WOODY PLANT RICHNESS AND NDVI RESPONSE TO DROUGHT EVENTS IN CATALONIAN (NORTHEASTERN SPAIN) FORESTS - Francisco Lloret

Ecology, 88(9), 2007, pp. 2270–2279! 2007 by the Ecological Society of America

WOODY PLANT RICHNESS AND NDVI RESPONSE TO DROUGHT EVENTSIN CATALONIAN (NORTHEASTERN SPAIN) FORESTS

F. LLORET,1,5 A. LOBO,2 H. ESTEVAN,3 P. MAISONGRANDE,4 J. VAYREDA,3 AND J. TERRADAS1

1CREAF (Center for Ecological Research and Forestry Applications) - Departament de Biologia Animal, Biologia Vegetal i Ecologia,Universitat Autonoma de Barcelona, 08193 Bellaterra, Barcelona, Spain

2Institut de Ciencies de la Terra ‘‘Jaume Almera’’ (CSIC), Lluis Sole Sabarıs s/n, 08028 Barcelona, Spain3CREAF (Center for Ecological Research and Forestry Applications), Universitat Autonoma de Barcelona, 08193 Bellaterra,

Barcelona, Spain4Centre d’Etudes Spatiales de la Biosphere (CESBIO), bpi 2801, 18, avenue Edouard Belin, 31401 Toulouse Cedex 9, France

Abstract. The role of species diversity on ecosystem resistance in the face of strongenvironmental fluctuations has been addressed from both theoretical and experimentalviewpoints to reveal a variety of positive and negative relationships. Here we exploreempirically the relationship between the richness of forest woody species and canopyresistance to extreme drought episodes. We compare richness data from an extensive forestinventory to a temporal series of satellite imagery that estimated drought impact on forestcanopy as NDVI (normalized difference vegetation index) anomalies of the dry summer in2003 in relation to records of previous years. We considered five different types of forests thatare representative of the main climatic and altitudinal gradients of the region, ranging fromlowland Mediterranean to mountain boreal-temperate climates.

The observed relationship differed among forest types and interacted with the climate,summarised by the Thorntwaite index. In Mediterranean Pinus halepensis forests, NDVIdecreased during the drought. This decrease was stronger in forests with lower richness. InMediterranean evergreen forests of Quercus ilex, drought did not result in an overall NDVIloss, but lower NDVI values were observed in drier localities with lower richness, and in moremoist localities with higher number of species. In mountain Pinus sylvestris forests NDVIdecreased, mostly due to the drought impact on drier localities, while no relation to speciesrichness was observed. In moist Fagus sylvatica forests, NDVI only decreased in plots withhigh richness. No effect of drought was observed in the high mountain Pinus uncinata forests.

Our results show that a shift on the diversity–stability relationship appears across theregional, climatic gradient. A positive relationship appears in drier localities, supporting a nullmodel where the probability of finding a species able to cope with drier conditions increaseswith the number of species. However, in more moist localities we hypothesize that theproportion of drought-sensitive species would increase in richer localities, due to a higherlikelihood of co-occurrence of species that share moist climatic requirements. The study pointsto the convenience of considering the causes of disturbance in relation to currentenvironmental gradients and historical environmental constraints on the community.

Key words: climate change; ecosystem resistance; forest dieback; Mediterranean forests; NDVI; speciesdiversity.

INTRODUCTION

The role of species diversity in the functioning ofecosystems has become one of the most challengingtopics in recent ecological research. Evidence of greathuman-induced transformations ranging from the localto the global scale has highlighted the importance of thisquestion. One particular issue concerns the role of speciesdiversity on ecosystems’ ability to face strong environ-mental fluctuations, such as extreme climate events,which in some cases can be considered as disturbances.Experimental results point to a positive relationship

between diversity and the resistance of the ecosystem(Tilman and Downing 1994, McGrady-Steed et al. 1997,Naem and Li 1997). However, Loreau (2000) used amodel approach to conclude that resistance may decreaseor increase with increasing diversity. A selection mech-anism (Huston 1997) (equivalent to the sampling effectdescribed by Tilman et al. 1997), by which ‘‘diversityincreases the range of trait variation and a selectiveprocess promotes dominance by species with extremetrait values’’ (Loreau 2000), would explain these differentrelationships. A positive relationship would be found ifthe disturbance has a negative effect on most species anddiversity enhances the probability of finding the speciesbest adapted to afford the new conditions (Loreau 2000).This response would be enhanced by historically-fluctuating selective pressures that remain promoting

Manuscript received 13 July 2006; revised 7 February 2007;accepted 12 February 2007. Corresponding Editor: T. J.Stohlgren.

5 E-mail: [email protected]

2270

the occurrence of traits fitting the new conditions.Alternatively, a greater number of species may implygreater changes in ecosystem properties under the newdisturbance conditions when competitive pressuresunder non-disturbance conditions tended to eliminatedisturbance-tolerant species from the community.Climate models and current empirical trends point to an

increase of climatic variability in some regions, resulting ina greater number and intensity of extreme events (Voss etal. 2002). In the particular case of the MediterraneanBasin, where water availability is the major environmentalconstraint to plant growth, and where episodes of droughtresulting in forest dieback have been recorded in the lastdecades (Penuelas et al. 2001, Lloret et al. 2004), currentmodels predict warmer and drier conditions (Houghon etal. 2001, Gibelin and Deque 2003), with increasedseasonal and inter-annual variability resulting in morecommon and intense drought events (Houghton et al.2001, Sanchez et al. 2004).As experimental manipulation of climate and richness

