Introduction 

Chlorophylls are pigments responsible for capturing light for photosynthesis and are essential for the conversion of light radiation to chemical energy ([25]). Several studies showed that variations in chlorophyll content are related to plant photosynthetic capacity, development stage and environmental stresses ([17], [16], [23], [12], [3], [7]). Therefore, chlorophyll content is an important indicator of photosynthetic efficiency, carbon uptake, and fitness of plants and ecosystems ([26], [11]).

Spectrophotometry is the most conventional and precise technique for the determination of leaf chlorophyll content. It requires tissue homogenization with an organic extraction solvent, such as acetone or dimethyl sulfoxide (DMSO), and a spectrophotometer for the direct quantification of pigments ([19], [35]). Nonetheless, this technique is destructive and usually has to be performed under laboratory conditions. Other disadvantages of this technique are the low number of samples that can be processed per day and the necessity of transporting leaves from the field to the laboratory, which could interfere with the results, since more delicate leaves may suffer dehydration and pigment oxidation after harvesting.

With the development of optical instrumentation, the estimation of chlorophyll content has become easier, faster, nondestructive, and can be performed in the field. Currently, with the use of hand-held chlorophyll meters and portable reflectometers, it is possible to analyze a large number of samples in a single day and to measure the same leaf several times as it grows and develops. Hand-held chlorophyll meters are often calibrated to a single index and measure the absorption of electromagnetic radiation in the red and infrared regions ([21]). Red light is strongly absorbed by chlorophyll, whereas near-infrared light is used as an internal reference wavelength to adjust for differences in leaf structure. Conversely, portable reflectometers measure the radiation reflected by the leaf at a wide range of wavelengths ([28], [31]). Methods based on the reflection of electromagnetic radiation by leaves originated from remote sensing technology and are based on vegetation indices ([10], [9], [28], [31], [2]). Conceptually, vegetation indices are mathematical transformations of spectral bands used to estimate the contribution of vegetation in multispectral observations. The equations of most vegetation indices are mainly derived from bands of red and near infrared, such as the normalized difference vegetation index (NDVI) at an ecosystem scale ([26], [2]), or the chlorophyll normalized difference index (ND₇₀₅) at a leaf scale ([10], [28]).

Hand-held chlorophyll meters and portable reflectometers are more convenient than extraction methods because they offer a more rapid, efficient, and nondestructive assessment of the leaf chlorophyll content. Nevertheless, optical methods also have limitations. First, they do not provide absolute chlorophyll content, but only relative indices that express the relative chlorophyll content ([28], [2]). Thus, the indices need to be previously calibrated with a more precise extraction method in the laboratory to provide absolute values of chlorophyll content. There are several mathematical models available in the literature for the calibration of hand-held absorbance chlorophyll meters. However, most of the available models are species specific and developed either for crop species ([32], [30], [4]) or temperate and sub-tropical forests species ([5], [34], [27], [13]). There are limited studies on the calibration of nondestructive methods for estimating leaf chlorophyll content in tropical tree species ([20], [8], [23]). Additionally, most of the available models are restricted to an estimation of total chlorophyll (Chl t) ([20], [8], [24]) and there are limited models for the estimation of chlorophyll a (Chl a) and b (Chl b - [28], [23]). Estimation of Chl a and b is important for the calculation of the ratio Chl a/b, which is an indicator of photosynthetic performance ([17]) and frequently used in studies on light acclimation capacity of seedlings when transferred from the nursery to the field in forest restoration programs ([12], [3]). Another limitation of optical methods refers to the physical properties of instruments that can result in non-linear relationships between chlorophyll indices and absolute chlorophyll content ([28], [31], [33], [32]). For instance, high leaf mass per area (LMA) may lead to the underestimation of chlorophyll content because high leaf thickness is usually associated with high mutual chloroplast shading and high overlapping of optical properties from distinct cell constituents that can affect the reflectance captured by the instrument ([18]). Heterogeneity in chlorophyll distribution may also create bias, from different leaf anatomical structures in distinct species, leaf developmental stages, and leaf vein networks. Finally, hand-held absorbance chlorophyll meters may lose sensitivity at high chlorophyll contents because of saturation of absorption in the red range ([32]). Therefore, the development of standard models that account for different leaf morphologies to be applied to different species and environmental conditions is necessary.

