Auxiliary Material Submission for Paper 2004gb002390 Empirical and mechanistic models for the particle export ratio John P. Dunne (NOAA Geophysical Fluid Dynamics Laboratory) Robert A. Armstrong (Marine Sciences Research Center, Stony Brook University, Stony Brook, NY, 11794) Anand Gnanadesikan (NOAA Geophysical Fluid Dynamics Laboratory) Jorge L. Sarmiento (AOS Program, Princeton University, P.O. Box CN710, Princeton, NJ 08544) Introduction Supplemental Material 1: The first supplemental dataset describes observations utilized in the particle export ratio synthesis. Variables include average latitude (Lat), average longitude (Lon), mixed-layer/sea surface temperature (SST; degrees C), depth integrated chlorophyll (mg Chl m-2), depth integrated productivity (Prod; mmol C m-2 d-1), particle export estimate at the depth of the euphotic zone (Part; mmol C m-2 d-1), euphotic zone depth (Zeu; m; 1% light level or sampling zone, if less), the carbon to chlorophyll ratio in phytoplankton (C:Chl, mg C mg Chl-1), mixed-layer/sea surface nitrate (SSNO3; micromolar), and the number of samples (n; where TS is used to indicate a long time series). The general philosophy was to group similar conditions together so as to minimize data uncertainty and maximize the natural variability captured. In order to maximize the global coverage, we include data from estimates of new (NO3-based) production, nutrient, oxygen or carbon based estimates of export production, and particle export estimates based on sinking flux from sediment traps and/or 234Th. If we assume steady state, we can use these observations to constrain particle export, because in the absence of net accumulation of organic nitrogen in the euphotic zone, the amount of new (NO3-based and N2-fixation based) production is balanced by the sum of sinking particle export, advection of (primarily dissolved) organic nitrogen and nitrification. We assume that nitrification is insignificant within the euphotic zone due to photo-inhibition (Ward et al., 1982). We also assume that N2-fixation is a small contributor to the upper ocean nitrogen budget for the sites that we consider here. We adjust the estimates of new production and export production downward on a site-to-site basis to account for export of carbon through dissolved organic matter (DOM) transport and thereby convert all estimates into particle export. Estimates of the fraction of new production going to semi-labile DOM range from 11% for polar regions (Carlson et al., 2000), to 52% for the subtropical north Pacific (Emerson et al., 1997) with intermediate values of 33% for the subtropical North Atlantic (Carlson et al., 1994) and 19% for the tropical Pacific (Hansell et al., 1997; Dunne et al., 2000). We assume a multiplication factor of 0.813 to account for DOM transport in estimates of new production unless otherwise noted. This correction ended up having only a small impact on results, partially as most of the oligotrophic sites (where the correction is large) were based on direct sediment trap measurements rather than inferred from new production. In order to maintain internal consistency, estimates of particle export from 15N-based NO3 production were converted to carbon units by multiplying the f-ratio (NO3 production divided by total n production) by the 14C-based primary production. Primary production was from 14C-based incubations unless otherwise noted. Where necessary, a carbon to nitrogen ratio of 7.3125 was assumed after Anderson and Sarmiento (1994). Supplemental Material 2: The second supplemental dataset describes the subset of sites given in Supplemental Material 1 for which observations of ecosystem size structure were available. Variables include mixed-layer temperature (SST; degrees C), depth integrated chlorophyll (Chl; mg m-2), depth integrated productivity (Prod; mmol C m-2 d-1), euphotic zone depth (Zeu; m; 1% light level or sampling zone, if less), the carbon to chlorophyll ratio in phytoplankton (C:Chl; mg C mg Chl-1), and the fraction of large (nominally > 5 micrometers) phytoplankton of total phytoplankton (fracL). The data source and reference is also provided for each estimate. The following is a detailed description of each study site: Tropical Panama Basin - This data set was taken as a single average of four sediment trap and productivity deployments in the eastern equatorial Pacific. Primary production was taken from Murray et al. (1989). Particle export was taken as the average of sediment trap fluxes and the 15N-based f-ratio multiplied by the 14C-based primary production (Murray et al., 1989). C:Chl was taken from Eppley et al. (1992). WEC88 - This data set from the central equatorial Pacific during boreal spring 1988, El Nino conditions was divided into five groups, corresponding to north oligotrophic, north equatorial, equatorial, south equatorial and south oligotrophic conditions. Primary production was taken from Barber et al. (1992). Particle export was taken as the 15N-based f-ratio multiplied by the 14C-based primary