**4.1. Chlorophyll content based on SPAD indices**

Models 1– 7 shown in **Table 3** were applied to the SPAD-502 chlorophyll meter readings from the tropical forest study sites and the descriptive statistics of the estimates are shown in **Table 3**.


**Table 3.** Descriptive statistics of leaf chlorophyll content (µg cm−2) based on seven published SPAD-502 chlorophyll meter models.

**Figure 5** illustrates the chlorophyll content estimations for each model, its average values across models, and the confidence interval of 95% for the binned SPAD-502 readings. Estimations for the first six bins (range 15–80 SPAD-502 index) reported similar values. Average values at the higher SPAD index bin (80–95) show increase differences between models.

**Figure 5.** (a) ASD Hand Held 2 spectrophotometer (b) plant probe + leaf clip.

### **4.2. Chlorophyll content based on reflectance indices**

**3.9. Vegetation indices from satellite images: MTCI index**

62 Tropical Forests - The Challenges of Maintaining Ecosystem Services while Managing the Landscape

chlorophyll content in the tropical forest.

**4.1. Chlorophyll content based on SPAD indices**

**4. Results**

meter models.

**3**.

USGS EO-1 Hyperion image was that acquired on 15 February 2005. Hyperion data have a spatial resolution of 30 m2 with each pixel covering the spectral range, 400–2500 nm. A single image is 7.65 km wide (cross-track) by 185 km long (along-track) covering the study sites 1 and 2 (secondary disturbed sites). After atmospheric and radiometric corrections, see more details in [62], MTCI index was derived to assess from space the main impacts of land use changes on

Models 1– 7 shown in **Table 3** were applied to the SPAD-502 chlorophyll meter readings from the tropical forest study sites and the descriptive statistics of the estimates are shown in **Table**

**Model Reference Max. Min. Mean SD** [35] 292.83 11.62 67.02 39.70 [15] 203.45 13.48 72.16 28.30 [31] 191.73 10.06 72.82 30.21 [33] 150.27 9.71 62.43 23.21 [28] 154.04 13.48 69.46 23.10 [28] 194.22 10.31 72.78 29.87 [28] 187.56 10.18 70.80 28.82

**Table 3.** Descriptive statistics of leaf chlorophyll content (µg cm−2) based on seven published SPAD-502 chlorophyll

at the higher SPAD index bin (80–95) show increase differences between models.

**Figure 5.** (a) ASD Hand Held 2 spectrophotometer (b) plant probe + leaf clip.

**Figure 5** illustrates the chlorophyll content estimations for each model, its average values across models, and the confidence interval of 95% for the binned SPAD-502 readings. Estimations for the first six bins (range 15–80 SPAD-502 index) reported similar values. Average values

Reflectance indices and their respective models were applied to the reflectance spectra to the samples collected for this study. The resulting descriptive statistics are shown in **Table 4**. Most of the mean chlorophyll estimations are lower than their counterpart based on SPAD-502 index.


**Table 4.** Descriptive statistics of chlorophyll concentration (µg cm**<sup>−</sup>**<sup>2</sup> ) from the reflectance models based on the spectroradiometer data.

**Figure 6.** Estimated chlorophyll content for each SPAD-502 calibration model applied to the total samples of our dataset. The black line represents the average value across models and its confidential interval of 95% for the binned SPAD-502 readings.

**Figure 6** illustrates the estimations of chlorophyll content for each reflectance model. It includes the average values across models and the 95% confidence interval for the binned SPAD-502 readings. It is interesting to observe that chlorophyll estimations become insensitive for SPAD reading greater than 80.

#### **4.3. Comparison between the three methods for chlorophyll estimation**

**Figure 7** shows the comparison between average chlorophyll estimations from the three methods used in this study. Estimations until bin 50–60 are relatively similar. Estimation from SPAD then increased exponentially while estimations from reflectance and PROSPECT model are close to each other until bin 70–80, differences then increased since the asymptotic behavior of reflectance models estimations.

**Figure 7.** Average chlorophyll content estimates from five reflectance models (errors bars at 1.96 standard deviations) compared to estimated ground truth chlorophyll content based on SPAD-502 chlorophyll meter readings (error bars at 1.96 standard deviations).

**Figure 8.** Comparison of average chlorophyll content estimates from the SPAD-502 chlorophyll meter index and the averages of all spectroradiometer-based chlorophyll estimates (error bars at 1.96 standard deviations).

**Figure 8** illustrates the comparison of average chlorophyll content estimates from the SPAD-502 chlorophyll meter index and the averages of all spectroradiometer-based chlorophyll estimates. **Figure 9** presents the correspondent boxplots for the three approaches used in this study.

**Figure 9.** Boxplots of the three estimation of chlorophyll content (outliers not included).

**4.3. Comparison between the three methods for chlorophyll estimation**

64 Tropical Forests - The Challenges of Maintaining Ecosystem Services while Managing the Landscape

of reflectance models estimations.

1.96 standard deviations).

in this study.

**Figure 7** shows the comparison between average chlorophyll estimations from the three methods used in this study. Estimations until bin 50–60 are relatively similar. Estimation from SPAD then increased exponentially while estimations from reflectance and PROSPECT model are close to each other until bin 70–80, differences then increased since the asymptotic behavior

**Figure 7.** Average chlorophyll content estimates from five reflectance models (errors bars at 1.96 standard deviations) compared to estimated ground truth chlorophyll content based on SPAD-502 chlorophyll meter readings (error bars at

**Figure 8.** Comparison of average chlorophyll content estimates from the SPAD-502 chlorophyll meter index and the

**Figure 8** illustrates the comparison of average chlorophyll content estimates from the SPAD-502 chlorophyll meter index and the averages of all spectroradiometer-based chlorophyll estimates. **Figure 9** presents the correspondent boxplots for the three approaches used

averages of all spectroradiometer-based chlorophyll estimates (error bars at 1.96 standard deviations).