in forest ecosystems is unlikely to be performed, directobservations of the effect of extreme climate events arecurrently the most reliable approach for investigatingthe relationship between forest species richness and theeffects of extreme drought on forest properties. Inaddition, extensive sampling is recommended to inves-tigate regional patterns, on a similar scale to that atwhich drought occurs. The scaling-up of the level ofobservation requires integrative variables related toecosystem functioning, although accurate details offunctioning at stand level may be missed. Remotesensing imagery has been revealed as a useful tool todescribe structural patterns of vegetation determiningprimary productivity in relation to between-year climat-ic variability at local and regional scales (Tucker andSellers 1986, Paruelo et al. 2001). NDVI (normalizeddifference vegetation index) is a normalized contrastbetween near infrared and red reflectance, thus empha-sizing the presence of green vegetation in the areacorresponding to the pixel. Since NDVI integrates boththe abundance of green vegetation and its greenness, ithas a certain consubstantial ambiguity. Also, as NDVIis a two-dimensional approximation to a three-dimen-sional reality in which leaves tend to be arranged indifferent layers, the relationship between NDVI andPAR tends to saturate in the case of highly multilayeredcanopies. Notwithstanding, if time series are availablefor post-processing and certain observations (i.e.,extreme angles) are disregarded, NDVI is a linearestimate of the fraction of photosynthetically activeradiation (PAR) that is intercepted by the photosynthe-sizing tissue of vegetation present in the pixel. Becauseof the aforementioned multilayering, NDVI correlationsbetween NDVI and LAI (leaf area index; i.e., Chen andCihlar 1996, Cohen et al. 2003) are more variable, with aless close fit and quicker saturation than in the case ofAPAR. NDVI has also been found to correlate toecosystem CO2 flux (Wylie et al. 2003, Li et al. 2005).

In addition, extensive forest inventories providereliable information about the forest composition ofwoody species across large territories, and both types ofinformation may be integrated into geographical infor-mation systems (GIS).

During the summer of 2003 an intense droughtepisode occurred in southwestern Europe, which waswell documented by remote sensing imagery (Gobron etal. 2005, Lobo and Maisongrande 2005). In the presentstudy, we analyze the relationship between woodyspecies richness obtained from extensive forest invento-ries in Catalonia (northeastern Spain) and droughtimpact on the canopy of forests, estimated as NDVIanomalies of the year 2003 in relation to median NDVIrecords of previous years. We assume that the canopyNDVI values in the dense forests under study weremostly determined by woody species. We consider fivedifferent types of forests, defined by their dominant treespecies, which are representative of the main climaticand altitudinal gradients of the region. Our mainhypothesis is that species richness correlates positivelywith canopy resistance to drought, this pattern beingmore evident in those communities where droughtintensity was higher, and where drought-driven selectionhas been historically prevalent.

METHODS

Study area

Catalonia (08150 E, 408300 N to 38150 E, 428400 N,;31 900 km2) is located in the northeast of the IberianPeninsula, including the Pyrenean range to the northand bounded to the east by the Mediterranean Sea.Most of the area falls under different degrees ofMediterranean climate, with a main climatic-topograph-ic gradient from a temperate-boreal climate (meanannual precipitation of up to 1500 mm) in the Pyrenees,to a Mediterranean climate of mild winters and drysummers (mean annual precipitation from 530 mm) inthe southern extreme. There is also a continental,semiarid gradient inland towards the west, with colderwinters, very hot summers, and low rainfall (meanannual precipitation around 400 mm). Vegetation typesacross the climatic-topographic gradients include scle-rophyllous, evergreen shrublands and forests, deciduousforests, and coniferous forests (Folch 1981), all with along history of human management.

Climatic data and remote sensing estimationof forest drought

From the Atlas Climatic Digital de Catalunya, weobtained for each plot locality the Thorntwaite index asa standard index that effectively describes the aridity–moisture climatic gradient, from the Mediterraneancoast to the mountains inland (Lloret et al. 2005; atlasavailable online).6 The Thorntwaite index wascalculated as follows:

6 hhttp://magno.uab.es/atles-climatic/i

September 2007 2271DROUGHT RESISTANCE AND PLANT RICHNESS

Thorntwaite index ! "P# ETP$3 100=ETP

where P is annual precipitation (mm), and ETP is annualpotential evapotranspiration (mm), which is the sum ofmonthly evapotranspiration values calculated followingthe Thorntwaite formula (Thorntwaite 1948). Thiscalculation considers for each plot locality the meanmonthly average values of temperature, that wereobtained from the Atlas Climatic Digital de Catalunya,and a coefficient estimated by the numbers of days in amonth and the daily hours of sun as a function oflatitude.The difference between precipitation and evapotrans-

piration can be considered as an estimation of thebalance between atmospheric water supply and demand(Lobo and Maisongrande 2005). During the droughtepisode of the summer of 2003, in most of our region ofstudy, the difference between total precipitation andevapotranspiration showed a negative anomaly of morethan 100 mm (Lobo and Maisongrande 2005). Thisinformation was obtained by comparing climate andmeteorological data of June, July, and August 2003 todata of the same months for the period 1961–2000,provided as grids of 53 5 km resolution by the SpanishInstituto Nacional de Metereologia (INM) (Lobo andMaisongrande 2005).The vegetation response to water deficit was assessed

from the normalized difference vegetation index (NDVI;Tarpley et al. 1984) included in the S10 products that arederived from images acquired by the VEGETATIONinstrument onboard satellite SPOT. VEGETATION isan optical multi-spectral instrument that acquires a dailyand almost complete cover of the Earth’s surface at 1-km2 resolution in four spectral bands (Hagolle et al.2004). S10 products are calibrated, atmospherically andgeometrically corrected images.The VEGETATION S10 collection is produced from

