Herbivory Rate on Woody Species of the Caatinga and NDVI as Indicators of Plant Stress

Especies de plantas distribuidas em uma paisagem sao submetidas a um mosaico de condicoes abioticas que podem ter efeito negativo sobre o desenvolvimento (stress geracao) e expo-las a predacao por herbivoros. Esse estresse pode causar adicionalmente assimetria foliar e uma reducao na producao primaria. A taxa fotossintetica, relacionada com a produtividade da planta, pode ser medida por indices espectrais, tais como o NDVI (indice de vegetacao da diferenca normalizada), calculado a partir de imagens de satelite. No presente trabalho, testou-se a hipotese de que ambientes com baixa produtividade primaria (NDVI baixo) ira possuir maior assimetria foliar e maiores taxas de herbivoria. Os resultados mostram que na regiao de Caatinga semi arida de Pernambuco, Brasil, a folha de assimetria diminui com valores mais elevados de NDVI, indicando uma estreita relacao entre esta medida da planta e o indice espectral. Por outro lado, a correlacao entre herbivoria e producao primaria ou assimetria foliar nao foi significativa, sugerindo que os herbivoros vao alem da simples selecao de individuos mais estressados. Palavras-Chave: Assimetria flutuante, herbivoria, NDVI Taxa de Herbivoria em Especies Arboreas da Caatinga e o Uso do Indice de Vegetacao por Diferenca Normalizada (NDVI) como Indicador de Estresse em Planta ABSTRACT Plant species distributed on a landscape are submitted to a mosaic of abiotic conditions that may have a negative effect on development (generating stress) and expose them to predation by herbivores. This stress can cause additionally leaf asymmetry and a reduction on primary production. The photosynthetic rate, related to plant productivity, can be measured by spectral indexes, such as the NDVI (normalized difference vegetation index), calculated from satellite images. In the present work, we test the hypothesis that environments with low primary productivity (low NDVI) will possess larger leaf asymmetry and higher herbivory rates. Our results show that in the Caatinga semi-arid region of Pernambuco, Brazil, the leaf asymmetry reduces with higher NDVI values, indicating a close relationship between this plant measure and the spectral index. On the other side, the correlation between herbivory and primary production or leaf asymmetry was not significant, suggesting that herbivores go beyond just selecting more stressed individuals. Keywords: Leaf asymmetry, NDVI, herbivory


Introduction
Several studies show that there is a high correlation between part of the incident solar radiation on plants and their physiologic state (Méndez-Barroso et al., 2008;Wessels et al. 2004).Normally vegetation has low reflectance on the visible red part of the spectrum, and high on the green and near infrared -as the wavelength of the former is absorbed by plants for photosynthetic activity (Weier and Herring, 1999).In this way, the fraction of the photosynthetic active radiation (PAR) captured by the photosynthesizing tissues is, among other factors, related to the productivity of a vegetated area (Hill and Donald, 2003).
The ability to translate spectral data to meaningful biological variables is a key step in the increase of use and value of information collected by satellites (Paruelo et al., 1997).
Based on this statement, several indexes were developed involving algebraic operations on satellite image bands.There is substantial evidence that PAR is related to some of these vegetation indexes, although in several cases pixel heterogeneity, atmospheric distortions and solar angle can degrade the quality of the data (Myneni and Williams, 1994).The most used of these indexes is the Normalized Difference Vegetation Index (NDVI), which was developed and proven to be indicator of photosynthetic activity, even for heterogeneous areas (Gamon et al., 1995), different scales (Stefanov and Netzband, 2005) and different ecosystems (Myneni and Williams, 1994;Paruelo et al., 1997).
The NDVI has been used as a sensible measure for disturbed area identification (Weier and Herring, 1999), varying in relation to the aridity of the environment (Barbosa et al. 2006), and according to some authors (White, 1974;Rhoades, 1979), plants that suffer stress of any sort must be more susceptible to predation by herbivores than less stressed individuals, although higher productivity (an aspect of a low stress plant) can lead to higher herbivory rates (Carmo and Penedo, 2004).Stress is defined as any environmental factor that can cause potential damage to the biological system (Hoffman and Parsons, 1991).
One of the indicators of plant stress and thus vulnerability to herbivory (e.g., Moller, 1995;Wiggins, 1997;Zvereva et al., 1997) is the level of leaf asymmetry (LA), which represents a small random deviation from the bilateral symmetry in a morphologically symmetrical leaf (Van Valen, 1962;Palmer and Strobeck, 1986;Bjorksten et al., 2000).

Study Area
The study was carried out between the 15 and 16 th of march 2007, in a Caatinga area of the Catimbau Valley National Park (8º 32' 27''S; 37º 14' 51''W), municipality of Buíque, Pernambuco state, Brazil, situated 285km from Recife (Figure 2).The closest meteorological station of the area is located on the town of Buíque and recorded an annual average air temperature and rainfall of 25ºC and 1,095.9mm, respectively, with the highest rainfall rates occurring during the months of April to June (SUDENE, 1990).

Data Sampling
In the study area, two trails were

Leaf Asymmetry
With the help of a precision ruler of 0,1mm, measures were executed on the wider part of the leaves, from the central nerve to the edge on the right and left sides.To determine the Leaf Asymmetry (LA) of each plant an index was calculated according to (2), along with a correction, in case the asymmetry is sizedependent (Cornelissen and Stiling, 2005).
Ri: Right leaf measure.Li: Left leaf measure.