**Figure 10.** Scatter plots, histograms, and Pearson correlation between three chlorophyll estimations (SPAD, reflectance, and PROSPECT) and MTCI index and REP.

**Figure 10** shows the correlations between the three chlorophyll estimations (SPAD-502, reflectance, and PROSPECT) applied in this study. Additionally, correlations with MTCI and REP are presented. Pearson correlation demonstrates a strong correspondence between the three methods calculated at leave level (SPAD-502, reflectance indices, and PROSPECT). Chlorophyll content estimates by the second-order polynomial based on SPAD-502 models and reflectance models agree in 0.76 while SPAD-502 models and PROSPECT agreed in 0.71. The lowest correlation (*r* = 0.67) is presented by estimations from reflectance models and PROSPECT model despite the fact that both methods are estimated from reflectance measurements. A strong correlation between them was found. MTCI and SPAD-502 correlate in 0.74, MTCI and reflectance models correlate in 0.88, and MTCI and PROSPECT correlate in 0.69. Correlation coefficients between REP and SPAD-502 model, reflectance models, PROS-PECT, and MTCI are 0.66, 0.81, 0.59, and 0.87, respectively.

**Figure 11** shows the estimations of leaf chlorophyll content based on SPAD index, MTCI and Ratio of derivatives. For the first two methods, chlorophyll content in the oil spill is significantly lower compared to the non-polluted sites.

**Figure 11.** (a) SPAD chlorophyll index for the three study sites; (b) MERIS terrestrial chlorophyll index and (c) REP red-edge position-first derivatives for the three study sites.

#### **4.4. Chlorophyll content evaluation**

SPAD 502 chlorophyll content index and REP index were estimated for the three study sites. The results from **Figure 10(a)** and **(c)** shows that chlorophyll content was significantly lower (99.9%) at the secondary forest affected by pollution (Site 1) which allow us to conclude that forest degradation at local level can be detected using a portable chlorophyll content instrument. On the other hand, MTCI index derived from the satellite image also shows significantly lower values in the Site 1 **(Figure 10b)**, which confirm that chlorophyll content is a suitable indicator of land uses changes, and it can be applied at regional level to detect forest degradation caused by land use changes in the tropical forest.

MTCI index at regional level was computed using the Hyperion satellite images of the area corresponding to Site 1 and Site 2. **Figure 12** illustrates the results. Lower levels of chlorophyll (less than four) are found around the petroleum facilities and routes. On the other hand, higher levels of chlorophyll content (more than four) were found in areas still covered by the secondary forest.

Detection of Amazon Forest Degradation Caused by Land Use Changes http://dx.doi.org/10.5772/65493 67

0.69. Correlation coefficients between REP and SPAD-502 model, reflectance models, PROS-

**Figure 11** shows the estimations of leaf chlorophyll content based on SPAD index, MTCI and Ratio of derivatives. For the first two methods, chlorophyll content in the oil spill is significantly

**Figure 11.** (a) SPAD chlorophyll index for the three study sites; (b) MERIS terrestrial chlorophyll index and (c) REP

SPAD 502 chlorophyll content index and REP index were estimated for the three study sites. The results from **Figure 10(a)** and **(c)** shows that chlorophyll content was significantly lower (99.9%) at the secondary forest affected by pollution (Site 1) which allow us to conclude that forest degradation at local level can be detected using a portable chlorophyll content instrument. On the other hand, MTCI index derived from the satellite image also shows significantly lower values in the Site 1 **(Figure 10b)**, which confirm that chlorophyll content is a suitable indicator of land uses changes, and it can be applied at regional level to detect forest degra-

MTCI index at regional level was computed using the Hyperion satellite images of the area corresponding to Site 1 and Site 2. **Figure 12** illustrates the results. Lower levels of chlorophyll (less than four) are found around the petroleum facilities and routes. On the other hand, higher levels of chlorophyll content (more than four) were found in areas still covered by the secon-

PECT, and MTCI are 0.66, 0.81, 0.59, and 0.87, respectively.

66 Tropical Forests - The Challenges of Maintaining Ecosystem Services while Managing the Landscape

lower compared to the non-polluted sites.

red-edge position-first derivatives for the three study sites.

dation caused by land use changes in the tropical forest.

**4.4. Chlorophyll content evaluation**

dary forest.

**Figure 12.** MTCI index computed from the Hyperion Satellite images of the study area of Site 1 and Site 2.

**Figure 13.** Comparison of three generalized models derived from SPAD-502 readings. The second-order polynomial model proposed in this study (black line), the homographic model proposed by Cerovic et al. [36] (dotted line), and the homographic model proposed by Coste et al. (2012) for trees from the Amazon forest.

Based on the results of the seven SPAD-502 published calibration models we compute their average in order to obtain a general model for chlorophyll content estimation which accomplish for a wide range of vegetation species and physiological stage. The resulting general model is a second order polynomial in a range of 15 to 95 SPAD index readings. This general model is proposed as ground truth chlorophyll which is assessed by comparing it to a reference published generalized model based on SPAD-502 readings and traditional methods in a laboratory. The first reference model is a homographic model proposed by Cerovic et al. (2012) and computed from seven (polynomial, exponential and homographic) models applied to a variety of plant species. The second model is the generalised homographic model for tropical trees proposed by Coste et al. (2010) which was discussed before as Model 2 in **Table 1**. **Figure 13** illustrates the comparison of the three models.