temporal compositing of calibrated, atmosphericallyand geometrically corrected images, by the combinationof daily images in periods of 10 days. The aim is tocreate a synthetic reconstruction from cloud-free imagesthat is assumed to be representative of each 10-dayperiod. The maximum value composite (MVC; Holben1986) procedure is used for compositing in VEGETA-TION S10 products. MVC selects, for each pixel, themaximum NDVI value among 10 daily images. Al-though alternative compositing methods have beendeveloped (Hagolle et al. 2005), MVC is the onlymethod currently implemented in the processing chainof the S10 collection. Time series of NDVI compositesproduced by this method have been used successfully tomonitor surface dynamics at global and regional scales(Zhou et al. 2001, Lucht et al. 2002, Nemani et al. 2003).We used a data set of S10 products of the region of

study from 1999 to 2003, with a resolution of 3200

(seconds). NDVI values in S10 products are linearlyscaled from the observed j#0.1, 0.9j range to the integerrange j0, 250j, and we kept this scale. We computed an

average annual series of monthly mean NDVI imagesfor the period 1999–2002, which we used as the‘‘normal’’ reference, and an annual series of monthlymean NDVI images for 2003. We calculated thedifference between the respective 2003 and 1999–2002mean NDVI values, and we finally estimated NDVIanomalies as the difference between the respective 2003and 1999–2002 mean NDVI values (Fig. 1). Consideringthat the period of drought concluded by the end ofAugust in southwestern Europe, we use in this study theanomaly of NDVI in August 2003 as an estimate of theimpact of the drought in the canopy. Also, this month isat the peak of the dry season, when a large part of theherbaceous vegetation has probably burned off, causingthe minimal interference to the NDVI scores.

Species richness data

Richness data are based on circular field plotsestablished in the Third National Forest Inventory ofSpain (IFN3; Direccion General de Conservacion de laNaturaleza 2006). Sampling was conducted from 2000 to2001 at a density of 1 plot/km2, following a regulardesign within the forested surface of the whole territory(Fig. 1). Each plot was located in the field by a globalpositioning system (GPS), allowing for cross-referenceswith climatic databases (see footnote 6).The IFN3 survey includes exhaustive information on

the composition of canopy and understory woodyspecies, as well as on production and structure. Werestricted our analysis to this group as no consistentinformation is available for non-woody species. Shruband regenerative trees (defined as those with a normaldiameter, measured 50 cm above the ground surface,below 75 mm) were sampled within circular plots with aradius of 10 and 5 m, respectively). Extensive stands ofregenerative trees after clearing are not common in theregion and are not included in the study. Plot size forsampling the rest of the trees varied in order to sampleenough individuals belonging to the largest size classespresent in the plot. In order to minimize species richnessbias due to plot size, we compared the richness shown byplots of different size in each forest type. In the caseswhere differences arose (Fagus sylvatica and Pinussylvestris forests), we only selected the most commonsize (15 m radius), which was considered to be the mostrepresentative for that type of forest. When there wereno significant differences in species richness betweenplots of different size (Quercus ilex, Pinus halepensis, andPinus uncinata forests), we pooled these plots in theanalysis. This process reduced the original 9126 IFN3plots for Catalonia to 7567, most of them with a radiusof 15 m. Finally, since large variation in richness amongplots may produce no significant differences betweenplot size classes, we minimized the effect of plot size onthe relationship between NDVI anomaly and speciesrichness by including it as an explicative variable in thestatistical analysis, as detailed below.

F. LLORET ET AL.2272 Ecology, Vol. 88, No. 9

Data analysis

In order to ensure a reliable correspondence betweenIFN3 information and pixel information, we applied anadditional filter to avoid plots located in heterogeneousareas, so that reflectance from the forest canopy doesnot dominate the corresponding pixel values. Thus, wedefined a 1-km2 grid coincident with the pixels of theimagery and selected those cells including inventoryplots and satisfying two criteria: (1) land cover categoryhaving dense forest in more than 50% of the cell, and (2)tree canopy cover being more than 50% in the respectiveinventory plots. This second filter reduced the number ofplots (and pixels) to 3504. After applying the filters, wecrossed the selected inventory data and the NDVI valuesobtained in the respective 1-km2 pixel surrounding theplot location, following the nearest neighbor procedureof the Miramon GIS (Pons 2000). Since 1-km2 NDVIanomalies are obtained from averaged continuous forestcanopies, we assumed that they can be applied to smallerpieces of forest, such as those sampled at the plot scale.We also assumed that the number of woody species at

plot level is a reliable estimate of woody species richnessin the surrounding forest areas. More specifically, weassumed that the spatial pattern of richness in smallsampling units 1 km apart follows the same pattern asadjacent 1-km2 units. If this assumption is true, richness

from plots should show some degree of autocorrelationin short distances (1 km). Alternatively, a randomspatial distribution of richness values from sites 1 kmapart would indicate that the richness estimated in smallsampling units is not representative of spatial patternson a larger scale. This assumption was supported by aspatial autocorrelation analysis of the selected plots,which indicated that the pattern of species richness doesnot change across the neighbouring areas (Mantel test inwhich the main factors were the absolute differencesbetween species richness for each pair of inventory plotsand the geographical Euclidean distance between them, r! 0.125, P , 0.001; for distances up to 15 km, r! 0.075,P , 0.001; for distances up to 60 km, r ! 0.036, P ,0.001).

