Biospheric Primary Production During an ENSO Transition

provides open access, scholarly publishing services to the University of California and delivers a dynamic research platform to scholars worldwide. eScholarship provides open access, scholarly publishing services to the University of California and delivers a dynamic research platform to scholars worldwide. Abstract: The Sea-viewing Wide Field-of-view Sensor (SeaWiFS) provides global monthly measurements of both oceanic phytoplankton chlorophyll biomass and light harvesting by land plants. These measurements allowed the comparison of simultaneous ocean and land net primary production (NPP) responses to a major El Niño to La Niña transition. Between September 1997 and August 2000, biospheric NPP varied by 6 petagrams of carbon per year (from 111 to 117 petagrams of carbon per year). Increases in ocean NPP were pronounced in tropical regions where El Niño-Southern Oscillation (ENSO) impacts on upwelling and nutrient availability were greatest. Globally, land NPP did not exhibit a clear ENSO response, although regional changes were substantial. The Sea-viewing Wide Field-of-view Sensor (SeaWiFS) provides global monthly measurements of both oceanic phytoplankton chlorophyll biomass and light harvesting by land plants. These measurements allowed the comparison of simultaneous ocean and land net primary production (NPP) responses to a major El Nin(cid:247)o to La Nin(cid:247)a transition. Between September 1997 and August 2000, biospheric NPP varied by 6 petagrams of carbon per year (from 111 to 117 petagrams of carbon per year). Increases in ocean NPP were pronounced in tropical regions where El Nin(cid:247)o—Southern Oscillation (ENSO) impacts on upwelling and nutrient availability were greatest. Globally, land NPP did not exhibit a clear ENSO response, although regional changes were substantial. Temporal changes in the physical environ-ment are manifested in the light-harvesting capacity of plant communities throughout the biosphere and can be monitored remotely by changes in surface chlorophyll concentration ( C sat ) in the oceans and the Normalized Difference Vegetation Index (NDVI) on land. A continuous, 20-year global record of satellite NDVI has permitted characterization of interannual, climate-driven changes in terrestrial photosynthesis ( 1–5 ). Coincident changes in ocean productivity have not been assessed because an analogous long-term global C sat record does not exist. The first C sat measure-R

where ⌬G act NAT ϭ (30.9 Ϯ 0.3) Ϫ (0.14 Ϯ 0.0004)S NAT where J is the freezing rate of a single particle in sec Ϫ1 , T is temperature in K, r is the particle radius in cm, ⌬G act is the activation free energy for the formation of hydrate germs in solution in units of kcal mol Ϫ1 (20), S is hydrate saturation in solution, and R is the universal ideal gas constant in units of kcal mol Ϫ1 K Ϫ1 . To account for the effect of particle size on freezing, we have slightly modified (35) the preexponential factor from that given in Salcedo et al. (20). 22. The 1D model uses vertical temperature profiles that are constructed using the archived NASA Goddard National Centers for Environmental Prediction (NCEP) data (36). To obtain a vertical temperature profile, a single position is first selected on the 450 K surface (ϳ50 mbar) inside a cloudy region (T Ͻ 192 K). Next, air parcel trajectories (36) are run forward and backward for 2 weeks from the chosen location to obtain a 4-week-long temperature and position history of an air parcel. Positions along the 4-weeklong trajectory at the 450 K level are used to read off vertical temperature profiles at each location and time along the parcel's path using the NCEP data. 23 Heymsfield, J. Geophys. Res. 99, 10443 (1994). 24. Cloud model sensitivity studies (22) indicate that hourly NAD and/or NAT hydrate particle production rates Ͻϳ10 Ϫ5 cm Ϫ3 (Fig. 1) can result in 10% denitrification at most in typical air parcels before particle evaporation occurs. Thus, NAD particle production within the nucleation window is the dominant process that controls the overall atmospheric number density of the nitric acid hydrated particles. Direct NAT particle production rate from STS is Ͻ10 Ϫ5 cm Ϫ3 in the atmosphere (Fig. 1 (21). Sensitivity studies were also performed by using upper and lower limits for ⌬G in the J expressions for both hydrates. 28. To account for NAD to NAT conversion in the model, the assumptions were used for J values as follows: J NAT ϭ standard J NAD ϩ standard J NAT and J NAD ϭ 0 (27). Most laboratory observations indicate that the metastable NAD phase is the nucleus that initially forms in aerosol particles (13,19,20,37). In general, kinetically favorable phases usually nucleate first in particles, and in time such metastable phases often but not always transition into the most stable form (37). Thus, the possibility of NAD to NAT conversion in the atmosphere is likely, and recent in situ NAT particle observations (38) seem to support this hypothesis. 29. Due to cold temperatures roughly 1 to 4 ppbv of HNO 3 between 17 to 23 km is sequestered in the STS phase at the end of the Antarctic calculations in most air parcels. The STS content is sensitive to both J uncertainties (27) and the assumption of NAD to NAT conversion (28).