Data Analysis
We verified the existence of natural LA in the plant species.This step is necessary because many plants exibit natural leaf asymmetry.We applied the Kolmogorov-Smirnov test comparing right and left measure averages in each species (Palmer and Strobeck, 1986).The hypothesis is only viable to test on plants that do not exhibit natural asymmetry.
Next, we tested using Kruskal-Wallis and post-hoc Dunn tests to verify differences in asymmetries and average herbivory rates between species.This step was taken to test for natural differences on herbivory and asymmetry between species.
Finally, using the Spearman correlation test, we verified if the NDVI, LA and herbivory who correlated among themselves.

Results
Herbivory caused by chewing insects differed significantly between plant species (Kruskal-Wallis, H=44.717; p=0.001) (Table 1 and Figure 4).In crescent order, Tocoyena formosa and Bauhinia pentandra demonstrated higher rates of herbivory, while Passiflora luetzelburgii and Croton campestris suffered a lower herbivory.In the study area, 62% of the sampled leaves were undamaged, and 22% had up to 5% of damage (Figure 5).The test for LA showed a slight tendency for natural asymmetry in the analyzed plant species.Even with this result, we still thought it would be interesting to test LA with the correlations.The LA index between species did not differ significantly (Kruskal-Wallis, H=9.4961, p=0.3022) (Table 1).Next, we performed correlation analysis (5% significance) comparing two on two the indexes (NDVI, LA and Herbivory).In the correlation test between NDVI and LA, we found that out of the nine analyzed species, only Tocoyena formosa and Bauhinia pentandra were not significantly correlated (Figure 6 and Table 2).The inverse was found when we correlated Herbivory and NDVIthe two cited species where the only ones to correlate significantly.

Discussion
Out of the nine analysed species in the transect, seven presented significant correlations of NDVI with LA.The NDVI is known to be an indicator of photosynthetic activity (Myneni and Williams, 1994), with good efficiency for analysis in drier climates (Liesenberg et al., 2007) and the correlations indicated that individuals with higher LA values were located in areas with lower photosynthetic rates (as measured by the NDVI).
Therefore, using the NDVI index as a reference, we were able to identify environments less favourable to plant growth.Although several studies suggest the LA index as one of the measures of plant stress and thus vulnerability to herbivory, (i.e., Moller, 1995;Wiggins, 1997;Zvereva et al., 1997), this relation is not so evident, and each species can present specific responses (Bañuelos et al., 2004).
Another explanation for the lack of significant differences between the studied species is that they respond in a similar manner to environmental conditions.
Our hypothesis that LA and herbivory index were positively correlated between themselves and negatively correlated with NDVI was only partially corroborated (Figure 7).The relationship between LA and the herbivory index was inconclusive, and the correlation between the herbivory index and NDVI was only significant for one species, Tocoyena formosa.The statistics tests for the correlation between NDVI and LA were significant for 7 out of the 9 species, making this element of the hypothesis system the most likely to be correlated in future studies.
Besides from differing in relation to other species with respect to NDVI-LA correlation, the Tocoyena formosa and Bauhinia pentandra were also the taxons with highest rates of herbivory.This fact may shed a light in this distinctive patternthe LA increases its value in very high peaks of herbivory (Bañuelos et al., 2004).In this way, high values of herbivory could mask the relationship between LA and NDVI.Another possibility is that there is not in fact any relation between LA and stress in these two species, as found in some recent studies (Lens et al., 2002;Siikamäki et al., 2002).More detailed studies are needed to elucidate these patterns.
In the studied Caatinga area, the plants differed significantly in relation to herbivory rate.As this ecosystem has a wide variety of biotic (species richness, population density and distribution) and abiotic (light, temperature, humidity, precipitation, soil, relief, etc.) factors, this result can stem from the large number of specific interactions between plants and herbivorous insects (Lowman, 1985;Coley and Aide, 1991).
In this study the NDVI provided a rather interesting quantification of the environment Zvereva, E., Kozlov, M., Haukioja, E. (1997).
Stress responses of Salix borealis to pollution and defoliation.Journal of Applied Ecology 34, 1387-1396.

Figure 1 .
Figure 1.Hypothesis system of this work.The connectors with (-) indicate a negative expected correlation, and connectors with (+) indicate a positive expected correlation.

Figure 2 .
Figure 2. Delimitation of the Catimbau Valley National Park (green), Pernambuco, Brazil, with emphasis to the study area (red), and the Landsat 7 ETM+ image from 12/01/2001 in the background.The map units are UTM, zone 24 south, georreferenced to the South America 69 datum.

Figure 3 .
Figure 3. NDVI processing of the study area, indicating the sampling points through the trails.The trail 1 was sampled on 03/15/2007, and trail 2 in the next day.The bright areas indicate an index closest to +1 (highest photosynthetic activity) and the dark areas indicate an index close to -1 (lowest photosynthetic activity).The images on the lower half of the figure were taken from Google Earth (http://earth.google.com)for illustrative purposes

Figure 4 .Figure 5 .
Figure 4. Mean Damage Index (Garcia-Gusman & Dirzo, 2001) of plant species in a Caatinga area, Catimbau Valley National Park, Pernambuco Brazil productivity, being probably related to plant stress.The results suggest the NDVI as a long range habitat stress indicator, in this way assisting natural reserve management and the definition of priority areas for conservation.a plantation of Pinus radiata in New Zealand.an African silk moth Imbrasia belina.Oikos 79, 484-488.

Table 1 .
Sampled plant species in a Caatinga area, Catimbau Valley National Park, Pernambuco, Brazil.Averages of herbivory and LA indexes, observed between the 14 th and 15 th March 2007 Correlation plots of NDVI vs. LA index for the species a. Croton sonderianus, b.

Table 2 .
Correlations between the NDVI and LA index in an area of Caatinga, Catimbau Valley, Pernambuco, Brazil.N.S.:Not significantly correlated.Result of the hypothesis tests for this study.