We also tested this assumption by comparing plotrichness values to records of species richness on largerspatial scales obtained from the BIOCAT data bank(available online).7 BIOCAT provides the list of plantspecies recorded by botanists at 103 10 km UTM gridscovering the whole Catalonia territory. We selected 33 103 10 km grids where forest cover was largely dominant(more than 75% of the territory) and included at least 10plots, from which we calculated the mean species

FIG. 1. Map of the studied region (Catalonia, Spain) showing the pattern of the woody species richness obtained from fieldplots (left) and the 2003 summer NDVI (normalized difference vegetation index) anomaly (right). The darker the gray, the greater isthe species richness and drought impact, respectively. Negative values of the NDVI anomaly correspond to lower values of NDVIin August 2003 than in the reference August. Values of NDVI are linearly scaled to the range j0, 250j.

7 http://biodiver.bio.ub.es/biocat/homepage.htmli


richness. We obtained a significant positive correlation (r! 0.765, F1,32! 43.74, P , 0.0001, slope! 4.39) betweenwoody species richness at the plot scale and at the 10310km scale, supporting our assumption that richness valuesobtained at plot level may be correlated to NDVI 1-km2

estimates at intermediate scales.We selected five types of forests that were represen-

tative of the climatic conditions of the area of study.These forest types were defined by its respectivedominant species (accounting for .50% of the plotbasal area): Pinus halepensis (Mediterranean, coniferousforests), Quercus ilex (broadleaf, evergreen, Mediterra-nean), Pinus sylvestris (mesic and mountain coniferousforests), Fagus sylvatica (moist, broadleaf, deciduousforests), and Pinus uncinata (high-mountain, coniferousforests) (Folch 1981, Gracia et al. 2004). As wedisregarded those plots located in other types of forests,the total number of plots (and pixels) was 2087: 464 P.halepensis plots, 565 Q. ilex plots, 749 P. sylvestris plots,91 F. sylvatica plots, and 218 P. uncinata plots.We built a general linear model (GLM), including the

NDVI anomaly (difference between August 2003 andmean values for the same month of the period 1999–2002) as the dependent variable, and forest type, woodyspecies richness, and Thorntwaite index as independentfactors. We included the interactions between theseindependent variables in the model. Since some sam-pling points may experience some degree of spatialautocorrelation, we included in the model two spatialterms corresponding to UTM x and y coordinates.Including the x and y coordinates may not compensatefor small scale spatial biases. Given the large number ofsampling points, small scale heterogeneity is unlikely tobias strongly our main results, but we tested theautocorrelation pattern of NDVI anomalies by correlo-gram based on Mantel tests and we found that whenconsidering larger distances, NDVI anomaly tends todiverge gradually without relevant leveling that couldindicate small scale heterogeneity at range lower thanabout 40 km. We also checked GLM analysis includinghigher order spatial coordinate parameters, but they didnot improve the model, instead in several cases theyresulted in collinearity problems. Finally, since plot size

may determine richness values and consequently mayindirectly influence the relationship between NDVIanomaly and species richness we also included plot sizein the model as an additional explicative variable.

We also built similar GLMs for each type of forest,including woody species richness, Thorntwaite index,UTM x and y coordinates and plot size as explicativevariables. Since all the plots in F. sylvatica and P.sylvestris were of the same size (15 m radius), thisvariable was not included in the analysis of these twotypes of forests.

RESULTS

The relationships between NDVI anomaly and woodyspecies richness and between NDVI anomaly andThorntwaite index were not the same in the differenttypes of forests, as shown by the significant interactionsforest type 3 richness and forest type 3 Thorntwaite(F4,2063 ! 3.37, P , 0.005 and F4,2063 ! 2.56, P! 0.037,respectively). NDVI anomaly was also significantlyinfluenced by the UTM y coordinate (F4,2063 ! 74.42,P , 0.001). Therefore, we analyzed these relationshipsfor each forest type separately. The effect of plot sizewas not significant in the analysis of the modelconsidering the five types of forests (F2,2065 ! 0.63, P !0.533), nor was it in the analysis of each forest separately(Table 1).

NDVI values in P. halepensis forests significantlydecreased in the 2003 summer (anomaly mean: #2.14,SD! 7.98, Student’s t test of significant differences fromzero, t ! 5.77, P , 0.001). NDVI loss was lower in P.halepensis plots with higher woody species richness(Table 1, Fig. 2). Although NDVI loss was notsignificantly influenced by the Thorntwaite index, amarginally significant interaction between woody speciesrichness and Thorntwaite index points to more vulner-ability to the 2003 drought episode in moist localities(within the distribution range of these forests) with lessspecies (Table 1, Fig. 1). NDVI anomaly was signifi-cantly influenced by the UTM y coordinate, with lessdrought impact in northern localities occupied by thistype of forest.

TABLE 1. GLM results for the different forest types considering NDVI anomaly (August 2003) as the dependent variable andwoody species richness, Thorntwaite index, UTM x and y coordinates, and plot size as explicative variables.

Parameter

Pinus halepensis Quercus ilex

F1, 457 PParameterestimate F1, 556 P

Parameterestimate

Richness (R) 7.02 0.008 0.238 (0.009) 10.33 ,0.001 0.402 (0.105)Thorntwaite index (T ) 1.82 0.178 #0.118 (0.009) 14.61 ,0.001 0.115 (0.036)R 3 T 3.16 0.076 0.076 (0.011) 6.74 0.010 #0.007 (0.003)Coordinate x 0.08 0.776 0.79 0.374Coordinate y 58.97 ,0.001 1.76 0.185Plot size 0.01 0.909 0.27 0.603

Notes: Values in parentheses are SE.


FIG. 2. Contour graphs of the 2003 summer NDVI (normalized difference vegetation index) anomaly (linearly standardized) inrelation to woody species richness and Thorntwaite index (lower values are found under drier conditions) for the five types offorests. Darker surfaces indicate more negative anomalies, i.e., lower NDVI than in previous years, as shown by anomaly valuesdrawn over the contour lines. Arrows show the direction of the drought impact, estimated by NDVI decrease in relation to thespecies richness and the climatic gradient. Dots indicate the situation of field-sampled plots in the richness–Thorntwaite indexspace. The graphs were obtained after fitting a second-order polynomical surface, following the GLM module of the Statistica 5.1.package (StatSoft, Tulsa, Oklahoma, USA).