The Sea-viewing Wide Field-of-view Sensor (SeaWiFS) provides global monthly measurements of both oceanic phytoplankton chlorophyll biomass and light harvesting by land plants. These measurements allowed the comparison of simultaneous ocean and land net primary production (NPP) responses to a major El Niño to La Niña transition. Between September 1997 and August 2000, biospheric NPP varied by 6 petagrams of carbon per year (from 111 to 117 petagrams of carbon per year). Increases in ocean NPP were pronounced in tropical regions where El Niño-Southern Oscillation (ENSO) impacts on upwelling and nutrient availability were greatest. Globally, land NPP did not exhibit a clear ENSO response, although regional changes were substantial.
Temporal changes in the physical environment are manifested in the light-harvesting capacity of plant communities throughout the biosphere and can be monitored remotely by changes in surface chlorophyll concentration (C sat ) in the oceans and the Normalized Difference Vegetation Index (NDVI) on land. A continuous, 20-year global record of satellite NDVI has permitted characterization of interannual, climate-driven changes in terrestrial photosynthesis (1)(2)(3)(4)(5). Coincident changes in ocean productivity have not been assessed because an analogous long-term global C sat record does not exist. The first C sat measure- ments were made with the Coastal Zone Color Scanner (CZCS: 1978 -86), but this proofof-concept sensor collected data on a highly irregular basis that yielded incomplete global coverage even after integration over the entire 8-year mission. Eleven years later, Sea-WiFS was launched, marking the beginning of the first multiyear satellite measurements of phytoplankton biomass since CZCS. Sea-WiFS now provides greater global coverage of C sat each month than was achieved throughout the lifetime of CZCS. In addition, SeaWiFS is the first satellite instrument with the spectral coverage and dynamic range necessary to derive both C sat and NDVI. Here we report spatial and temporal changes in the photosynthetic biosphere for an El Niño to La Niña transition period, as recorded during the first 3 years of the SeaWiFS mission.
We analyzed global, 4-km resolution, monthly SeaWiFS C sat and NDVI data collected between September 1997 and August 2000. Stability of the sensor was characterized from monthly lunar-based calibrations and derived products verified by comparison with field measurements (6 -8). Biospheric net primary production (NPP) was estimated following the approach of Field et al. (9), which integrates the Vertically Generalized Production Model (VGPM) for the oceans (10) with the Carnegie-Ames-Stanford Approach (CASA) for land (11,12). Variations in NPP for the CASA-VGPM model arise from changes in three factors: (i) incident photosynthetically active radiation (PAR), (ii) the fraction of radiation absorbed by plants (related to C sat and NDVI), and (iii) light use efficiency (ε). Coincident changes in these factors collectively control NPP. Unlike previous calculations that used C sat , NDVI, and climate data from different periods (9), all data used in the current NPP estimates were collected during the SeaWiFS period (13). The CASA-VGPM model was operated on a monthly time step.
