is important but ignored?
 We assessed the influence of IGF on carbon fluxes and budgets in two small adjacent watersheds at La Selva Biological Station in the lowland tropical rainforest of Costa Rica (Figure 1). The watersheds are identical or nearly so in all major features (rainfall, temperature, forest cover, soils, etc.), with one exception: the Arboleda watershed receives a significant influx of 3000–4000 year old regional groundwater via IGF, while the Taconazo does not [Genereux et al., 2009, 2005]. The Taconazo has only young local groundwater several years or less in age [Solomon et al., 2010]. We utilized multiple chemical and isotopic signals for carbon, combined with hydrologic data to estimate fluxes.
 Stream discharge (m3/s) was measured every 15 minutes at V notch weirs on the two watersheds during 2006–2009 (data are available at http://www.ots.ac.cr/meteoro/default.php?pestacion=2). Precipitation was measured at two tipping bucket rain gauges (one above the forest canopy on a tower and one about 2 m above ground in a forest clearing). Water samples for chemical analysis were collected on a weekly basis at the weirs, supplemented with additional event-based sampling. Details are in Zanon , Nagy , and Zanon et al. . Carbon export by streamflow was estimated using the flow-weighted mean concentration approach [Walling and Webb, 1985; Birgand et al., 2010]. Measurements of the CO2 content of riparian air were made with gas analyzers (a LI-840 from LI-COR Inc., Lincoln, NE, and GMT 222 from Vaisala, Helsinki, Finland), simultaneous with collection of air samples for isotopic measurement in Exetainer sample vials (Labco, Buckinghamshire, UK) using a 30 ml syringe. Carbon isotopic measurements were made at the NOSAMS facility at the Woods Hole Oceanographic Institution [dissolved inorganic carbon (DIC)], at the University of California at Davis (CO2 in air samples), and at North Carolina State University [dissolved organic carbon (DOC)].
 Comparing between the Taconazo and Arboleda streams (at the weirs), IGF increases the concentration of DIC by a factor of about 12, and stream export of DIC by a factor of about 70 (Table 1). IGF lowers the stream DOC concentration (old regional groundwater is lower in DOC than young groundwater) but increases DOC export by a factor of 3.5 (because of the large additional water throughput from IGF). The Taconazo DOC and DIC export values fall within published ranges for other small watersheds, while the Arboleda values augmented by IGF are higher (Figure 2 and Supplemental Table 1).
Table 1. Comparison of Lowland Rainforest Watersheds With (Arboleda) and Without (Taconazo) Interbasin Groundwater Flow (IGF) of Old Regional Groundwater, La Selva Biological Station, Costa Ricaa
 Elevated DIC concentration and DIC and DOC export in the Arboleda are due to the large IGF into the Arboleda (about 10 m3 of water per m2 of watershed per year, or 10 m per year), much of which is high-DIC (14 mM) regional groundwater. The carbon input to the Arboleda by IGF was estimated to be about 870 gC/m2yr, a value that is 24–32% of the magnitude of whole ecosystem respiration at La Selva [Cavaleri et al., 2008; Loescher et al., 2003]. The IGF carbon input is also at the upper end of the range for net ecosystem exchange (NEE) of CO2 with the atmosphere at La Selva: −5 to 800 gC/m2yr (a positive value indicates an ecosystem sink), depending on the year and method of NEE estimation [Loescher et al., 2003]. In other words, the net carbon input to the Arboleda watershed “from below” (by IGF) is at least as large as the net input “from above” (NEE).
 The low 14C and high δ13C of the DIC reaching the Arboleda from below are strongly consistent with the isotopic signature of magmatic CO2 [Genereux et al., 2009, and references therein]. That is, the carbon entering the Arboleda in association with the water from IGF is from a geological source and not a result of any modern ecosystem process that differs between the Arboleda and the Taconazo.