TABLE 1. Extended.

Pinus sylvestris Fagus sylvatica Pinus uncinata

F1, 743 PParameterestimate F1,85 P

Parameterestimate F1, 210 P

Parameterestimate

0.17 0.680 #0.015 (0.073) 5.96 0.017 #0.187 (0.114) 0.28 0.595 0.502 (0.943)11.15 ,0.001 0.119 (0.031) 2.71 0.104 #2.194 (0.899) 0.84 0.361 0.038 (0.042)0.17 0.678 #0.0004 (0.0003) 1.38 0.243 0.014 (0.012) 0.06 0.813 #0.002 (0.007)9.48 0.002 6.54 0.012 46.80 ,0.00122.15 ,0.001 14.08 ,0.001 1.68 0.197

0.35 0.704


In Q. ilex forests, NDVI in August 2003 wassignificantly higher than in previous years (anomalymean! 0.98, SD! 6.61, t! 3.53, P , 0.001). Althoughboth richness and the Thorntwaite index were positivelycorrelated to this increase (Table 1), a significantinteraction between them indicates that lower NDVIvalues were observed in drier localities with lowerspecies richness and they also appear to be in moisterlocalities with a higher number of species, althoughrelatively few plots shared this profile (Table 1, Fig. 2).Pinus sylvestris forests showed a significant decrease

in NDVI values (mean!#3.80, SD! 7.02, t! 14.81, P, 0.001). This decrease was not related to woody speciesrichness, but was lower in moister localities (Table 1,Fig. 2). The NDVI anomaly was significantly influencedby the UTM x and y coordinates, indicating less droughtimpact in southern and western inland localities.NDVI values of Fagus sylvatica forests were not

significantly lower in August 2003 (mean!#0.44, SD!8.03, t !#0.52, P ! 0.606), and consequently there wasno significant correlation with the Thorntwaite index(Table 1). However, NDVI values decreased significant-ly in plots with higher woody species richness (Table 1,Fig. 2). NDVI anomaly was significantly determined bythe UTM x and y coordinates, with less drought impactin northern and eastern localities.Pinus uncinata forests did not show significantly

different values of NDVI in August 2003 whencompared to previous years (mean !#0.22, SD ! 7.10,t ! #0.46, P ! 0.643) (Table 1), and this patternremained present in the plots with higher speciesrichness (Fig. 2). NDVI anomaly was significantlyinfluenced by the UTM x coordinate, with less droughtimpact in eastern localities.

DISCUSSION

The regional decrease in vegetation green cover(NDVI) in southern Europe (Gobron et al. 2005, Loboand Maisongrande 2005) during the 2003 droughtepisode exhibited important differences between foresttypes, although water deficit occurred across the wholeterritory under study. Lobo and Maisongrande (2005),after crossing the 2003 remote-sensing anomaly withCORINE Land Cover 2000 cartography, reportedhigher anomalies in herbaceous than in woody vegeta-tion, and in deciduous than in evergreen broadleafforests, but no analysis of coniferous forests wasundertaken. Our study excludes herbaceous communi-ties but also found more NDVI loss in F. sylvaticaforests than in Q. ilex ones. Evergreen canopies areexpected to be more resistant to NDVI changes thandeciduous canopies, which should be able to respondfaster to changes in water availability by shedding leavesor losing other photosynthetic tissues. Unexpectedly, Q.ilex forest even showed higher NDVI values in the 2003summer than in previous years, perhaps due to the lossof old leaves that are more likely to fall under droughtconditions (Ogaya and Penuelas 2003). As expected,

among coniferous forests, NDVI decreased in Mediter-ranean (P. halepensis) and mesic (P. sylvestris) forests,while it did not change significantly in mountain (P.uncinata) pine forests. The lack of drought effect on P.uncinata forests also concurs with the patches of positiveNDVI anomaly found by Lobo and Maisongrande(2005) in high-mountain areas, where water availabilitywas sufficient to compensate for the 2003 drought event.This general concordance with studies based on land

cover maps supports our approach based on scaling-down parameters obtained for large spatial units, inwhich remote-sensing information is available, to standlevel information taken from field surveys. In fact,special care was taken to select representative foreststands that were located in the middle of forestedpatches of land. This is relevant because an importantsource of variability in NDVI signal comes fromherbaceous vegetation, which experiences importantphenological changes over the course of a summer(Lobo and Maisongrande 2005). Furthermore, upscal-ing approaches have been used when field measures of agiven variable are unlikely to be obtained for the wholeterritory (Williams and Rastetter 1999, Hernandez-Stefanoni and Ponce-Hernandez 2004). In our case, theassumption that species richness recorded at plot level isa reliable estimate of woody plant diversity on a largerscale is supported by correlation to records for largerpieces of land (10 3 10 km) obtained from plantdistribution data banks. Also, the observed autocorre-lation in Mantel tests ensures that local heterogeneitydoes not result in random patterns on a scale of 1 km.Finally, extensive forestry inventories are the mostreliable source of information at a regional level onspecies coexisting in the field. This empirical informationavoids the problems of other estimates of regionalpatterns of species richness based on merging individualspecies distribution maps, which do not take intoaccount species interactions or local habitat variability(Terradas et al. 2004). However, our study does notconsider habitat heterogeneity within 1-km2 pixels. Thismay result in a source of error when estimates of speciesrichness are scaled-up.The relationship between species richness and drought