SeaWiFS measurements began near the peak of the 1997-98 El Niño event (by some measures, one of the strongest on record) (14) and then continued through an equally strong La Niña period. A pronounced seasonal cycle dominated temporal variability in global mean C sat throughout the SeaWiFS record (Fig. 1A). Summer phytoplankton blooms in the Northern Hemisphere exceeded those in the Southern Hemisphere, causing minima in global mean C sat between November and March and maxima between May and September. Superimposed on this prominent seasonal cycle was a clear El Niño-Southern Oscillation (ENSO)-related change in ocean productivity, as illustrated by the monthly C sat anomaly record (Fig. 1A) (15). The El Niño to La Niña transition altered ocean nutrient distributions, causing nearly a 10% increase in global mean C sat between September 1997 and December 1998. Changes in C sat during this period were not restricted to the equatorial belt but rather were global in extent. During the subsequent La Niña period of January 1999 to August 2000, C sat continued to increase at the reduced rate of 2.2% per year, primarily reflecting increased phytoplankton biomass in the Pacific Ocean.
Temporal changes in ocean NPP exhibited seasonal and interannual patterns similar to C sat , increasing from 54 to 59 Pg C year Ϫ1 (Pg ϭ 10 15 g) over the 3-year SeaWiFS period. Regionally, NPP was highest near equatorial and eastern margin upwelling centers, at high latitudes in the Northern Hemisphere, and within the southern subtropical convergence zone (Fig. 2, A and B). Seasonal changes in Southern Hemisphere NPP mirrored those of the Northern Hemisphere, except between 40°and 75°S latitude from October to April (Fig. 3). At Ͼ40°N, phytoplankton growth is restricted by deep mixing and low PAR during winter months and then increases markedly in the summer when surface waters rich in nutrients become stratified and PAR is high. Consequently, NPP was strongly seasonal in this region, varying from 0 to 49 g C m Ϫ2 month Ϫ1 (Fig. 3). In contrast, seasonality in NPP was greatly dampened poleward of 40°S, with summer values decreasing from 27 to 7 g C m Ϫ2 month Ϫ1 between 40°and 70°S (Fig. 3). This absence of a high-latitude, Southern Hemisphere bloom results from weak seasonality in factors limiting phytoplankton growth, particularly iron and vertical mixing (16 -19). We calculated that a 9 Pg C year Ϫ1 increase in NPP would result if seasonal changes in phytoplankton biomass between 40°and 75°S paralleled those in the Northern Hemisphere (20).
On land, temporal changes in global mean NDVI were dominated by strong seasonal fluctuations, with minima of 0.44 Ϯ 0.01 (dimensionless) between December and February and maxima of 0.55 Ϯ 0.01 between June and September (Fig. 1B). Land NPP peaked between 15°S and 10°N, reaching 87 g C m Ϫ2 month Ϫ1 , and varied seasonally at Ͼ35°N from 0 to 75 g C m Ϫ2 month Ϫ1 (Fig. 2, A and B). Despite the strong El Niño and La Niña, monthly anomalies indicated little systematic impact on global mean NDVI for the 3-year SeaWiFS record ( Fig.  1B) (15). Land NPP was nearly constant for both climate regimes, ranging from 57 to 58 Pg of C year Ϫ1 between September 1997 and August 2000. Substantial ENSO-related regional changes, however, are hidden in these global integrals.
Biospheric distributions of NPP register   (21) and reduced terrestrial NPP in eastern Africa related to decreased precipitation (Fig. 2C).
Additional features of the transition included (i) a change in Indian Ocean circulation that increased NPP in the northeast while decreasing productivity west of Indonesia (22), (ii) precipitation-related changes in NPP over Amazonia and Argentina, and (iii) nutrient-driven increases in ocean NPP east of Argentina and in the Mauritanian upwelling plume off western Africa (Fig.  2C). Persistent La Niñ a conditions between the Boreal summers of 1998 and 1999 led to spatially heterogeneous changes in NPP, including a large equatorial decrease and off-equatorial increase in the Pacific Ocean that likely reflected broad-scale shoaling of the thermocline (23) (Fig. 2D).
