NCEP and GISS solar radiation data sets available for ecosystemmodeling: Description, differences, and impacts on net primaryproductionJeffrey A. HickeNatural Resource Ecology Laboratory, Colorado State University, Fort Collins, Colorado, USAReceived 20 October 2004; revised 21 January 2005; accepted 8 February 2005; published 15 April 2005.[1] Downwelling surface solar r adiation is an important input to ecosystem models, andglobal models require spatially extensive data sets that vary interannually to captureeffects that potentially drive changes in ecosystem function. In this paper, I describeand compare solar radiation data sets from two representative sources, National Centersfor Environmental Prediction (NCEP) reanalyses and Goddard Institute for SpaceStudies (GISS) calculations that included satellite observations of cloud properties.The CASA ecosystem model, which uses solar radiation and satellite-derivedvegetation information, was run with the two solar radiation data sets to explore howdifferences affect estimated net primary production (NPP). GISS solar radiation matchedground-based observations better than NCEP solar radiation. Mean global NCEPsolar radiation exceeded that from GISS by 16%, likely as a result of lower cloudinesswithin the NCEP reanalyses compared to satellite observations. Neither data set resultedin a significant trend over the study period (1984–2000). Locally, relative differenceswere up to 40% in the mean and 10% in the trend of solar radiation and NPP, andvaried in sign across the globe. Because reanalysis solar radiation is only indirectlyconstrained by observations in contrast to the satellite-derived data, it is recommendedthat studies use the GISS solar radiation when possible.Citation: Hicke, J. A. (2005), NCEP and GISS solar radiation data sets available for ecosystem modeling: Description, differences,and impacts on net primary production, Global Biogeochem. Cycles, 19, GB2006, doi:10.1029/2004GB002391.1. Introduction[2] Models, input data sets, and computing power haveadvanced sufficiently such that in the last decade, estimatesof ecosystem properties have become available at the globalscale. Interest in the global carbon (C) cycle and itsrelationship to future climate change has led to studies ofglobal C stocks and fluxes. Results have been used toquantify regional or global net carbon balance, identifylocations with carbon sources or sinks, and propose mech-anisms that drive these responses [Cao et al., 2002; Hicke etal., 2002a; McGuire et al., 2001; Nemani et al., 2002;Nemani et al., 2003; Potter et al., 1999; Schimel et al.,2000].[3] Solar radiation, used to prescribe photosyntheticallyactive radiation (PAR) at the top of the canopy, is among theinputs required by ecosystem models. Although solar radi-ation inputs are sometimes limited to capturing the annualcycle due to the lack of data (e.g., historical runs such asreported by McGuire et al. [2001]), solar radiation doesvary in time. Downwelling surface solar radiation hasdecreased by 1 –3% per decade over the last 50 years asindicated by ground-based measurements (though the timeperiods varied among studies) [Cohen et al., 2004; Gilgen etal., 1998; Liepert, 2002; Stanhill and Cohen, 2001], consis-tent with decreases in pan evaporation over the last 50 years[Roderick and Farquhar, 2002]. Behavior in the 1990ssuggests a recovery, however [Cohen et al., 2004]. Respon-sible mechanisms include changing aerosol and cloud prop-erties [Liepert, 2002; Stanhill and Cohen, 2001].[4] Spatiotemporal variability in solar radiation can sig-nificantly affect carbon fluxes. In their study of global NPPtrends, Nemani et al. [2003] reported large increases in NPPin the tropics driven by increasing solar radiation fromNCEP. Hicke et al. [2002b] compared conterminous UnitedStates NPP calculated using National Centers for Environ-mental Prediction (NCEP) solar radiation with that computedwith data from the VEMAP project [VEMAP Members ,1995]. NPP trends over 12 years were similar with the twosets of input data, but the higher NCEP solar radiation resultedin higher NPP by 10%.[5] For model inputs of temperature and precipitation,multiple sources of data exist that satisfy the necessaryrequirements of global extent and interannual variability(e.g., from NCEP reanalyses [Kistler et al., 2001]). Incontrast, solar radiation is not recorded operationally byweather stations, and coverage of ground-based instrumentsGLOBAL BIOGEOCHEMICAL CYCLES, VOL. 19, GB2006, doi:10.1029/2004GB002391, 2005Copyright 2005 by the American Geophysical Union.0886-6236/05/2004GB002391$12.00GB2006 1of18is sparse. Global estimates of solar radiation therefore relyon models of the propagation of radiation thr ough theatmosphere together with information about atmosphericconditions that affect this propagation. The physical pro-cesses governing radiative transfer through the atmosphereare well known, and with adequate knowledge of the stateof the atmosphere, the solar radiation incident at the surfacecan be calculated with sufficient accuracy for use inecosystem models. Even highly parameterized, fast radia-tive transfer models suitable for inclusion into generalcirculation models (GCMs) calculate incident solar radia-tion to within 4–11% (compared to results from moredetailed models) [Fouquart et al. , 1991].[6] Several types of global solar radiation data sets exist.The first type is produced with a GCM that assimilatesobservations of the atmosphere. In addition to producingglobal, physically consistent estimates of temperature, pres-sure, and winds that are constrained by observations, theGCM uses representations of ot her import ant processessuch as clouds, radiation, and the land surface to model asuite of atmospheric and surface variables.[7] A second type of solar radiation data set is calculatedusing a radiative transfer model (similar to ones used inreanalysis projects) together with satellite-derived atmo-spheric properties, most importantly clouds. Global satelliteobservations of cloud variables have become available aspart of the International Satellite Cloud Climatology Project(ISCCP). Several groups have used the ISCCP cloudvariables together with other information about the atmo-sphere (e.g., water vapor, aerosols) to compute radiationfluxes [Darnell et al., 1992; Pinker and Laszlo,1992;Zhang et al., 1995, 2004]. While limited to the time periodof ISCCP data availability and thus satellite observations,the