Estimation of leaf chlorophyll content of tree seedlings in areas of high biodiversity, such as tropical rainforests, that contain many species with different biochemical and anatomical structures, still represents a challenge for researchers. Coste et al. ([8]) developed a homographic model for the calibration of the SPAD-502 index to estimate chlorophyll content in a group of 13 Neotropical tree species with a wide range of LMA and chlorophyll contents. The developed homographic model was more accurate than the linear, polynomial, and exponential models and provided accurate estimates at high and low values of chlorophyll content. However, Coste et al. ([8]) did not calibrate their model for other chlorophyll indices, and did not test the model during initial tree development. Therefore, we conducted this study to test whether the multi-species model proposed by Coste et al. ([8]) is suitable for the calibration of the hand-held chlorophyll absorbance meters SPAD-502 and ClorofiLog, and of the leaf reflectance index ND₇₀₅ ([10]) for the estimation of Chl a, b, and t in the leaves of seedlings at the nursery stage of six Neotropical rainforest tree species.

  Materials and methods 

Data were collected from seedlings of six Neotropical rainforest tree species representing six different botanical families (Tab. 1). The seedlings were grown under 50% artificial shading, in the nursery of Floresta Viva Institute (⇒ http:/­/­www.­florestaviva.­org.­br/­), Serra Grande, Uruçuca, Bahia, (Brazil). When seedlings were 2-4-months old, they were transported to the Plant Physiology Laboratory of the Universidade Estadual de Santa Cruz (UESC), Ilhéus, Bahia (Brazil), shortly before measurements. In the laboratory, we collected 30 to 37 fully expanded leaves from each species, with a total of 200 leaves. All the leaves collected had preserved edges and no sign of herbivory or disease symptoms. We measured the leaf absorbance, leaf reflectance, Chl a, b, and t contents, leaf area (LA), and leaf mass per area (LMA).

Leaf absorbance and reflectance were measured immediately after their collection. Leaf absorbance was measured using the handheld chlorophyll absorbance meters ClorofiLog^® (Falker, Porto Alegre, Brazil) and SPAD-502^® (Minolta Inc., Osaka, Japan), for the determination of the Falker chlorophyll index (FCI) and the SPAD index (SPADi), respectively. Leaf reflectance of the adaxial leaf surface was measured with an USB4000-UV-VIS spectrometer using a LS-1 tungsten-halogen light source (Ocean Optics Inc., Dunedim, FL, USA). Based on the values of reflectance at 705 nm (R₇₀₅) and 750 nm (R₇₅₀), the normalized difference reflectance index was calculated as ND₇₀₅ = (R₇₅₀ - R₇₀₅)/(R₇₅₀ + R₇₀₅) ([10]). We performed one measurement per leaf for each device. Measurements were always performed in the same place, at the middle of the leaf blade.

After absorbance and reflectance measurements, leaves were digitalized with an HP 2400 scanner. The digitalized images were saved in a TIFF format and the LAs were calculated using the ImageJ software (⇒ http:/­/­rsb.­info.­nih.­gov/­ij/­). Then, three leaf-discs (6 mm diameter) were collected from the same area where the absorbance and reflectance measurements were taken. The leaf discs were placed in Eppendorf tubes with 5 mL dimethyl sulfoxide (DMSO) saturated with CaCO₃ ([14]). The tubes were incubated at 65 °C for 24 h, for the total extraction of leaf pigments. An aliquot of 3 ml of the extract was used to measure the pigment absorbance at 649 nm and 665 nm, using a V-1100D^® spectrophotometer (Mesu Lab Enterprise Co., Ltd, China). Chl a and b contents were calculated according to Wellburn ([35]). Finally, the leaves were dried at 65 °C until a constant mass. The dry mass and the leaf area were used for the calculation of LMA for each individual leaf.

Parameters of the homographic model Chl = (α · index)/(β - index) ([8]) were calibrated for chlorophyll contents (Chl a, b, and t) and chlorophyll indices (ND₇₀₅, FCI and SPADi) using the entire data set (n = 200). The relative root mean square calibration errors (RMSε_c, %) were calculated for each of the six rainforest tree species cited in Tab. 1 (n = 30-37), based on chlorophyll values estimated by the multi-species models and on the chlorophyll values measured by the indices. RMSε_c was used as an indicator of the accuracy of predictive models and as a comparison with data from the literature ([28]).

  Results 

The average values for LA among the studied species ranged between 14.2 and 29.5 cm², and for LMA, values ranged within 34.8 and 98.9 g m^-2 (Tab. 2). The lowest and highest values for LA and LMA were found in T. guianensis and B. rubescens, and in C. myriantum and C. brasiliense, respectively. The average chlorophyll content obtained via traditional DMSO extraction methods ranged between 142.3-411.6 mg m^-2 for Chl t, 98.1-270.1 mg m^-2 for Chl a, and 44.2-141.5 mg m^-2 for Chl b. The lowest and highest mean values for Chl t, b, and a were found in C. myrianthum and B. rubescens, respectively. The lower and higher individual values of Chl t contents for each of the six species used in this study are represented by the spectral properties of the leaves in Fig. 1.