production Wilkerson et al (1992). C:Chl was taken from Eppley et al. (1992). NWIO - Open Sea - This data set taken during the boreal fall of 1986 in the Northwest Indian Ocean that were all away from the coast and had consistently low values of primary production and new production. Primary production was taken from Owens et al. (1992). Particle export was taken as the 15N-based f-ratio (Owens et al., 1992) multiplied by the 14C-based primary production. C:Chl was taken from Jochem et al. (1992). EQPAC surveys - This data set from the U.S. Joint Global Ocean Flux Study (JGOFS) process study in the central equatorial Pacific during boreal spring 1992, El Nino conditions (Survey I) and during boreal fall 1992, normal upwelling conditions (Survey II) was divided into three groups, corresponding to north equatorial, equatorial and south equatorial conditions. Primary production was taken from Barber et al. (1996). Particle export was taken as the average of 234Th-calibrated sediment trap fluxes and the 15N-based f-ratio multiplied by the 14C-based primary production multiplied (Murray et al., 1996). C:Chl was taken from Eppley et al. (1992). EQPAC time-series - This data set from the U.S. JGOFS process study in the central equatorial Pacific during boreal spring 1992, El Nino conditions (TS-I) and during boreal fall 1992, normal upwelling conditions (TS-II) was divided into three groups, corresponding to north equatorial, equatorial and south equatorial conditions. Primary production was taken from Barber et al. (1996). Particle export was taken as the average of 234Th fluxes (multiplied by sediment trap C:234Th ratios from EQPAC Survey I) and the 15N-based f-ratio multiplied by the 14C-based primary production (Dunne et al., 2000). C:Chl was taken from Eppley et al. (1992). IronEx-II - This data set from an iron fertilization experiment in the eastern equatorial pacific was divided into three phases, a control, a mean over the onset of the bloom, and a survey during the peak of the bloom. Primary production for the control was taken from the average phytoplankton concentration multiplied by the average growth rate in the upper 36 m while particle export was taken from 15NO3-based production. Primary production for the mean bloom was taken from the average phytoplankton concentration multiplied by the average growth rate in the upper 36 m while particle export was taken from the mean, decrease in dissolved inorganic carbon, decrease in NO3 and increase in O2 over six days. Primary production for the peak was taken from 14C-based incubations while particle export was taken from 15NO3-based production (Coale et al., 1996). C:Chl was taken from Landry et al. (2000). OLIPAC - This data set from the French JGOFS process study in the central equatorial Pacific during boreal fall 1994, El Nino conditions was divided into three groups, corresponding to equatorial, south equatorial and south oligotrophic conditions. Primary production was taken from Raimbault et al. (1999). Particle export was taken as the 15N-based f-ratio multiplied by the 14C-based primary production (Raimbault et al., 1999). C:Chl was taken from Eppley et al. (1992). Arabesque - Open Sea - This data set from the U.K. JGOFS process study in the boreal fall (intermonsoon) of 1994 in the Northwest Indian Ocean includes all stations that were all away from the coast. Primary production was taken as the sum of 15N-based NO3, NH4 and urea incubations while particle export was taken as 15N-based NO3 production (Watts et al., 1999). C:Chl was estimated from Barlow et al. (1999) chlorophyll and Tarran et al. (1999) carbon. FLUPAC - This data set includes time-series sites from the French JGOFS process study in the eastern (165E) and central (150W) equatorial Pacific during boreal fall 1994, El Nino conditions. Primary production was taken from Dunne et al. (2000). Particle export was taken as the average of 234Th-calibrated sediment trap fluxes and the 15N-based f-ratio multiplied by the 14C-based primary production (Dunne et al., 2000). C:Chl was taken from Eppley et al. (1992). Zonal Flux - This data set includes sites along the equator from the eastern (165E) to central (150W) Pacific during boreal spring 1996, La Nina conditions. Primary production was taken from Dunne et al. (2000). Particle export was taken as the average of 234Th-calibrated sediment trap fluxes and the 15N-based f-ratio multiplied by the 14C-based primary production multiplied by 0.813 to account for dissolved organic matter (Dunne et al., 2000). C:Chl was taken from Eppley et al. (1992). Arabian Sea - These data sets are taken from five survey cruises of the U.S. JGOFS process study in the Arabian Sea under various monsoonal conditions. Primary production was taken from Barber et al. (2001). Particle export was taken as the average of 234Th fluxes (multiplied by large volume filtration C:234Th ratios; Buesseler et al., 1998) and the 15N-based f-ratio (McCarthy et al., 1999) multiplied by the 14C-based primary production. C:Chl was