The CASA-VGPM model gave biospheric NPP estimates of 111 to 117 Pg C year Ϫ1 for the September 1997 to August 2000 period (24). Using the same model and remote sensing data collected between 1978 and 1990, Field et al. (9) estimated biospheric NPP at 105 Pg C year Ϫ1 . Their estimate for the land component (56 Pg C year Ϫ1 ) was about the same as that reported here. However, their estimate of ocean NPP (49 Pg C year Ϫ1 ) was considerably lower than our results, largely because of higher C sat values from SeaWiFS (1997-2000) than from CZCS (1978 -86) (9,25).
Since September 1997, SeaWiFS has provided the first multiyear measurements of ocean plant biomass in over a decade, as well as the first single-sensor global observations of the photosynthetic biosphere. SeaWiFS NDVI and C sat data provide a basis for quantifying temporal changes in NPP, which is a critical component of global carbon and nutrient cycles. Land and ocean productivity responds to changes in the physical environ-    (Fig. 1). Our initial analysis of the first 3 years of SeaWiFS data suggests that this sensor will have the capacity to detect longer time scale, lower amplitude responses of the photosynthetic biosphere to climate change. Achieving this goal will require a long-term commitment to intercalibrated global observations and improved ε models (26) and remote sensing algorithms (27). As these developments are realized, the SeaWiFS record will provide a basis against which future estimates of Earth system elemental cycling can be compared. 6. The SeaWiFS sensor is mounted on the OrbView-2 spacecraft (28) and follows a sun-synchronous orbit with a noontime equatorial crossing. The sensor provides complete 4-km global coverage every 2 days. The mission has operationally produced oceanic, atmospheric, and terrestrial data products since September 1997. The complete SeaWiFS data set has been reprocessed three times to incorporate refinements in sensor calibration, atmospheric correction, and bio-optical algorithms. The present study uses data from the most recent, May 2000 reprocessing (29). Monthly lunar calibration images indicate a progressive decrease in sensitivity of Ͻ1% for the six visible wave bands and up to 10% for the nearinfrared wave bands. By applying calibration corrections based on the lunar time series, temporal changes in radiometric sensitivity are corrected to within Ͻ1% (29, 30). 7. SeaWiFS wave bands relevant for deriving C sat values are 443, 490, 510, and 555 nm. The third reprocessing of SeaWiFS data used the OC4 chlorophyll algorithm (29). Comparison of OC4 modeled and 2849 in situ chlorophyll values indicated a correlation coefficient (r 2 ) of 0.892 and root mean squared (RMS) error of 0.222 (29). OC4 is a switching algorithm that calculates C sat from the maximum value of the ratios: 443:555, 490:555, and 510:555. OC4 outperforms earlier single wave band ratio algorithms, such as OC2 (490:555), at the high and/or low end of naturally occurring chlorophyll concentrations (29). Comparison of SeaWiFS OC4-derived C sat values and 103 coincident in situ measurements indicated an average difference between modeled and measured C sat of Ͻ6% (RMS error ϭ 0.301) for chlorophyll (Chl) concentrations between 0.029 and 6.42 mg m Ϫ3 (29). It is noteworthy that, between 0.07 and 5.0 mg Chl m Ϫ3 (i.e., 93% of the global ocean), OC4based C sat values are not substantially different than values derived from a single wave band ratio algorithm. Thus, the global-scale spatial and temporal patterns in C sat and ocean NPP described here are not strongly dependent on the choice of C sat algorithm. 8. For SeaWiFS NDVI data, cloud artifacts were reduced by applying a maximum 5 by 5 pixel spatial filter over evergreen broadleaf forest regions and a median 5 by 5 pixel spatial filter and a Fourier Adjustment over all other land surfaces (31,32). Missing data for needle leaf evergreen regions during winter at Ͼ45°N were estimated from October values. No statistically significant trends (95% confidence level) were introduced by these corrections. SeaWiFS NDVI data were calibrated with