Fig. 1 - Spectral properties of seedlings of six Neotropical rainforest tree species. The percentage of the irradiance reflected by the adaxial leaf surface was obtained with a spectrometer USB4000-UV-VIS using a LS-1 tungsten-halogen light source (Ocean Optics Inc. Dunedim, FL, USA). Dotted and continuous lines correspond to the lowest and the highest values of total chlorophyll content (mg m^-2) for single-leaf measurements of each species, respectively, determined by dimethyl sulfoxide extraction and spectrophotometric measurements. (A): B. rubescens; (B): C. brasiliense; (C): C. myrianthum; (D): E. macrophylla; (E): I. capitata; (F): T. guianensis.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

A strong linear relationship was observed between the two handheld chlorophyll meters, suggesting that data obtained from ClorofiLog (FCI) can be easily converted into SPADi (Fig. 2). Relationships between chlorophyll content and the nondestructive methods tested in this study were all curvilinear (Fig. 3) and the R² values obtained for the calibrated homographic models were above 0.8 (Tab. 3). For all methods, the highest values of R² were observed for Chl a, followed by Chl t and b. Although the R² values were very similar among the three methods, the best fit was obtained for the data from ND₇₀₅ (R² ≥ 0.86) followed by SPAD-502 (R² ≥ 0.83), and ClorofiLog (R² ≥ 0.82).

Fig. 2 - Relationships between two non-invasive chlorophyll indexes based on leaves of seedlings of six Neotropical rainforest tree species (n = 200). Dotted line represent the 1:1 relationship between FCI and SPADi.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Fig. 3 - Relationships between chlorophyll contents and noninvasive chlorophyll indexes. Parameters of the homographic models based on leaves of seedlings of six Neotropical rainforest tree species (n = 200) are shown in Tab. 3.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

For all species and indices, the calculated values of RMSε_c were below 20% (Tab. 4). Among species, the lowest values of RMSε_c were observed in C. myrianthum and E. macrophylla, whereas the highest value of RMSε_c was found in T. guianensis. The ND₇₀₅ index had the lowest average RMSε_c values for Chl t, a, and b, followed by SPADi and FCI. The lowest and highest values of RMSε_c for Chl a and b were found in C. myrianthum and E. macrophylla (ND₇₀₅), and T. guianensis (FCI), respectively. The lowest and highest values of RMSε_c for Chl t were found in E. macrophylla (ND₇₀₅) and T. guianensis (FCI), respectively.

  Discussion 

The results from our study show that the homographic model proposed by Coste et al. ([8]) is suitable for calibration of the relationship between absolute chlorophyll content and the chlorophyll indices ND₇₀₅, SPADi, and FCI. As expected, there is a tendency towards a non-linear relationship between chlorophyll content and the chlorophyll indices ([28], [31], [33]). The nonlinear response may reflect the physical limitation of optical methods at higher chlorophyll concentrations ([21], [32], [30]), and the non-uniform distribution of chlorophyll in the leaf blade, particularly when there is high chlorophyll density in chloroplasts ([4]). Nonetheless, good fits can be obtained for Chl a (R² ≥ 0.91), t (R² ≥ 0.89), and b (R² ≥ 0.82). A greater fit was observed for Chl a in all indices, which is particularly important because Chl a is always present in a greater amount than Chl b in the light harvesting complexes, thus having a significant contribution to Chl t ([25]).

Both handheld chlorophyll meters used in the present study operate in a similar way, with small differences. Although SPAD-502 measures absorption from leaves at two wavelengths, 660 nm (red) and 940 nm (near-infrared), ClorofiLog measures the absorption at three wavelengths, 635 nm and 660 nm (red) and 880 nm (near infrared - [1], [29]). We expected ClorofiLog, which measures a third wavelength near the second absorption peak of Chl b (642 nm), to provide greater accuracy for chlorophyll estimates. However, the values of R² and RMSε_c for both devices were similar, which means that SPAD-502 and ClorofiLog provide equally accurate chlorophyll estimates. Additionally, the indices provided by these devices (FCI and SPADi) showed a strong linear relationship. This result has an important application. Most regression models available in the current literature predicting chlorophyll content in tropical trees have been developed using SPAD-502 ([20], [8], [23]). ClorofiLog is newer and is currently being used more frequently in some countries. Based on the results from our study, the data from ClorofiLog can now be safely converted to the SPAD index and used for the estimation of chlorophyll content using regression with the models already available in literature ([20], [8], [23]).