taken from Garrison et al. (2000). Subtropical VERTEX - This data set includes three sites in the Northeast Pacific. Primary production was taken from Knauer et al. (1984). Particle export was taken from sediment traps (Knauer et al. (1984). C:Chl was taken from Welschmeyer and Lorenzen (1985). BATS - Synth - This is a synthesis of U.S. JGOFS Bermuda Atlantic Time-Series data by Carlson et al (1994) which compares six different geochemical methods of estimating export production, DOC export and particle export with the annual averages for particle export for three separate years giving a total of nine separate estimates. The reason for including all estimates separately is that each spans a different time and spatial scale and so are somewhat independent. C:Chl was taken from Fasham et al. (1990). BATS - Upon investigation of the full U.S. JGOFS Bermuda Atlantic Time-series Site (BATS) data set from 1989 through 1998, it was found that seasonal (3 month) averages over all years captured 41% more of the variance in the sediment trap-based particle export data than annual averages, demonstrating more intra-annual variability than inter-annual variability. Primary production and sediment trap-based particle export data was thus grouped into four seasonal averages to capture the most variability without weighting the entire data synthesis to fit only the BATS study site (from a total of 108 data values). All data were taken from the BATS data access portal (http://www.bbsr.edu/cintoo/bats/bats.html). Sediment trap fluxes were each multiplied by a factor of 1.078 to account for zooplankton migration (Steinberg et al., 2000). C:Chl was taken from Fasham et al. (1990). HOT - Upon investigation of the full U.S. JGOFS Hawaii Ocean Time-series (HOT) data set from 1989 through 1999, it was found that annual averages over all years captured 46% more of the variance in the sediment trap-based particle export data than seasonal (3 month) averages, demonstrating more inter-annual variability than intra-annual variability. Primary production and sediment trap-based particle export data was thus grouped into 11 annual averages to capture the most variability without weighting the entire data synthesis to fit only the HOT study site (from a total of 94 data values). All data were taken from the HOT data access portal (http://hahana.soest.hawaii.edu/hot/hot-dogs/interface.html). For years that geochemical estimates based on O2, N2 and Ar and DIC/DI13C (1990, 1992, 1994 and 1995), an average of all three estimates was used, multiplied by 0.476 to account for dissolved organic matter (Emerson et al., 1997). Sediment trap fluxes were each increased by 0.55 mmol C m-2 d-1 to account for zooplankton migration (Emerson et al., 1997). C:Chl was taken from Welschmeyer and Lorenzen (1985). N. Atl - This data set from three survey cruises boreal fall of 1992 and spring of 1993 includes three separate regions of the North Atlantic, the Gulf Stream (GFST), the western North Atlantic subtropical gyre (NASW) and the eastern North Atlantic subtropical gyre (NASE). Primary production was taken from 13C incubations while particle export was taken from 15N incubations (Harrison et al., 2001). C:Chl was taken from Li and Harrison (2001). Subpolar-polar Weddell Sea - This data set from two surveys in the austral spring of 1983 and fall of 1986 by Smith and Nelson (1990). Primary production was taken as the sum of 15N-based NO3, NH4 and NO2 production multiplied by 1.12 to account for urea production (after the work of Cochlan and Bronk, 2001) while particle export was taken as 15N-based NO3 production multiplied by 0.91 to account for dissolved organic matter (rather than 0.813). C:Chl was taken from Nelson et al. (1989). NABE - These two independent data sets are from the U.S. JGOFS North Atlantic Bloom Experiment during the boreal spring of 1989. Two independent syntheses of primary production and particle export were made during this study, one by Buesseler et al. (1992) combining 15N-based NO3 production and 234Th-calibrated sediment traps from three studies over 6 weeks and another by Bender et al (1992), who formulated a mass balance over the first 11 days of the study using 18O-based export production, sediment traps and temporal variability in the organic carbon inventory. C:Chl was taken from Lochte et al. (1993). Super - This data set from the Subpolar northeast Pacific was divided into boreal spring of two years (1984, 1988) and summer of two years (1987, 1988). Primary production and C:Chl were taken from (Welschmeyer et al., 1993). Particle export was taken from 15N-based NO3 production (Wheeler et al., 1993). Station P - This data set was taken from two stations in the subpolar Northeast Pacific during August of 1988. Primary production was taken as the sum of 15N-based NO3, NH4 and urea production while particle export was taken as the average of 15N-based NO3 production and an O2, N2 and Ar mass balance (Emerson et al., 1993). C:Chl was taken from Welschmeyer et al.