**3.4 FRB cartography validation**

312 Remote Sensing of Biomass – Principles and Applications

Nº plots 182 130 182 130 182 130 182 130 182 130 TM1 -0.595\* -0.616\* -0.600\* -0.621\* -0.599\* -0.617\* -0.606\* -0.628\* -0.598\* -0.630\* TM2 -0.585\* -0.617\* -0.580\* -0.613\* -0.573\* -0.609\* -0.579\* -0.619\* -0.585\* -0.632\* TM3 -0.594\* -0.638\* -0.591\* -0.640\* -0.588\* -0.629\* -0.597\* -0.641\* -0.602\* -0.657\* TM4 -0.224\* -0.182\* -0.243\* -0.210\* -0.220\* -0.192\* -0.224\* -0.188\* -0.240\* -0.205\* TM5 -0.622\* -0.678\* -0.642\* -0.696\* -0.635\* -0.684\* -0.638\* -0.696\* -0.650\* -0.698\* TM7 -0.605\* -0.683\* -0.625\* -0.696\* -0.621\* -0.685\* -0.632\* -0.704\* -0.641\* -0.701\* PC1 -0.592\* -0.653\* -0.608\* -0.663\* -0.601\* -0.651\* -0.606\* -0.664\* -0.615\* -0.668\* PC2 -0.059 -0.131 -0.063 -0.147 -0.084 -0.156 -0.082 -0.154 -0.082 -0.159 PC3 -0.563\* -0.621\* 0.579\* 0.634\* 0.573\* 0.621\* 0.591\* 0.636\* 0.605\* 0.649\* TC1 -0.578\* -0.629\* -0.588\* -0.637\* -0.581\* -0.624\* -0.584\* -0.636\* -0.595\* -0.643\* TC2 0.437\* 0.477\* 0.443\* 0.484\* 0.445\* 0.479\* 0.453\* 0.484\* 0.449\* 0.491\* TC3 0.609\* 0.670\* 0.629\* 0.694\* 0.622\* 0.678\* 0.636\* 0.696\* 0.654\* 0.700\* NDVI 0.626\* 0.665\* 0.627\* 0.660\* 0.626\* 0.655\* 0.634\* 0.657\* 0.630\* 0.665\* SAVI 0.624\* 0.663\* 0.625\* 0.657\* 0.624\* 0.653\* 0.631\* 0.655\* 0.628\* 0.663\* MSI -0.592\* -0.640\* -0.629\* -0.667\* -0.617\* -0.652\* -0.633\* -0.667\* -0.643\* -0.676\* MID57 -0.618\* -0.684\* -0.638\* -0.699\* -0.632\* -0.687\* -0.639\* -0.702\* -0.648\* -0.701\*

Table 7. Pearson's coefficient of correlation (R) between spectral variables and data obtained using segmentation and 3x3 pixel windows with restrictions and without restrictions

*<sup>a</sup> β0 β1 β2*

N2 ln\_MID57 0.535 16.409 -3.888 - 5.657 35.58 N3 ln\_TM7 0.560 11.395 -3.411 - 8.840 50.07 N4 ln\_MID57 0.596 16.563 -3.928 - 9.688 66.67 N5 ln\_MID57 0.533 16.217 -3.846 - 8.231 51.01 Table 8. Linear regression models obtained using segmentation and 3 x 3 pixel windows

ln\_TM1 0.547 7.746 0.303 -1.087 9.064 44.95

**Segmentation S3** 

**CV7 CV6 CV7 CV6 CV7 CV6 CV7 CV6 CV7 CV6** 

**Segmentation S4** 

> *RMSE (ton/ha)*

*RMSEr (%)* 

**Fixed 3x3 without restrictions** 

**Segmentation S2** 

**Segmentation S1** 

(\* correlation is significant at the 0.05 level)

**Model Variable** *R2*

TC3,

N1

with restrictions

The accuracy assessments of every regression equation (RMSE and RMSEr) were completed using plots that showed the same homogeneity criteria as was used to run the model. However, since the selected estimation models have been applied to each one of the Landsat pixels located in forested areas in Teruel Province, the degree of success in these must also be evaluated at that scale.

To accomplish this, the NFI-2 plots excluded from the estimation models and their validation were considered. In order to guarantee the results, those plots that were affected by inaccuracies in their field location and/or by the radiometric response of different landscape elements located in their immediate vicinity, were removed from the validation sample. Consequently, group CV8 was used since it includes a high number of plots, which ensures that the validation results were not biased by using only the ideal plots.

As it can be seen in figure 9, the results show few differences between the three maps. Those obtained from 3 x 3 fixed windows yieded a RMSEr of 64.26%, while spectral homogeneous forest areas had RMSEr values of 66, 71% and 65.06%, respectively. These results at pixel level can be considered tolerable for the study area considering previous experiments using similar methodologies for boreal environments less affected by heterogeneity than Mediterranean forests. Thus, Tokola et al. (1996), Tokola & Heikkila (1997), Mäkkelä & Pekkarien (2001) and Katila & Tomppo (2001) reported RMSEr to estimate forest parameters such us timber volume or total volume from about 65% to more than 100%. In this respect, it is important to emphasize that estimation error in cartography derived from satellite images decreases with an increase in the size of the area used to validate it. For example, Fazakas et al. (1999) showed a RMSEr of 66.5% at pixel level, but when using an aggregation area of 598 ha, the RMSEr was 8.7%. However, it was not possible to carry out a similar analysis in our study area because no other FRB data were available at any scale.

Using Remote Sensing to Estimate a Renewable Resource: Forest Residual Biomass 315

This study demonstrates the utility of Landsat TM images and forest inventory data in estimating FRB in Mediterranean areas. The methodology employed provides a continuous and complete estimation of FRB that overcomes problems associated with the limitations of point inventory. This information can be very useful in determining the most suitable areas in which to install power stations that make use of this resource; the lack of this type of information is one of the main problems currently facing the industry. As a result, an increase in the use of this kind of biomass resource will help to achieve the stated objectives of renewable energy production in Spain. Moreover, this is especially important considering two facts: (i) the current socio-economic context, with an increase in the price of petrol due to international instability, and public concerns regarding nuclear power, owing to Fukushima incident; and (ii) biomass is currently the only renewable energy that can be used as a strategic source of energy, as it is always available independent of weather

The three sources of information typically used for biomass estimation (data from field sampling, satellite imagery and ancillary data) (Lu, 2006) have been carefully integrated, bearing in mind the specific characteristics of the Mediterranean environment. Thus, to reduce the problem that heterogeneity introduced into estimation models, three different methods were tested to extract radiometric data from a TM image. In agreement with previous studies focused on AGB estimation (i.e. Foody, 2001; Lu et al., 2004; Lu, 2005; Lu & Batistiella, 2005; Steininger, 2000) and with previous analyses carried out with our data (García-Martín et al., 2006, 2008a, 2008b), the derived spectral variables related to wetness showed the strongest correlations to FRB, independent of the method used to extract the

Fig. 10. FRB in the study area in July 2008

conditions and can be stored easily prior to use (Jarabo, 1999).

**4. Conclusions** 

Fig. 9. RMSEr differences between FRB maps obtained using the three extraction methods