The α parameter of the model described by Coste et al. ([8]) for Chl t (1.171 mg m^-2) was similar to values found in our study for both chlorophyll meters, but the value of β (148.84) was lower than the values found in our study. Compared with the model reported by Coste et al. ([8]), our model would result in an underestimation of the chlorophyll content as the SPADi values increase. We attribute these discrepancies to differences in the maximum values of chlorophyll contents. In the present study, the maximum value of Chl t was 815 mg m^-2 (B. rubescens), while Coste et al. ([8]) reported maximum values of Chl t around of 1500 mg m^-2. The differences in chlorophyll content can be explained by differences in the ontogenetic stages of the plants, environmental conditions, and species used in the two studies. The plants used by Coste et al. ([8]) were around 2 years old, whereas the plants used in the present study were 2-4-months old. Different studies have shown that seedlings of tropical trees have lower nitrogen and chlorophyll contents per leaf area unit than adults ([16], [15]). Differences in chlorophyll content can also be explained by the distinct environmental conditions in which plants are grown. In the present study, seedlings were grown in nursery conditions, under 50% of full sun, whereas the plants studied by Coste et al. ([8]) were grown under three different light treatments: 5%, 10%, and 20% full sun. The reduction in light intensity induces an increase in chlorophyll content of leaves ([25]), which may explain the higher chlorophyll content reported in the study of Coste et al. ([8]). Although the maximum values of Chl t found in the present study were lower than the values reported by Coste et al. ([8]), they are in agreement with values reported for other tropical trees, for instance, Scleronema micranthum (600 mg m^-2), Swietenia macrophylla, and Ceiba pentandra (900 mg m^-2 - [20]). Additionally, the β values calculated individually for each of the 13 species ranged from 88.2 (Cecropia obtusa) to 299.2 (Eperua falcata), with a large dispersion of data in the SPADi range between 40 and 60, and chlorophyll content between 500 and 1000 mg m^-2 ([8]). Based on this, the model proposed by Coste et al. ([8]) can be applied to different indices and equipment, but the range of chlorophyll contents should be taken into account.

The average RMSε_c value calculated individually for each species is low compared with the values available in literature ([28], [8], [24]), indicating the high accuracy of our models. The average values of RMSε_c for Chl a, b, and t in the present study were lower than 10%. Richardson et al. ([28]), for example, reported values of RMSε_c for ND₇₀₅ varying from 11.9% (Chl a) to 15.2% (Chl b), and values of RMSε_c for SPADi from 15.6% (Chl b) to 20.3% (Chl a). The RMSε_c values lower than 10% prove the feasibility of the multi-species model for estimating Chl a and b from hand-held absorbance chlorophyll meters and leaf reflectance. Our highest value of RMSε_c was obtained for the estimation of Chl b in T. guianensis (18.7%). Yet, this value is similar to those reported by Richardson et al. ([28]) for estimates of Chl b in Betula papyrifera, using different hand-held chlorophyll meters and leaf reflectance indices (14.6% to 36.2%). It was also similar to values of RMSε_c (14.6% to 29.2%) calculated by models based on reflectance indices in Eugenia uniflora by Mielke et al. ([24]).

Neotropical rainforests are recognized for their high biodiversity ([22]) and increasing human pressures. The growth of human populations and changes in land use are the main causes of forest degradation and losses of biodiversity. Sustainable use of forest remnants and restoration of degraded forest ecosystems are challenges for scientists and civil organizations in the tropics ([6]). Therefore, the suitability of scientific technology for the improvement the production and plantation of seedlings of native tree species is crucial to the success of commercial forest plantations and ecosystems restoration programs ([12], [3]). In the present study, we showed the feasibility of the multi-species model proposed by Coste et al. ([8]) to estimate leaf chlorophyll content with the use of different instruments, aiming the application at seedling production in forest nurseries and monitoring the physiological performance of seedlings after introduction to reforestation areas. All indices tested in the present study showed excellent fits for Chl t, b, and for Chl a in particular. We also demonstrated that the SPADi and FCI indices can be safely interconverted.

  Acknowledgements 

The authors thank Gerson J. Sales Neto, Nilson A. dos Santos, and Rones F. Souza of Floresta Viva Institute, and M.Sc. Murilo F. C. de Jesus for assistance with data collection. We thank Dr. Fábio P. Gomes of DCB/UESC for providing the SPAD-502 used in this study. Funding for Daniela V. Silva during this study was provided by a scholarship from Capes (Brazilian Higher Education Council). Marcelo S. Mielke and Ediófila Brito-Rocha gratefully acknowledge fellowships provided by CNPq (Brazilian National Council for Scientific and Technological Development). This study was supported by CNPq (Proc. 561933/2010).

  References