(1993). Bell. Sea - Bury - This data set from three sites in the Bellinghausen Sea was taken in the austral spring of 1992 by Bury et al. (1995). Primary production was taken as 13C-based production while particle export was taken as 15N-based NO3 production multiplied by 0.91 to account for dissolved organic matter. C:Chl was taken from Robins et al. (1995). Bell. Sea - Regions II, III, IV - This data set from three time-series sites in the Bellinghausen Sea was taken in the austral spring of 1992 by Waldron et al. (1995). Primary production was taken as the sum of 15N-based NO3, NH4 and urea production while particle export was taken as 15N-based NO3 production multiplied by 0.91 to account for dissolved organic matter. C:Chl was taken from Robins et al. (1995). NEW Polyna - This data set from the Northeast Water Polynya study (off the northeast coast of Greenland) in July and August of 1992 and 1993 and was divided into Northerly ice-covered and ice free, central and southern regions after Yager et al. (1995). Primary production was taken as the sum of 15N-based NO3 and NH4 production (Smith, 1995; except for the central region where primary production was taken from 234Th fluxes multiplied by large volume filtration C:234Th ratios) while particle export was taken as the average of 234Th fluxes multiplied by large volume filtration C:234Th ratios (Cochlan et al., 1995) and 15N-based NO3 production (Smith, 1995) multiplied by 0.91 to account for dissolved organic matter. C:Chl was taken from Smith et al. (1995). Arctic Sea - This data set was taken from a geographically broad survey of the Arctic in the boreal summer of 1994. Primary production was taken from Gosselin et al. (1997). Particle export was taken from 234Th fluxes multiplied by large volume filtration C:234Th ratios (Moran et al. 1997). C:Chl was taken from Booth and Horner (1997) Line P - This data set, taken from the Canadian JGOFS process study in the subpolar Northeast Pacific, includes surveys in the boreal winter (1993, 1994), spring (1993, 1994) and summer (1992, 1994). Primary production was taken as the sum of 15N-based NO3, NH4 and urea production (Varela and Harrison, 1999) while particle export was taken as the average of 234Th fluxes (multiplied by large volume filtration C:234Th ratios; Charette et al., 1999) and 15N-based NO3 production (Varela and Harrison, 1999). C:Chl was taken from Welschmeyer et al. (1993). Ross Sea - These four data sets from the U.S. JGOFS process study in the Ross Sea during the austral spring of 1996 through spring of 1997 were taken as independent annual averages. Temporal averages were chosen in the hopes of most closely approximating steady state. They include primary production from 14C-based incubations (Smith et al., 2000) with particle export from 234Th fluxes multiplied by large volume filtration C:234Th ratios (Cochran et al., 2000), primary production from 18O gross production incubations multiplied by 0.5 to account for phytoplankton respiration with particle export from 18O net production incubations multiplied by 0.91 to account for dissolved organic matter (Bender et al., 2000), primary production from 15N-based NO3, NH4, NO2 and urea incubations with particle export from 15N-based NO3 incubations (Cochlan and Bronk, 2001) multiplied by 0.91 to account for dissolved organic matter, and primary production from 14C-based incubations (Smith et al., 2000) with particle export from a nutrient budget over the growing season (Sweeney et al., 2000). C:Chl was taken from Olson et al.(2000). ASPZ - This data set from the U.S. JGOFS process study in the Antarctic and Subantarctic Polar Zones from austral spring of 1997 to austral fall of 1998 was divided into four latitude bands corresponding to different biogeographical regions. Temporal averages were chosen in the hopes of most closely approximating steady state. Primary production was taken as the average of 14C-based incubations (Hiscock et al., 2003), 15N-based NO3, NH4, NO2 and urea incubations (Sambrotto and Mace, 2000) and 18O gross production incubations multiplied by 0.5 to account for phytoplankton respiration (Bender et al., 2000) while particle export was taken as the average of 15N-based NO3 incubations (Sambrotto and Mace, 2000) multiplied by 0.91 to account for dissolved organic matter, 18O gross production incubations (Bender et al., 2000) also multiplied by 0.91 to account for dissolved organic matter and 234Th fluxes multiplied by large volume filtration C:234Th ratios (Buesseler et al. 2003). C:Chl was taken from Robins et al. (1995). SOIREE - This data set from an iron fertilization experiment in the southern ocean by Boyd et al. (2000) was divided into two phases, a mean over the onset of the bloom, and a survey during the peak of the bloom. Particle export during the mean bloom was taken as the change in the organic carbon budget over 9 days multiplied by 0.91 to account for dissolved organic matter. Particle export during the peak was taken as the dissolved inorganic carbon decrease