Evaluating water stress controls on primary production inbiogeochemical and remote sensing based modelsQiaozhen Mu,1Maosheng Zhao,1Faith Ann Heinsch,1Mingliang Liu,2Hanqin Tian,2and Steven W. Running1Received 9 February 2006; revised 8 September 2006; accepted 6 October 2006; published 8 February 2007.[1] Water stress is one of the most important limiting factors controlling terrestrialprimary production, and the performance of a primary production model is largelydetermined by its capacity to capture environmental water stress. The algorithm thatgenerates the global near-real-time MODIS GPP/NPP products (MOD17) uses VPD(vapor pressure deficit) alone to estimate the environmental water stress. This papercompares the water stress calculation in the MOD17 algorithm with results simulatedusing a process-based biogeochemical model (Biome-BGC) to evaluate the performanceof the water stress determined using the MOD17 algorithm. The investigation study areasinclude China and the conterminous United States because of the availability of dailymeteorological observation data. Our study shows that VPD alone can capture interannualvariability of the full water stress nearly over all the study areas. In wet regions, whereannual precipitation is greater than 400 mm/yr, the VPD-based water stress estimate inMOD17 is adequate to explain the magnitude and variability of water stress determinedfrom atmospheric VPD and soil water in Biome-BGC. In some dry regions, where soilwater is severely limiting, MOD17 underestimates water stress, overestimates GPP, andfails to capture the intraannual variability of water stress. The MOD17 algorithm shouldadd soil water stress to its calculations in these dry regions, thereby improving GPPestimates. Interannual variability in water stress is simpler to capture than the seasonality,but it is more difficult to capture this interannual variability in GPP. The MOD17algorithm captures interannual and intraannual variability of both the Biome-BGC-calculated water stress and GPP better in the conterminous United States than in thestrongly monsoon-controlled China.Citation: Mu, Q., M. Zhao, F. A. Heinsch, M. Liu, H. Tian, and S. W. Running (2007), Evaluating water stress controls on primaryproduction in biogeochemical and remote sensing based models, J. Geophys. Res., 112, G01012, doi:10.1029/2006JG000179.1. Introduction[2] Water availability is the primary limiting factor forvegetation growth over 40% of the Earth’s vegetatedsurface, while an additional 33% is limited by cold temper-atures and frozen water, further limiting water availabilityfor plant growth [Nemani et al., 2003]. Vegetation respondsto water deficits in several ways [Waring and Running,1998]. Even mild soil water deficits begin to inhibit cellularexpansion, xylem water flow from roots to leaves, andphloem sugar transfer in the stems of growing plants. Lackof mobile water has similar impacts on reducing plant leafarea, transpiration, growth, and related ecosystem activity[Running and Kimball, 2005].[3] Terrestrial net primary production (NPP), equal to thedifference between gross primary production (GPP) andautotrophic respiration (Ra), plays an important role in thecarbon balance of the biosphere. NPP is receiving increasedattention not only because it is related to the global carboncycle, but also because it is greatly influenced by theassociated effects of changing climate on the carbon cycle[Prentice et al., 2001]. Photosynthesis is the only process bywhich to assimilate CO2from the atmosphere into terrestrialprimary production, and stomata are the major pathways fortransfer of trace gases between vegetation and the atmo-sphere. The stomatal conductance and photosynthetic as-similation rate are largely controlled by environmentalfactors such as irradiance, temperature, water availability,and nutrition [Wong et al., 1985a, 1985b, 1985c; Lange etal., 1987]. Therefore global terrestrial primary productionmodels incorporate these environmental factors to study theterrestrial carbon balance and vegetation dynamics under achanging climate [Cramer et al., 1999; Arora, 2002]. Plantwater deficits induce progressive leaf stomatal closure,reducing plant water loss via transpiration while also slow-JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112, G01012, doi:10.1029/2006JG000179, 2007ClickHereforFullArticle1Numerical Terradynamic Simulation Group, Department of Ecosystemand Conservation Sciences, University of Montana, Missoula, Montana,USA.2School of Forestry and Wildlife Sciences, Auburn University, Auburn,Alabama, USA.Copyright 2007 by the American Geophysical Union.0148-0227/07/2006JG000179$09.00G01012 1of13ing photosynthesis and canopy-atmosphere gas exchange.Sustained drought will produce early leaf senescence andshedding and may impact ecosystem leaf area for a numberof years [Running and Kimball, 2005]. During most of thegrowing season, when light and temperature are greatenough to maximize stomatal conductance, water limitationis the dominant climatic controller of stomatal conductance.There are different approaches to introducing water limi-tations on primary production in different models, and formost process-based ecosystem models, water limitations arecalculated for both the soil and the air [Churkina et al.,1999].[4] The Biome-BGC ecosystem model is a process-based biogeochemical model, simulating ecosystem cyclesof carbon, water and nitrogen at regional and global scales[Running and Hunt, 1993; White et al., 2000; Thornton etal., 2002]. The water stress calculation in Biome-BGC isdetermined by the combined stresses from soil-leaf waterpotential (PSI) and the atmospheric water vapor pressuredeficit (VPD; details in section 2.1). T he operationalMODIS (Moderate Resolution Imaging Spectroradiometer)Production Efficiency Model (MOD17) on the Terra satel-lite is used to generate 8-day near-real-time vegetationprimary production [Zhao et al., 2005] (also F. A. Heinschet al., User’s Guide GPP and NPP (MOD17A2/A3) Prod-ucts NASA MODIS Land Algorithm, 2003, available athttp://www.ntsg.umt.edu/modis/MOD17UsersGuide.pdf).MOD17 is based on the radiation use efficiency logicsuggested by Monteith [Monteith, 1972, 1977; Runninget al., 2000, 2004], and the model is similar to existingproduction efficiency models [Prince and Goward, 1995;Potter et al., 1993; Ruimy et al., 1994; Field et al., 1995].One key difference between MOD17 and Biome-BGC isthe implementation of water stress control through VPD(MOD17)