Combined effects of landscape composition and heterogeneity on farmland avian diversity

Abstract Conserving biodiversity on farmland is an essential element of worldwide efforts for reversing the global biodiversity decline. Common approaches involve improving the natural component of the landscape by increasing the amount of natural and seminatural habitats (e.g., hedgerows, woodlots, and ponds) or improving the production component of the landscape by increasing the amount of biodiversity‐friendly crops. Because these approaches may negatively impact on economic output, it was suggested that an alternative might be to enhance the diversity (compositional heterogeneity) or the spatial complexity (configurational heterogeneity) of land cover types, without necessarily changing composition. Here, we develop a case study to evaluate these ideas, examining whether managing landscape composition or heterogeneity, or both, would be required to achieve conservation benefits on avian diversity in open Mediterranean farmland. We surveyed birds in farmland landscapes of southern Portugal, before (1995–1997) and after (2010–2012) the European Union's Common Agricultural Policy (CAP) reform of 2003, and related spatial and temporal variation in bird species richness to variables describing the composition, and the compositional and configurational heterogeneity, of the natural and production components of the landscape. We found that the composition of the production component had the strongest effects on avian diversity, with a particularly marked effect on the richness of farmland and steppe bird species. Composition of the natural component was also influential, mainly affecting the richness of woodland/shrubland species. Although there were some effects of compositional and configurational heterogeneity, these were much weaker and inconsistent than those of landscape composition. Overall, we suggest that conservation efforts in our area should focus primarily on the composition of the production component, by striving to maximize the prevalence of biodiversity‐friendly crops. This recommendation probably applies to other areas such as ours, where a range of species of conservation concern is strongly associated with crop habitats.

. Percentage of occurrence of bird species recorded in 73 transects sampled annually during the breeding season in southern Portugal, in 1995Portugal, in -1997Portugal, in and 2010Portugal, in -2012. Species are classified according to their habitat affinities (F -farmland; S -steppe; W -woodland; O -other), conservation status (SPEC #), and phenology (R -resident, M -migratory

Compositional heterogeneity
Land cover richness (no., CR) Total number of different natural/production land cover types.
Land cover diversity (SHDI) a Shannon's diversity index computed on the proportion of different natural/production land cover types.
Land cover evenness (SHEI) b Shannon's evenness index computed on the proportion of different natural/production land cover types.

Configurational heterogeneity
Largest patch index (%, LPI) Percentage of area of the largest natural/production land cover type patch.
Patch size (ha, AREA) Mean area of natural/production land cover type patches.
Edge density (m 2 /ha, ED) Density of edges between natural and production land cover type patches.

Shape complexity (SHAPE)
Mean perimeter-to-area ratio of natural/production land cover type patches. a SHDI = 0 when the landscape contains only 1 or 0 cover types; b SHEI = 0 when the landscape contains only 1 or 0 cover types. SHEI = 1 when distribution of area among patch types is perfectly even (i.e., proportional abundances are the same). Table S3. Formulation of candidate models (g 1-63 ) based on all possible combinations of the six sets of landscape variables listed in Table 1.

No. variable sets No. models Model formulation
One Five sets 6 g 57 = Set 1 + Set 2 + Set 3 + Set 4 + Set 5 g 58 = Set 1 + Set 2 + Set 3 + Set 4 + Set 6 g 59 = Set 1 + Set 2 + Set 3 + Set 5 + Set 6 g 60 = Set 1 + Set 2 + Set 4 + Set 5 + Set 6 g 61 = Set 1 + Set 3 + Set 4 + Set 5 + Set 6 g 62 = Set 2 + Set 3 + Set 4 + Set 5 + Set 6 Six sets 1 g 63 = Set 1 + Set 2 + Set 3 + Set 4 + Set 5 + Set 6  Table S4. Summary of average models relating spatial variation in bird species richness in [1995][1996][1997] to landscape variables. In each case we provide the model-averaged partial standardized coefficients (Coef) and their partial standardized standard error (SE). The relative importance of each variable in the model (Imp) was calculated as the ratio between the respective partial standardized coefficient and the largest standardized coefficient in the model (Cade 2015 Cover diversity (Natural) 0.03 0.24 0.07 Figure S1. Classification tree of land cover categories used to model the relations between bird species richness and landscape characteristics in southern Portugal. Categories were defined considering the main nesting and foraging habitats of bird species in the study area (Moreira 1999;Delgado & Moreira 2000;Stoate et al. 2000;Reino et al. 2009Reino et al. , 2010, and assuming that habitat preferences are often influenced strongly by structural characteristics (e.g. tree density, shrub cover, sward density and height, and amount of bare ground -ground cover). Characteristics of the herbaceous sward were considered during the sampling months (April-May), though they are known to vary strongly during the annual cycle (e.g., dry annual crops are sown in autumn and thus the sward is tall and dense during the breeding season, whereas irrigated annual crops are generally sown in spring, and so during the breeding season the sward tends to be short, sparse, and with a high proportion of bare ground). Figure S2. Spline correlograms describing spatial autocorrelation for total bird species richness and for the residuals of models relating species richness to landscape variables (Tables S4 -S6). Separate correlograms are presented for 1995-97 (a, d), 2010-12 (b, e), and temporal variation (c, f). Lines represent the estimate (in the middle) and the 95% confidence envelopes (external lines) using 1000 bootstrap resamples (Bjørnstad & Falck 2001). Figure S4. Spline correlograms describing spatial autocorrelation for farmland bird species richness and for the residuals of models relating species richness to landscape variables (Tables S4 -S6).