Results: Testing the ELIA model against biodiversity empirical data
Our study provides evidence that depending on the interplay between energy storage and its distribution pattern (E ·I ), the societal metabolic flows driven by farming imprint different landscapes (L ) where agroecosystems may either enhance or decrease populations and species richness of butterflies and birds in the Barcelona Metropolitan Region. Significant results of Structural Equation Models (SEM) are summarized in Fig. 2 and 3 , while complete data are shown in Tables S1 to S8 inAnnex A of the Supplementary Information .
[Insert Table 1 here]
The butterfly’s Principal Component Analysis (Table 1 ) shows that the first factor is the land cover composition of the landscape (Cm1; 43.9% of variance) that obtains higher loading for forest, while the second (Cm2; 35.5%) presents greater loading for cropland. The first component of land metrics which assesses landscape configuration through land metrics (Cn1; 72.4%) shows negative loadings for diversity and fragmentation and positive loadings for grain size and connectivity, while the second component is more heterogeneous and contributes much less to explain the variance (Cn2; 12.2%) despite being mostly associated to connectivity metrics (including effective mesh size).
[Insert Figure 2 here]
The SEM results (Fig. 2 ) show that total butterfly observations (TBOB) are positively associated to ELIA values while species richness (TBSR) does not show any significant correlation with this model. ELIA is negatively associated to Cm1 in the TBSR model, but not in the TBOB one. In both models, Cm1 and Cn1 were negatively associated. The R2 for the endogenous variables of TBSR and TBOB models were 0.316 and 0.266 respectively. When disentangling the effects of ELIA components (E , Iand L ), I shows a positive association with both TBSR and TBOB, while E and L are only significant and positively associated to TBOB. Landscape composition (Cm) and configuration (Cn) are not significantly associated to any biodiversity component, yet Cm1 is negatively associated to ELIA values in both models and Cm2 is negatively associated to I only in the TBOB model. The R2 for the endogenous variables of TBSR and TBOB are 0.420 and 0.334 respectively.
In the bird’s Principal Component Analysis, the first factor of landscape composition (Cm1; 39.3% of variance) shows higher loading for forest, while the second factor (Cm2; 31.8%) has greater loadings for scrubland and cropland (Table 1 ). The first factor of landscape configuration (Cn1; 53.6%) records higher (negative) loadings for landscape diversity and fragmentation, while the second (Cn2; 27.3%) shows greater (positive) association for ecological connectivity and effective mesh size.
[Insert Figure 3 here]
The SEM results (Fig. 3 ) show that breeding bird species richness (BBSR) and wintering bird species richness (WBSR) are positively related to ELIA values and Cm2, and negatively related to Cn2. ELIA is negatively related to Cn1 in both models, and with Cm1 only in the BBSR model. Cn2 and Cm1 are also negatively associated in this model. The R2 for the endogenous variables of BBSR and WBSR are 0.244 and 0.210 respectively. If we disentangle the effects of ELIA components (E , Iand L ), I and E are positively and negatively correlated, respectively, with both BBSR and WBSR. L is only associated negatively to WBSR. E is negatively associated to Cm2 in both models, and to Cm1 only in the BBSR model. L and Cm1 are negatively associated in this last model. The R2 for the endogenous variables of BBSR and WBSR are 0.321 and 0.329 respectively.