Seasonal and interannual variability of North American isoprene emissions as determined by formaldehyde column measurements from space

[2] Isoprene (CH₂=CH−C(CH₃)=CH₂), a biogenic hydrocarbon whose main source is terrestrial vegetation [Guenther et al., 1995], plays an important role in the production of tropospheric ozone. Isoprene emission depends largely on vegetation type, vegetation density (leaf area index), leaf age, light, and temperature [Guenther et al., 2000; Geron et al., 2000]. Current inventories of isoprene emission use functional dependences of these parameters that are extrapolated from limited isoprene flux measurements. Large uncertainties are associated with this approach [Pierce et al., 1998]. Total global emission estimates of isoprene used in atmospheric chemistry models vary by at least a factor of 2 [Guenther et al., 1999] while the variation in regional emission estimates can be much larger.

[3] Satellite observations of formaldehyde (HCHO) columns from solar UV backscatter offer top-down constraints to reduce uncertainty in isoprene emission inventories [Palmer et al., 2003]. The HCHO column enhancement over that contributed by the methane background is due to oxidation of reactive organic gases. Over North America during the growing season, isoprene is the main regional contributor to this HCHO enhancement [Lee et al., 1998; Sumner et al., 2001]. The lifetime of isoprene against atmospheric oxidation is typically less than one hour, and the lifetime of HCHO is a few hours, so that isoprene emission is collocated with and linearly related to the resulting HCHO enhancement over a smearing length scale of order 10–100 km [Palmer et al., 2003]. Biogenic terpenes are thought to contribute little to HCHO columns because their HCHO yield per unit carbon is low [Orlando et al., 2000]. We showed previously that HCHO column observations from the Global Ozone Monitoring Experiment (GOME) satellite instrument [Burrows et al., 1999] over North America during July 1996 are consistent with in situ HCHO data and with the expected distribution of isoprene emission [Palmer et al., 2003]. Biomass burning is also a major source of enhanced HCHO observed by GOME [Thomas et al., 1998] but this source is not important over North America except for occasional wildfires. In this paper we further demonstrate the utility of HCHO column observations from GOME for mapping isoprene emission by examining the seasonal and interannual variability of these data over North America and comparing it to current understanding of isoprene emission patterns.

[4] The GOME instrument measures backscattered solar radiation (237–794 nm) in the nadir with a 320 × 40 km² field of view. Global coverage is achieved in 3 days. “Slant” column HCHO along the line of sight is calculated from GOME spectra in the 336–356 nm wavelength region with a mean fitting uncertainty of 4 × 10¹⁵ molec cm⁻² [Chance et al., 2000]. Air mass factors (AMFs), used to convert slant to vertical columns, are calculated with the LIDORT radiative transfer model [Spurr, 2002] using local relative vertical profiles of HCHO concentration (shape factors) from the GEOS-CHEM chemical transport model, and accounting for aerosol and cloud effects as well as Rayleigh scattering [Palmer et al., 2001; Martin et al., 2002; Martin et al., 2003a]. Using GOME cloud information [Kurosu et al., 1999] we exclude observation scenes with cloud or snow cover fraction >40%; these represent about 40% of the total data over North America during the summer.

[5] The GOME diffuser plate, used to attenuate solar radiation, introduces a global uniform bias in the retrieved slant column that varies with solar day and has no apparent seasonal or interannual pattern [Richter and Wagner, 2001; Martin et al., 2002]. There is also a zonally symmetric and seasonally varying pattern in the slant columns that we do not understand which results in slant columns over remote oceans that exceed the expected background due to methane oxidation by values approaching 1.0 × 10¹⁶ molec cm⁻² at high latitudes in summer. We quantify these biases as the daily difference between GEOS-CHEM and GOME background HCHO columns over the central North Pacific as a function of latitude [Martin et al., 2002; Palmer et al., 2003], and subtract them from the results presented here.

