Skip to main content
Dryad logo

Positive forest cover effects on coffee yields are consistent across regions

Citation

González-Chaves, Adrian; Carvalheiro, Luísa; Garibaldi, Lucas; Metzger, Jean Paul (2021), Positive forest cover effects on coffee yields are consistent across regions, Dryad, Dataset, https://doi.org/10.5061/dryad.612jm644g

Abstract

1. Enhancing biodiversity-based ecosystem services can generate win-win opportunities for conservation and agricultural production. Pollination and pest control are two essential agricultural services provided by mobile organisms, many depending on native vegetation networks beyond the farm scale. Many studies have evaluated the effects of landscape changes on such services at small scales. However, several landscape management policies (e.g., selection of conservation sites) and associated funding allocation occur at much larger spatial scales (e.g., state or regional level). Therefore, it is essential to understand whether the links between landscape, ecosystem services, and crop yields are robust across broad and heterogeneous regional conditions.

2. Here, we used data from 610 Brazilian municipalities within the Atlantic forest region (~50 Mha) and show that forest is a crucial factor affecting coffee yields, regardless of regional variations in soil, climate and management practices. We found forest cover surrounding coffee fields was better at predicting coffee yields than forest cover at the municipality level. Moreover, the positive effect of forest cover on coffee yields was stronger for Coffea canephora, the species with higher pollinator dependence, than for C. arabica. Overall, coffee yields were highest when coffee fields were near to forest fragments, mostly in landscapes with intermediate to high forest cover (> 20%), above the biodiversity extinction threshold.

3. Coffee cover was the most relevant management practice associated with coffee yield prediction. An increase in crop area was associated with a higher yield, but mostly in high forest covers municipalities. Other localized management practices like irrigation, pesticide use, organic manure, and honey-bee density had little importance in predicting coffee yields than landscape structure parameters. Neither the climatic or topographic variables were as relevant as forest cover at predicting coffee yields.   

4. Synthesis and application. Our work provides evidence that landscape relationships with ecosystem service provision are consistent across regions with different agricultural practices and environmental conditions. These results provide a way in which landscape management can articulate small landscape management with regional conservation goals. Policies directed towards increasing landscape interspersion of coffee fields with forest remnants favor spillover process, and can thus benefit the provision of biodiversity-based ecosystem services, increasing agricultural productivity. Such interventions can generate win-win situations favoring biodiversity conservation and increased crop yields across large regions.

Methods

We obtained productivity data from the Brazilian Institute of Geography and Statistics (IBGE, http://www.ibge.gov.br/), we calculated coffee yield (productivity) for each year per municipality by dividing the total production (tons) by the total coffee area (ha) planted per municipality per year. Mean coffee yields were calculated from three consecutive years for each municipality. The years considered for each municipality depended on data availability of the coffee fields’ maps, which was different for each state. To determine forest cover surrounding coffee plantations, we used coffee maps from the National Company of supply (CONAB, http://www.conab.gov.br/), which compiled maps from the five leading coffee producing states within the Atlantic forest. Additionally, we used annual forest remnants maps from MapBiomas (Project of annual mapping of land-use and land-cover of Brazil, http://mapbiomas.org/), both with a resolution of 30x30 m.

We gathered 19 bioclimatic variables from the Worldclim 30 seconds resolution database (www.worldclim.org). We calculated the mean values per municipality by extracting the climatic values from the coffee field maps. The bioclimatic variables include annual mean temperature and precipitation, and extreme or limiting factors relevant to coffee production (Table S1). Regarding soil properties, we obtained physical (bulk density, clay coarse and silt content at different depths) and chemical (cation exchange capacity, soil organic carbon content and soil pH) data from SoilGrids at 250 m (www.soilgrids.org), and then extracted mean values for the coffee fields at the municipality level (Fig. S1). 

Coffee agricultural management practices are from the IBGE database, based on a field survey done in 2006 that assessed the number of coffee farms under a particular management practice. We considered variables associated with management intensity and calculated the percentage of farms that use: irrigation, mechanical harvest, organic or chemical manure, and pesticides. Additionally, we also calculated the percentage of organic farms within each municipality.

Usage Notes

For model selection, transformation of the variables were made. For instance, yield values were log transformed. As well as coffee cover and forest cover variables. The rest of climatic and management were scaled before creating the full model.

Funding

Fundação de Amparo à Pesquisa do Estado de São Paulo, Award: 2013/234567-6

Fundação de Amparo à Pesquisa do Estado de São Paulo, Award: 2017/14911-1

Fundação de Amparo à Pesquisa do Estado de São Paulo, Award: 2018/06330-6