resistance showed an important variability between thedifferent types of forests. Forest type is a complexcategory involving species composition and foreststructure resulting from community assembly underdifferent climatic constrictions and human management.In fact, the five selected forests correspond to categoriesdistributed in a complex gradient from drier, warmerconditions for open P. halepensis and close Q. ilexforests to moister conditions for P. sylvestris and F.sylvatica ones, and to cold, mountain conditions for P.uncinata ones. However the climatic gradient, synthe-sized by the Thorntwaite index, failed in itself to explainthe pattern of NDVI anomaly variation, as reflected bythe significant interaction between forest type and theclimatic index. The Thorntwaite index is a complex


environmental variable correlated with altitude, temper-ature, and precipitation that did not account for thecomplete set of factors contributing to forest canopyresponse to drought. Structural differences amongforests types also emerge (Gracia et al. 2004), associatedwith different management including selection logging inQ. ilex and F. sylvatica forests, shelterwood systems inP. sylvestris forests and seed tree systems in P. uncinataforest, while logging is relatively rare in unproductive P.halepensis forests. Overall, these types of forest can beconsidered as distinct systems where the relationshipbetween species diversity and ecosystem function mayperform differently.Within each type of forest, patterns of drought impact

in relation to species richness may result from theinteraction with other variables, such as those related toclimate. As previously reported in herbaceous commu-nities (Tilman and Downing 1994), the hypothesis ofspecies richness enhancing drought resistance wassupported in forests growing under Mediterraneanconditions, such as P. halepensis forests. According tothe selection mechanism, the greater the number ofspecies, the greater the probability of finding species ableto cope with drying conditions. In this case, the negativeeffect of the drought episode on species performancecoincides with the historical selective pressures that havepromoted the acquisition of drought tolerance oravoidance traits in Mediterranean species (Mooney1989, Martınez-Ferri et al. 2000). This hypothesis isalso supported by the trend of drier localities of P.halepensis forests with more species to exhibit lowerNDVI anomalies than moister ones. In the more mesicQ. ilex forest, a more complex pattern of interactionbetween species richness and climate appears. As in P.halepensis forests, NDVI values were higher in richerforests growing in drier conditions, but the oppositetrend was observed in moister localities, where moreNDVI was observed in forests with low richness. Salaand Tenhunen (1994) also found that water deficitsduring dry summers were more severe in valley bottomlocalities of Q. ilex forests than in ridge top ones, wheretrees may have developed more conservative strategiesof water use.Significant relationships between species richness and

drought resistance tended to disappear in more humidP. sylvestris forests, where NDVI decrease becamesignificant in drier localities. In these types of forests,historical selective pressures may have not favouredthose species able to face drought, making the wholecommunity more sensitive to drier conditions. Withinthe moister F. sylvatica forests, the climatic gradientwould not be sufficiently contrasted to produce such apattern. The distribution of these forests is restricted inthe moister temperate localities of the region and theyare a refuge for many Euro-Siberian species (Folch1981). In these cases, species-rich communities made upof many drought-sensitive species exhibited high nega-tive NDVI anomalies. Finally, as expected, in high

mountain P. uncinata forests where the drought effectwas not significant, species richness was not related toNDVI variability.

Therefore, the proportion of drought-tolerant specieswithin the community would play an important role inexplaining NDVI anomaly. In drier P. halepensis forests,higher diversity would involve the recruitment of speciesthat are drought tolerant. In contrast, in F. sylvaticaforests, occurring in moister zones, higher diversitywould be caused by the recruitment of drought-intolerant species, and the more diverse canopy wouldlead to higher evapotranspiration rates. Quercus ilexforests, intermediate between dry and somewhat moisterzones, seem to display a more complex response,somewhere between that of the P. halepensis and F.syvatica forests.

The recovery of forest canopy after the droughtepisode is not considered in this study, but canopygreenness is expected to have achieved average valuesafter rainfall, as reported by field observations inprevious drought episodes (Penuelas et al. 2001, Lloretet al. 2004). Since an increase of extreme climateepisodes is expected in some regions, further explorationof the role of diversity in forest resistance in the face offrequent droughts may be achieved by increasing theeffort put into long-term monitoring and the surveyingof climate forest diebacks (Suarez et al. 2004).

Our exploration of the relationship between speciesrichness and ecosystem resistance may have also beeninfluenced by the parameter used to estimate ecosystemdrought-resistance. NDVI is a rough estimate of someglobal ecosystem properties, such as productivity, that isaffected by multiple factors, including canopy structureand moisture state. To minimize these limitations, weused forest homogeneous pixels, and the NDVI valueobtained in the drought period was compared to averagevalues of normal years. Therefore, NDVI loss alsoillustrates the relevant ecological consequences ofdrought on the physiological (i.e., loss of photosyntheticactivity), structural (i.e., crown partial dieback), func-tional (i.e., complete or partial foliage drop andreduction leaf longevity with consequences on nutrientcycling), and demographic (i.e., tree mortality) proper-ties of the forest ecosystem, although it reflects poorlythe details of these responses in the understoryvegetation.