 Degassing of CO2 from surface water has been shown to be a potentially significant carbon flux for terrestrial ecosystems [Oquist et al., 2009; Johnson et al., 2008; Teodoru et al., 2009; Richey et al., 2002; Hope et al., 2001]. We expect that the degassing flux from streams is much larger in the presence of IGF. An estimate of this flux based on the measured aqueous CO2 concentrations in the Taconazo and Arboleda streams, the approximate stream surface areas, and an estimate of 75 day−1 for the first-order degassing rate constant (a reasonable value for small shallow streams) [Hope et al., 2001; Genereux and Hemond, 1992] suggests about four times more CO2 degassing from the Arboleda stream than the Taconazo stream. Measurements of CO2 concentration and the δ13C of CO2 in air just above the stream water surface are consistent with an enhanced flux of isotopically heavy CO2 from the Arboleda stream. We sampled air above streams in the early morning (March, June, July, October, and November 2012) before daytime mixing of the canopy air began and found higher CO2 concentration and heavier δ13C–CO2 in air above the Arboleda stream compared to air above the Taconazo (Table 1). Concentrations of CO2 were highest in the Arboleda weir splash zone, but isotopically heavier CO2 was also found above more quiescent stream water about 150 m upstream of the Arboleda weir.
 Measurements of UV-visible light absorbance by dissolved organic matter (DOM) suggest that IGF of old groundwater alters the chemical nature of the DOM in streams as well as the concentration and export flux of its constituent carbon (DOC; Table 1). Slope ratio SR was determined for the Taconazo and Arboleda stream DOM, and for groundwater from Guacimo Spring, a large spring discharging high-DIC regional groundwater. SR is the ratio of the slope values from a linear fit, in two different wavelength ranges of light absorbance, of the logarithm of light absorbance versus wavelength. Larger SR values are associated with DOM that is relatively low in molecular mass and/or weakly aromatic [Helms et al., 2008; Spencer et al., 2012]. With regard to SR, the Arboleda was more variable than the Taconazo, and intermediate in magnitude between the Taconazo (local groundwater) and Guacimo Spring (regional groundwater), likely reflecting time-varying mixing of local and regional groundwaters (each with distinct DOM) in the Arboleda (Figure 3).
SR data indicate a qualitative difference in DOM chemistry between old regional groundwater and young local groundwater, likely that the former has become less aromatic and/or lower in molecular mass through partial microbial degradation during its long subsurface residence time. We hypothesize that older degraded DOM from regional groundwater is less bioavailable in rainforest streams compared to younger fresher DOM. If true, this would suggest that IGF alters watershed export of DOC by two mechanisms: additional input of DOM to the watershed, and input of DOM that is less bioavailable and thus more likely to experience hydrologic export from the watershed and longer riverine transport. Also, given the similarity in δ13C–DOC values between the Arboleda and Taconazo streams, it is unlikely that much of the geological DIC in the Arboleda is taken up there by photosynthesis. Uptake may occur downstream if streams receiving high-DIC IGF leave the rainforest where stream algae are light limited [Pringle and Triska, 1991] and enter pasture or other deforested areas.
 IGF clearly alters carbon concentrations and fluxes, and the chemistry of DOM, in the Arboleda watershed at La Selva. Using La Selva as an example, a preliminary conceptual carbon flux diagram (Figure 4) illustrates how knowledge of IGF could help avoid incorrect conclusions about the carbon source/sink status of a watershed.
 Estimates of NEE at La Selva span a wide range from −5 to 800 gC/m2yr, depending on the year and calculation method used [Loescher et al., 2003]. On either watershed, considering direct CO2 degassing from the stream separately from NEE (a possibility suggested for other sites by Dinsmore et al. , Cole et al. , Billett et al. , and Hope et al. ) could, at lower NEE values, shift the watershed from a net sink to a source of carbon. Bringing stream export of DIC and DOC into the budget picture does not have a major impact on source/sink status for the Taconazo, but it does for the Arboleda.
 Summarizing for the Arboleda, in the presence of significant IGF:
 assuming NEE alone represents the carbon budget would suggest that the watershed is on average a clear sink for CO2, and
 consideration of NEE, stream export, and stream degassing, without knowledge (based in part on the water budget) of the carbon input from IGF, would give the opposite conclusion, that the watershed is a clear source of CO2.
