**3.2 Sensitivity to forest structure and AGB**

Assuming that the constituted forest plots dataset is well distributed within the acquired scene(s), Fourier r-spectra can be computed for windows centred on each plot. For example, when applied to 1-m Ikonos (Proisy et al. 2007) or 0.5-m Geoeye panchromatic images (Fig. 9) r-spectra permit good discrimination of a wide array of canopy structures of mangroves (Fig. 9). Furthermore pre-adult, mature and decaying mangrove forests show contrasted signatures with dominant frequencies around 180, 80, 50 and 30 cycles per kilometre.

Inverting FOTO indices (the three first PCA axes) into AGB of forest plots distributed over two different sites (i.e. two different images) yielded good correlations and low errors, as presented in Fig. 10. Compared to estimations provided by the P-band HV polarisation channel, FOTO-derived AGB did not show saturations over the whole range of mangrove biomass (Fig. 9), i.e. up to 500 tDM.ha-1 and rmse error remains acceptable (33 tDM.ha-1). This result suggests that, in the case of closed canopies with sub-strata of low biomass (e.g. the mangrove ecosystem in French Guiana), the canopy grain approach is suitable to map AGB because crown size and spatial distribution are directly correlated to standing biomass of the dominant trees. However, one do not forget that the remote sensing-based model of AGB is assessed with respect to allometric predictions of "true" AGB, i.e. the aboveground dry mass of trees, from dendrometric data, so that the quality of the allometric model is potentially an additional source of bias (Chave et al. 2004; 2005).

Good correlations were also obtained between the first axis and tree density (r²= 0.8) or mean quadratic DBH (r²=0.71) in tropical evergreen *terra firme* forest Couteron et al (2005). However, forest heterogeneity and presence of relief makes the canopy approach to be used carefully, that is one must analyze visually whether the relief influences or not some of the PCA axes (e.g. Ploton, 2010). Only axes immune to relief influence should be used for biomass prediction otherwise the result may be biased or highly context-dependent. Moreover, due to the diversity of forest stand structures in tropical *terra firme* forests, a

Additionally, for a given species varying tree heights and crowns dimensions may yield important mass differences that the parsimonious relationships cannot take into account. Selecting an appropriate allometric model is then crucial and the sampling uncertainty relative to the size of the study plot should also be addressed carefully (e.g. Chave et al.

Tree location, crown shape, tree height and wood specific gravity also constitute useful information that will contribute to the characterization of the forest structure typology. Although it remains unrealistic in heterogeneous forests without the help of skilled botanists, identification of tree species is advisable in low-diversified situations, since the inclusion of a specific wood gravity parameter into allometric equations proved to improve significantly the model (Chave et al. 2005). Such additional data will also be valuable for initializing 3D forest templates. It is important to note that, in tropical forest, tree height measurements from the ground are problematic and cumbersome explaining the enthusiasm aroused by Lidar data (e.g. Gillespie et al. 2004). Another important point to improve AGB prediction would be to conduct forest inventories simultaneously to image

Assuming that the constituted forest plots dataset is well distributed within the acquired scene(s), Fourier r-spectra can be computed for windows centred on each plot. For example, when applied to 1-m Ikonos (Proisy et al. 2007) or 0.5-m Geoeye panchromatic images (Fig. 9) r-spectra permit good discrimination of a wide array of canopy structures of mangroves (Fig. 9). Furthermore pre-adult, mature and decaying mangrove forests show contrasted signatures with dominant frequencies around 180, 80, 50 and 30 cycles

Inverting FOTO indices (the three first PCA axes) into AGB of forest plots distributed over two different sites (i.e. two different images) yielded good correlations and low errors, as presented in Fig. 10. Compared to estimations provided by the P-band HV polarisation channel, FOTO-derived AGB did not show saturations over the whole range of mangrove biomass (Fig. 9), i.e. up to 500 tDM.ha-1 and rmse error remains acceptable (33 tDM.ha-1). This result suggests that, in the case of closed canopies with sub-strata of low biomass (e.g. the mangrove ecosystem in French Guiana), the canopy grain approach is suitable to map AGB because crown size and spatial distribution are directly correlated to standing biomass of the dominant trees. However, one do not forget that the remote sensing-based model of AGB is assessed with respect to allometric predictions of "true" AGB, i.e. the aboveground dry mass of trees, from dendrometric data, so that the quality of the allometric model is potentially an additional source of bias

Good correlations were also obtained between the first axis and tree density (r²= 0.8) or mean quadratic DBH (r²=0.71) in tropical evergreen *terra firme* forest Couteron et al (2005). However, forest heterogeneity and presence of relief makes the canopy approach to be used carefully, that is one must analyze visually whether the relief influences or not some of the PCA axes (e.g. Ploton, 2010). Only axes immune to relief influence should be used for biomass prediction otherwise the result may be biased or highly context-dependent. Moreover, due to the diversity of forest stand structures in tropical *terra firme* forests, a

2004).

acquisitions.

per kilometre.