Species homogeneity may be partly responsible for theunexplained effect of richness on drought impact: plotswith a high number of species may be dominated by justa few of them, while a few species on poor plots may beevenly represented in the canopy, greatly influencingNDVI. Unfortunately, our forestry inventories did notprovide enough reliable information to calculate thisparameter because the estimates of the abundance ofrare species within plots were not accurate enough.Nevertheless, we managed to detect that the number ofspecies significantly explained the variability of NDVIanomalies, probably because plot richness acts as a


robust indicator of diversity patterns in this group ofplants on a larger scale, as supported by correlationswith the number of woody species recorded at a scale of103 10 km in the BIOCAT data bank (see footnote 6).In spite of the difficulties of scaling up the species

diversity patterns, our study provides a novel insightinto large-scale empirical patterns in the relationshipbetween the species richness and drought-resistance oftemperate forests, in the face of increasing episodes ofrapid climate change. Thus, our study expands theinformation provided by experiments in diversitymanipulation on herbaceous communities across climat-ic gradients that have shown a positive correlationbetween species diversity and productivity (Hector et al.1999).Our study supports the idea that diversity–stability

relationships are not simple, as established by otherempirical and theoretical studies (Loreau 2000, Pfistererand Schmid 2002). Overall, a shift in this relationshipseems to appear across the regional aridity gradient.Forests from drier localities seem to have a pool ofspecies that allow a selection mechanism to operateunder extreme drought conditions. This results in animprovement in resistance in accordance with speciesrichness. However, in moister localities, drought-resis-tant species are likely to be competitively suppressed bydrought-sensitive ones, with higher growth rates undermoist climate regimes. In other words, there would be ahigher likelihood of co-occurrence of species sharingsimilar moist climatic requirements. Thus, if theproportion of drought-sensitive species increases withspecies richness in moist communities, these forestswould be more vulnerable to drought events. Finally,when the climatic episode is not intense enough, as in theP. uncinata forests of the Pyrenean Mountains, norelationship is found. Thus, the study points to thenecessity of incorporating into the analysis of therichness–resistance relationship an axis relating thecauses of disturbance (i.e., in our case, water availabil-ity) to current environmental gradients and to historicalenvironmental constraints on the community.

ACKNOWLEDGMENTS

We thank J. Penuelas for the comments to an earlier versionof the manuscript and J. J. Ibanez for his advice on theinventory data. We also are grateful to P. Lambdon and ananonymous reviewer for their suggestions. The study has beensupported by SI02IR0010 project of the Institut d’EstudisCatalans (IEC), by REN 2003-07198 and CGL2006-01293/BOSprojects of the Spanish MCYT and by project HIDROLIM, ofthe Departament d’Universitats, Recerca i Societat de laInformacio of the Generalitat de Catalunya (Spain) and Regiondes Midi-Pyrinees (France).

LITERATURE CITED

Chen, J. M., and J. Cihlar. 1996. Retrieving leaf area index ofboreal conifer forests using Landsat TM images. RemoteSensing of Environment 55:153–162.

Cohen, W. B., T. K. Maiersperger, S. T. Gower, and D. P.Turner. 2003. An improved strategy for regression of

biophysical variables and Landsat ETM%data. RemoteSensing of Environment 84:561–571.

Direccion General de Conservacion de la Naturaleza. 2006.Tercer Inventario Forestal Nacional 1997–2006, Barcelona.Ministerio de Medio Ambiente, Madrid, Spain.

Folch, R. 1981. La vegetacio dels Paısos Catalans. Ketres,Barcelona, Spain.

Gibelin, A. L., and M. Deque. 2003. Anthropogenic climatechange over the Mediterranean region simulated by a globalvariable resolution model. Climate Dynamics 20:327–339.

Gobron, N., B. Pinty, F. Melin, M. Taberner, M. M.Verstraete, A. Belward, T. Lavergne, and J. L. Widlowski.2005. The state of vegetation in Europe after following the2003 drought. International Journal of Remote Sensing 26:2013–2020.

Gracia, C., J. A. Burriel, J. J. Ibanez, T. Mata, and J. Vayreda.2004. Inventari Ecologic i Forestal de Catalunya, Catalunya.Centre de Recerca Ecologica i Aplicacions Forestals,Bellaterra, Spain.

Hagolle, O., A. Lobo, P. Maisongrande, B. Duchemin, and A.De Pereira. 2005. Quality assessment and improvement ofSPOT/VEGETATION level temporally composited productsof remotely sensed imagery by combination of VEGETA-TION 1 and 2 images. Remote Sensing of Environment 94:172–286.

Hagolle, O., J. M. Nicolas, B. Fougnie, F. Cabot, and B.Henry. 2004. Absolute calibration of VEGETATION de-rived from an interband method based on the sun glint overocean. IEEE Transactions of Geoscience and RemoteSensing 42:1–20.

Hector, A., et al. 1999. Plant diversity and productivityexperiments in European grasslands. Science 286:1123–1127.

Hernandez-Stefanoni, J. L., and R. Ponce-Hernandez. 2004.Mapping the spatial distribution of plant diversity indices ina tropical forest using multi-spectral satellite image classifi-cation and field measurements. Biodiversity and Conserva-tion 13–14:2599–2621.

Holben, B. N. 1986. Characteristics of maximum-valuescomposite images from temporal AVHRR data. Internation-al Journal of Remote Sensing 7:1417–1434.

Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. vander Linden, X. Dai, K. Maskell, and C. A. Johnson, editors.2001. Climate change 2001: the scientific basis. Contributionof Working Group I. Third Assessment Report of theIntergovernmental Panel on Climate Change. CambridgeUniversity Press, Cambridge, UK.

Huston, M. A. 1997. Hidden treatments in ecological experi-ments: re-evaluating the ecosystem function of biodiversity.Oecologia 110:449–460.

Li, S. G., J. Asanuma, A. Kotani, W. Eugster, G. Davaa, D.Oyunbaatar, and M. Sugita. 2005. Year-round measurementsof net ecosystem CO2 flux over a montane larch forest inMongolia. Journal of Geophysical Research: Atmospheres110(D9): Article No. D09303.

Lloret, F., H. Estevan, J. Vayreda, and J. Terradas. 2005. Fireregenerative syndromes of forest woody species across fireand climatic gradients. Oecologia 146:461–468.