 Correct understanding of the watershed source/sink status requires (1) knowledge of the carbon input by IGF (which supports the large stream export flux of carbon), and (2) field estimation of all fluxes during the same time period (inter-annual variation may be large, as noted above for NEE).
 Other examples of the importance of IGF to ecosystem carbon budgets are lacking in the literature, but the widespread occurrence of the two key factors (IGF and elevated dissolved carbon in regional groundwater) gives strong reason to believe that IGF may affect watershed carbon fluxes at other sites, with impacts ranging from small (and difficult to detect) to large (such as found at the Arboleda). Regional groundwater may acquire elevated dissolved carbon from magmatic outgassing, dissolution of carbonate minerals, dissolution-respiration-methanogenesis of sedimentary organic matter, or migration of carbon compounds from petroleum deposits. Even considering just magmatic outgassing alone, the extent of IGF-based effects on ecosystem carbon fluxes and budgets may be large, given that high topographic relief and active volcanism coincide over large areas (the entire Pacific rim, east Africa, parts of the northern Mediterranean, etc.). High DIC has been found in groundwater in many such areas, e.g., up to 45 mM [Evans et al., 2002] in the western United States, up to 28 mM [Chiodini et al., 2000] in Italy, and up to 65 mM [Ohsawa et al., 2002] in Japan (the 14 mM from La Selva is not at the high end of the range globally).
 Elevated DIC is well known from carbonate rock aquifers, e.g., up to 5 mM in the Floridan aquifer of Florida [Plummer and Sprinkle, 2001] and up to 15.9 mM in the Great Lakes region of the United States [McIntosh and Walter, 2006]. Significant DIC, DOC, and/or dissolved methane occur in many primarily clastic regional aquifers in nonvolcanic areas, such as the U.S. Atlantic coastal plain, where DIC of 10–14 mM occurs in confined aquifers in the Carolinas [Chappelle and Lovley, 1990; Kennedy and Genereux, 2007], and lower but significant DIC (up to 3.3 mM) occurs in the Aquia aquifer of Maryland [Aeschbach-Hertig et al., 2002]. In the central United States, Clark et al.  found DIC up to 9 mM in the Dakota aquifer, and McMahon et al.  found DIC up to 4.3 mM in the High Plains aquifer. Murphy et al.  found DIC up to 14.2 mM, DOC up to 1.4 mM, and methane up to 13.3 mM in the Milk River aquifer, Canada. Aravena and Wassenaar  found DOC up to 1.5 mM and methane up to 4.7 mM in the Alliston aquifer, Ontario, Canada.
 Also, IGF is an expected part of the hydrologic cycle with a theoretical foundation in the relationship between topography and groundwater flow paths at multiple spatial scales [e.g., Worman et al., 2007; Cardenas, 2008; Tóth, 2009]. IGF has been detected worldwide in both high- and low-relief topographic settings [e.g., Genereux et al., 2005, and references therein; Tóth, 2009; Kasper et al., 2010], though its global extent and magnitude are not fully known, in part because it can be costly to quantify, and perhaps in part because areas showing evidence of IGF may be avoided as long-term field research sites (IGF may be viewed as an unwanted complication in the determination of water or element fluxes and budgets in an experimental watershed). New research [e.g., Gleeson and Manning, 2008; Frisbee et al., 2011; Gardner et al., 2011; Smerdon et al., 2012] continues to advance the hydrogeology of IGF and large-scale groundwater flow paths to streams, but the significance for carbon budgets and fluxes remains relatively unexplored.
 We suggest that this is a significant gap, and an opportunity, in the study of carbon fluxes and the carbon source/sink status of watersheds and ecosystems. The connection between ecosystems and the deeper hydrogeological systems on which they sit may have strong relevance to understanding the carbon cycle and is ripe for further study.
 Financial support from the U.S. National Science Foundation (awards 0421178 and 1029371) and U.S. Department of Energy (award DE-SC0006703) is gratefully acknowledged. Logistical support at the field site was provided by the Organization for Tropical Studies, especially William Ureña. Preparation of Figure 1 was by Carlo Zanon.