[6] Figure 1 shows the monthly-mean GOME HCHO columns over North America during March–October 1997. Daily cloud data from GOME and HCHO shape factors for each observation scene were used in the AMF calculation to convert slant to vertical columns. We find a strong seasonal variation in HCHO columns correlated with the growing season and with high summertime temperatures. The GOME HCHO columns are highest in the southeastern United States, consistent with the known distribution of isoprene emissions [Guenther et al., 2000]. Also shown in Figure 1 are monthly mean HCHO columns from the GEOS-CHEM chemical transport model [Bey et al., 2001], which uses isoprene emissions from the Global Emission Inventory Activity (GEIA) [Guenther et al., 1995]. We use here GEOS-CHEM version 4.26 with additional model developments described in Martin et al. [2003b]. The model is broadly consistent with the observations in terms of both spatial and seasonal variations. It is biased high by 20% on average. Improvements in isoprene emission modeling have been made since the GEIA inventory, notably the accounting for the effect of leaf age and antecedent temperatures [Guenther et al., 1999; Geron et al., 2000]. These improvements will be incorporated into the model in future work.

[7] Outside of the growing season, observed and modeled HCHO columns are generally below the detection limit of 4 × 10¹⁵ molec cm⁻². The low winter values in the model reflect the relatively low emissions and long lifetimes of anthropogenic hydrocarbons. Large values over southern Mexico during April–May 1997 in the GOME data are due to agricultural fires; these fires are accounted for in the GEOS-CHEM simulation on the basis of satellite observations of firecounts [Duncan et al., 2003], but with apparently insufficient hydrocarbon emissions. Outside of this fire period the GOME HCHO columns over southern Mexico are low, consistent with the GEIA isoprene emission inventory.

[8] Figure 2 shows the interannual variability of observed monthly mean HCHO columns for August 1995–2001, together with the corresponding surface air temperature anomalies. Monthly mean AMFs from August 1997 were used to convert slant to vertical columns in other years. Comparison of monthly mean local AMFs over North America for September 1996 and 1997 suggests that interannual variation is generally within 10% so that the interannual variability in the GOME data is driven mainly by the slant columns fitted from the observed radiances. Cloud data specific to each GOME observation scene [Kurosu et al., 1999] was used to exclude scenes with cloud or snow fraction >40%. The general spatial pattern of HCHO columns from GOME in Figure 2 is reproduced from year to year. There is a 30% interannual variability over the eastern United States that appears to follow roughly variations in surface air temperature. The HCHO columns during August 1997 are about 15% lower than average over the southeastern United States, and the corresponding negative temperature anomaly is about 1 K, consistent with the exp[(0.12 K⁻¹)T] temperature dependence of isoprene emission in the temperature range of interest [Guenther et al., 1995]. The hottest August over the southeastern United States in the GOME record was in 2000, and HCHO columns for that month exceed 3.0 × 10¹⁶ molec cm⁻². The HCHO columns in August 2000 also show high values over the northwestern United States due to large wildfires.

[9] Not all of the interannual variability of GOME HCHO columns in Figure 2 can be explained by temperature (e.g., August 1996). Additional factors that might contribute to interannual variability of isoprene emissions include fires, water availability, and tree disease. Preliminary examination of satellite observations of firecount and drought indices did not reveal an obvious role for these factors but further investigation is needed.

[10] We have shown in this paper that satellite observations of the HCHO column from GOME are broadly consistent with our understanding of the seasonal and interannual variability of isoprene emission. This demonstrates the considerable potential for constraining isoprene emission inventories using satellite observations and improving our understanding of the factors controlling these emissions. Extending our analysis to the rest of the world is clearly desirable but will require careful separation of biomass burning and biogenic contributions to HCHO. Uncertainty in the HCHO yields from isoprene oxidation under low-NO_x conditions [Palmer et al., 2003], as frequently encountered in the tropics, is a key limitation that will need to be addressed through laboratory studies. Better understanding is also needed of the HCHO yields from oxidation of terpenes, sesquiterpenes, and other highly reactive organic compounds emitted by vegetation.