REPORTQuantifying mortality of tropical rain forest treesusing high-spatial-resolution satellite dataDavid B. Clark1*, CarlomagnoSoto Castro2, Luis Diego AlfaroAlvarado2and Jane M. Read31Department of Biology,University of Missouri-St Louis,St Louis, MO, USA and La SelvaBiological Station, Puerto Viejode Sarapiquı´, Costa Rica2Forest Science Department,School of the Environment,Universidad Nacional, Heredia,Costa Rica3Department of Geography,Maxwell School, SyracuseUniversity, Syracuse, NY, USA*Correspondence: OTS-Interlink341, P.O. Box 025635, Miami,Florida 33102, USAE-mail: [email protected] of forest responses to climate change is severely hampered by the limitedinformation on tree death on short temporal and broad spatial scales, particularly intropical forests. We used 1-m resolution panchromatic IKONOS and 0.7-m resolutionQuickBird satellite data, acquired in 2000 and 2002, respectively, to evaluate tree deathrates at the La Selva Biological Station in old-growth Tropical Wet Forest in Costa Rica,Central America. Using a calibration factor derived from ground inspection of treedeaths predicted from the images, we calculated a landscape-scale annual exponentialdeath rate of 2.8%. This corresponds closely to data for all canopy-level trees in 18 forestinventory plots, each of 0.5 ha, for a mostly-overlapping 2-year period (2.8% per year).This study shows that high-spatial-resolution satellite data can now be used to measureold-growth tropical rain forest tree death rates, suggesting many new avenues for tropicalforest ecology and global change research.KeywordsCosta Rica, IKONOS, La Selva Biological Station, QuickBird, remote sensing, treemortality rates, tropical rain forest.Ecology Letters (2004) 7: 52–59INTRODUCTIONThe planet’s biological systems are currently experiencingsteadily increasing temperature and atmospheric levels ofCO2and other greenhouse gases (Houghton et al. 2001;Root et al. 2003). The summed current and future effectsof global climate change on forest systems are poorlyunderstood, and even the directions of the expectedecological changes in many systems are controversial.A key factor for predicting and understanding forestresponses to global climate change is measurement of treemortality rates. For example, models that couple climatechange with vegetation response predict increased tree deathrates in tropical forests, leading to a positive feedback withCO2-induced global warming (Jones et al. 2003). Globalwarming will cause upward shifts in forest life zones alongmountain slopes (Enquist 2002), with increased treemortality and disappearance of species at their lowertemperature boundaries (Pen˜uelas & Boada 2003). Onemanifestation of massive global perturbations such asENSO events are increased death rates over large areas oftropical forests (Condit et al. 1995; Nakagawa et al. 2000;Williamson et al. 2000). Forest fragmentation can lead tosignificant carbon loss caused by increased tree mortalityalong fragment borders (Laurance et al. 2000), which in turnis a positive forcing factor on atmospheric CO2levels.Better knowledge of tree mortality responses to increas-ing temperature and greenhouse gases as well as changingland use patterns is of great interest in all forest systems.However, tropical rain forests (TRFs) are of special concern.TRFs contain a substantial fraction of terrestrial biodiversity(Reaka-Kudla et al. 1997), their soils and biomass accountfor a large percentage of the total world terrestrial carbonpool (Dixon et al. 1994), and they are estimated to accountfor roughly one-third of net primary productivity on land(Dixon et al. 1994).The decadal-level trajectory of tree growth and mortalityrates in these forests is controversial, and there is debateover the status of TRFs as carbon sources or sinks(reviewed in Clark 2003). There are several reasons for thiscontroversy. There are few data from long-term permanentforest inventory plots in the tropics, and in some cases,methodological issues limit the use of those few data that doexist (Clark 2002). Existing forest inventory plots in old-growth TRF are few in number, limited in size, and almostalways sited in a non-random fashion with little or noEcology Letters, (2004) 7: 52–59 doi: 10.1046/j.1461-0248.2003.00547.x2003 Blackwell Publishing Ltd/CNRSreplication (Brown 1996; Clark & Clark 2000a; Keller et al.2001). In addition, with only one exception that we areaware of (Clark et al. 2003a), published long-term data are atsupraannual time steps, which severely limit the ability torelate annual forest response to annual climate events suchas ENSO episodes.What is needed to assess forest dynamics in this biomeand its relation to global climate change is a way to measureforest growth and mortality rates over large areas of tropicalforests at annual time scales (Clark et al. 2003b). High-spatial-resolution (£1-m panchromatic, £4-m multispectral)commercial satellite data offer a potential solution to theseproblems. For example, 1-m resolution IKONOS (SpaceImaging Corp., Thornton, CO, USA) satellite data fromterra firme TRF have been used to measure tree crown sizes,to detect the effects of selective logging and to extendremote-sensing capacity to distinguish secondary forestsfrom old growth; these data have also been shown to becorrelated with key forest structural parameters such as plotbiomass and total gap area (Asner et al. 2002; Clark et al.2003b; Hurtt et al. 2003; Read 2003; Read et al. 2003). Thesestudies have also suggested (Read et al. 2003) that it shouldbe feasible to use high-spatial-resolution satellite data tomeasure TRF tree growth and survival.In this paper, we show that it is now possible to measureTRF tree mortality rates using such high-spatial-resolutionsatellite data. Forest-level tree mortality is a key ecosystemvariable, because it integrates the physiological and demo-graphic responses of all trees in a given forest, and it directlylinks to stand biomass and overall structure. We suggest thatthe ability to measure tree mortality over large regions ofTRF using satellite data will have extensive applications fortropical forest research, and that high spatial resolutionsatellite data will be a key tool for assessing the sensitivity ofthis biome to climate and land-use change.METHODSStudy siteThe study site was conducted at the La Selva BiologicalStation of the Organization for Tropical Studies (O.T.S.) inthe Atlantic lowlands of Costa Rica