Lloret, F., D. Siscart, and C. Dalmases. 2004. Canopy recoveryafter drought dieback in holm-oak Mediterranean forests ofCatalonia (NE Spain). Global Change Biology 10:2092–2099.

Lobo, A., and P. Maisongrande. 2005. Stratified analysis ofsatellite imagery of SW Europe during summer 2003: thedifferential response of vegetation classes to increased waterdeficit. Hydrology and Earth System Sciences Discussions 2:2025–2060.

Loreau, M. 2000. Biodiversity and ecosystem functioning:recent theoretical advances. Oikos 91:3–17.

Lucht, W. I., W. I. C. Prentice, R. B. Myneni, S. Sitch, P.Friedlingstein, W. Creamer, P. Bousquet, W. Buermann, andB. Smith. 2002. Climatic control of the high-latitude


vegetation greening trend and Pinatubo effect. Science 296:1687–1688.

Martınez-Ferri, E., L. Balaguer, F. Valladares, J. M. Chico,and E. Manrique. 2000. Energy dissipation in drought-avoiding and drought-tolerant tree species at midday duringthe Mediterranean summer. Tree Physiology 20:131–138.

McGrady-Steed, J., J. M. Harris, and P. J. Morin. 1997.Biodiversity regulates ecosystem predictability. Nature 390:162–165.

Mooney, H. A. 1989. Chaparral physiological ecology:paradigms revisited. Pages 85–90 in S. C. Keeley, editor.The California chaparral. Paradigms re-examined. NaturalHistory Museum of Los Angeles County, Los Angeles,California, USA.

Naeem, S., and S. Li. 1997. Biodiversity enhances ecosystemsreliability. Nature 390:507–509.

Nemani, R. R., C. D. Keeling, J. Hashimoto, W. M. Jolly, S. C.Pipar, C. J. Tucker, R. B. Myneni, and S. W. Running. 2003.Climate-driven increases in global terrestrial net primaryproduction from 1982 to 1999. Science 300:1560–1563.

Ogaya, R., and J. Penuelas. 2003. Phenological patterns ofQuercus ilex, Phyllirea latifolia and Arbutus unedo growingunder a field experimental drought. Ecoscience 11:262–270.

Paruelo, J. M., E. G. Jobbagy, and O. E. Sala. 2001. Currentdistribution of ecosystem functional types in temperate SouthAmerica. Ecosystems 4:683–698.

Penuelas, J., F. Lloret, and R. Montoya. 2001. Severe droughteffects on Mediterranean woody flora in Spain. ForestScience 47:214–219.

Pfisterer, A. B., and B. Schmid. 2002. Diversity-dependentproduction can decrease the stability of ecosystem function-ing. Nature 416:84–86.

Pons, X. 2000. Miramon. Sistema d’Informacio Geografica isoftware de Teledeteccio. CREAF, Bellaterra, Spain.

Sala, A., and J. D. Tenhunen. 1994. Site-specific water relationsand stomatal response of Quercus ilex in a Mediterraneanwatershed. Tree Physiology 14:601–617.

Sanchez, E., C. Gallardo, M. A. Gaertner, A. Arribas, and M.Castro. 2004. Future climate extreme events in the Mediter-ranean simulated by a regional climate model: a firstapproach. Global and Planetary Change 44:163–180.

Suarez, M. L., L. Ghermandi, and T. Kitzberger. 2004. Factorspredisposing episodic drought-induced tree mortality inNothofagus: site, climatic sensitivity and growth trends.Journal of Ecology 92:954–966.

Tarpley, J. P., S. R. Schneider, and R. L. Money. 1984. Globalvegetation indices from NOAA-7 meteorological satellite.Journal of Climate and Applied Meteorology 23:491–494.

Terradas, J., R. Salvador, J. Vayreda, and F. Lloret. 2004.Maximal species richness: an empirical approach forevaluating woody plant forest biodiversity. Forest Ecologyand Management 189:241–249.

Thorntwaite, C. W. 1948. An approach toward a rationalclassification of climate. Geographical Review 38:55–94.

Tilman, D., and J. A. Downing. 1994. Biodiversity and stabilityin grasslands. Nature 367:363–365.

Tilman, D., C. Lehman, and K. Thompson. 1997. Plantdiversity and ecosystem productivity: theoretical consider-ations. Proceedings of the National Academy of Science(USA) 94:1857–1861.

Tucker, C. J., and P. J. Sellers. 1986. Satellite remote-sensing ofprimary production. International Journal of Remote Sens-ing 7:1395–1416.

Voss, R., W. May, and E. Roeckner. 2002. Enhanced resolutionmodelling study an anthropogenic climate change: changes inextremes of the hydrological cycle. International Journal ofClimatology 22:755–777.

Williams, M., and E. B. Rastetter. 1999. Vegetation character-istics and primary production along an arctic gradienttransect: implications for scaling-up. Journal of Ecology 87:885–898.

Wylie, B. K., D. A. Johnson, E. Laca, N. Z. Saliendra, T. G.Gilmanov, B. C. Reed, L. L. Tieszan, and B. B. Worstell.2003. Calibration of remotely sensed, coarse resolutionNDVI to CO2 fluxes in a sagebrush–steppe ecosystem.Remote Sensing of Environment 85:243–255.

Zhou, L., C. J. Tucker, R. K. Kaufmann, D. Slayback, N.Shabnov, and R. B. Myneni. 2001. Variations in northernvegetation activity inferred from satellite data of vegetationindex during 1981 to 1999. Journal of Geophysical Research106(D17):20069–20083.


Education

WOODY PLANT RICHNESS AND NDVI RESPONSE TO DROUGHT EVENTS IN CATALONIAN (NORTHEASTERN SPAIN) FORESTS - Francisco Lloret