normalized to the change in the NO3 inventory from days 3-9 to days 10-12 multiplied by 0.91 to account for dissolved organic matter. C:Chl was taken from Robins et al. (1995). PRIME - This data set from the Northeast Atlantic during June of 1996 was divided into two groups: observations within an eddy (Rees et al., 2001), and observations along a meridional transect (Donald et al., 2001). Particle export was taken as from 15N-based NO3 incubations. C:Chl was taken from Lochte et al. (1993). Coastal Dabob Bay - This data set by Welschmeyer and Lorenzen (1985) from a time-series site in Dabob Bay, WA from Fall of 1978 to summer of 1981 utilized filters with a rather large pore size, and came before the advent of clean techniques. As a result, primary production by picoplankton was not sampled. To account for this, only data from spring and summer, periods of active diatom blooming, was included. Particle export was taken from sediment traps. Peru - This data set by Wilkerson et al. (1987) from two survey cruises in the boreal spring of 1976 during El Nino conditions and 1977 during normal upwelling conditions. Primary production was taken as the sum of 15N-based NO3 and NH4 incubations while particle export was taken as 15N-based NO3 production. C:Chl was taken from Lorenzen (1968). Santa Monica - This data set by Small et al. (1989) is a synthesis from a process study during 1985 to 1987 in the Santa Monica Basin along the southwestern U.S. coast and was divided into boreal spring and fall periods. Particle export was estimated from the average of NO3 changes during incubations and sediment traps. C:Chl was taken from Jackson and Eldridge (1992). NWIO - coast - as in NWIO - Open sea except for coastal stations and the Gulf of Oman. Agulhas - This data set by Probyn et al (1995) taken in the Agulhas current off the tip of South Africa during January of 1992 was divided into upwelling (surface temperature less than 24?C) and non-upwelling (surface temperature equal to 25?C) stations. Primary production was taken as the sum of 15N-based NO3, NH4 and urea production while particle export was taken from 15N-based NO3 production. C:Chl was taken from Jackson and Eldridge (1992). Scotland - This data set by Rees et al. (1995) from a time-series study of four sites in Loch Linnhe, Scotland was divided into three groups: spring, summer and fall/winter. Primary production was taken as the sum of 15N-based NO3 and NH4 and production while particle export was taken as 15N-based NO3 production multiplied by 0.91 to account for dissolved organic matter (rather than 0.813). C:Chl was taken from Jackson and Eldridge (1992). Arabesque - as in Arabesque - Open sea except for coastal stations and the Gulf of Oman. Santa Barbara Basin - This data set by Dunne (1999) from a time-series station in the Santa Barbara Basin along the southwestern U.S. coast during April of 1997. Primary production was taken from 13C incubations while particle export was taken as the average of 234Th-calibrated sediment trap fluxes and the 15N-based f-ratio multiplied by the 14C-based primary production. C:Chl was taken from Jackson and Eldridge (1992). Gulf of Riga - This synthesis by Donali et al. (1999) from the Baltic Sea utilizes data collected in the spring of 1995, summer of 1994 and fall of 1993. Primary production was taken from gross oxygen production incubations while particle export was taken as the sum of sediment trap fluxes and the time-rate-of-change of phytoplankton biomass. C:Chl was estimated from Olli and Heiskanen (1999). Monterey Bay - This data set by Wilkerson et al. (2000) from the central California, U.S. coast was taken during spring and fall of 1992 and 1993. Particle export was taken as 15N-based NO3 production. Note that these estimates are from surface samples only. C:Chl was taken from Jackson and Eldridge (1992). Gulf of St. Lawrence - This data set by Vezina et al. (2000) from the Canadian JGOFS process study in the Gulf of St. Lawrence from the boreal summer of 1992 to the boreal summer of 1994 was divided into summer and winter groups. Particle export was taken from a synthesis of sediment trap fluxes and the 15N-based f-ratio multiplied by the 14C-based primary production. C:Chl was taken from Jackson and Eldridge (1992). Chesapeake Bay - This data set by Smith and Kemp (2001) from surveys during 1996 and 1997 within Chesapeake Bay on the mid-Atlantic U.S. coast was divided into spring, summer and fall conditions. Primary production was taken from gross oxygen production incubations while particle export was taken as net oxygen production incubations. C:Chl was taken from Welschmeyer and Lorenzen (1985). E. China Sea - This data set by Chen et al. (2001) from two surveys in the East China Sea during 1997 was divided both into near-shore (coast) versus off-shore (open) sites and summer versus winter sites. Primary production was taken from 13C incubations while particle export was taken as 15N-based NO3 production. C:Chl was taken from Jackson and Eldridge (1992).