(Chave et al. 2004; 2005).

**3.2 Sensitivity to forest structure and AGB** 

sufficient number of studies in diversified locations and contexts are still needed before general conclusions can be reached about the robustness of such correlations. Independent ongoing studies suggest that the correlation with density is highly contextspecific while the correlation with the mean quadratic diameter may be a more robust feature.

Fig. 9. Radial spectra and associated 100 x 100 m images of different mangrove growth stages using a 0.5 m panchromatic Geoeye image acquired in 2009. Forest inventories dated of 2010 and 2011. Note the r-spectra of the open canopy decaying stage. A photograph of this plot is available in Fig. 11.

Biomass Prediction in Tropical Forests: The Canopy Grain Approach 73

analyzed after separating deciduous and evergreen forests than may be simultaneously present in a given region. Appropriate regional pre-stratification using multispectral satellite data and/or L- or P-band polarized signatures (Proisy et al. 2002) may help towards

Fig. 11. Two examples of specific forest structures for which canopy grain and total AGB relationships cannot be safely derived without prior-stratification of the main forest types. Left: Decaying mangrove, with both large surviving trees and large canopy gaps, French Guiana © C. Proisy. Right: *Maranthaceae* understorey, overtopped by a fairly continuous albeit deciduous forest canopy referred to as "*Maranthaceae forests"* in Cameroun, Africa,

The canopy grain approach is largely original. It combines common techniques, i.e. Fourier transform and principal component analysis to characterize tropical canopy aspect and beyond forest structure from images of metric resolution. It can be implemented without prior radiometric correction, such as reflectance calibration or histogram range concordance. Regarding the increasing availability of metric to sub-metric optical images, the FOTO canopy grain analysis demonstrated its potential to capture gradients of forest structural characteristics in tropical regions. Within this context, the possible contribution of the canopy grain approach to the challenging task of estimating tropical above-ground biomass is worth being assessed at very broad scale. Such aim requires conducting simultaneously observational and simulation studies aiming at better understanding how canopy grain is sensitive to forest structure or biomass in various types of forests under various conditions of image acquisitions. There is particularly an important field of research in simulating multi-spectral and metric reflectance images from realistic forest 3D templates to identify, for instance, the range of conditions for which inversing above ground biomass of tropical forests appears possible. Considering the extreme complexity of most the tropical forests, it would be illusory to believe that only one remote sensing technique can provide all the information required to the AGB inversion. We thus believe that combining canopy grain analysis with low frequencies radar-based studies can provide new insights on this problem.

note the absence of any intermediate tree strata © N. Barbier.

this purpose.

**4. Conclusion** 

Fig. 10. Comparison of FOTO- (Proisy et al. 2007) and P-HV-derived (from Mougin et al. 1999) biomass estimates in mangroves of French Guiana

#### **3.3 Present limitations of the methods and prerequisite**

In tropical forest, both gaps and multi-strata organization are often observed. Gaps are due to accidental tree falls or natural decaying of some canopy trees (Fig. 11, left). In presence of gaps, r-spectra tend to be skewed towards low frequencies and this may be erroneously interpreted as if the canopy contained large tree crowns (Fig. 9, r-spectrum of the decaying stage). In fact, gap-influenced r-spectra cannot be automatically related to the same biomass levels and must be removed from the PCA analysis to avoid biases in the AGB-FOTO relationship. Identically, the method was so far tested principally on evergreen forests. Further studies are needed regarding deciduous forests, not only because of the seasonal changes of the canopy aspect, but also because biomass of understorey vegetation often found in such forest type is not necessarily negligible. As spectral properties of the understorey may influence the overall reflectance of the corresponding pixels, this may be all more confusing if there is no intermediate stratum beneath the highest deciduous trees. An example of this is provided by the so-called *Maranthaceae* forest in Africa (Fig. 11, right), which presents a fairly closed albeit deciduous canopy and a very scarce intermediate tree storey. Such a structure allows the development of a dense herbaceous cover. Without relevant field information, results of the FOTO approach may be confusing in those forests. Their standing biomass is probably less than for evergreen closed forests since woody intermediate storey is missing, whereas both canopies are dominated by trees with large crowns. At least, statistical relationships between FOTO indices and AGB should be analyzed after separating deciduous and evergreen forests than may be simultaneously present in a given region. Appropriate regional pre-stratification using multispectral satellite data and/or L- or P-band polarized signatures (Proisy et al. 2002) may help towards this purpose.

Fig. 11. Two examples of specific forest structures for which canopy grain and total AGB relationships cannot be safely derived without prior-stratification of the main forest types. Left: Decaying mangrove, with both large surviving trees and large canopy gaps, French Guiana © C. Proisy. Right: *Maranthaceae* understorey, overtopped by a fairly continuous albeit deciduous forest canopy referred to as "*Maranthaceae forests"* in Cameroun, Africa, note the absence of any intermediate tree strata © N. Barbier.
