Skip to main content

Evidence of time-lag in the provision of ecosystem services by tropical regenerating forests to coffee yields

Cite this dataset

González-Chaves, Adrian et al. (2023). Evidence of time-lag in the provision of ecosystem services by tropical regenerating forests to coffee yields [Dataset]. Dryad.


Abstract Restoration of native tropical forests is crucial for protecting biodiversity and ecosystem functions, such as carbon stock capacity. However, little is known about the contribution of early stages of forest regeneration to crop productivity through the enhancement of ecosystem services, such as crop pollination and pest control. Using data from 610 municipalities along the Brazilian Atlantic Forest (30 m spatial resolution), we evaluated if young regenerating forests (less than 20 years old) are positively associated with coffee yield and whether such a relationship depends on the amount of preserved forest in the surroundings of the coffee fields. We found that regenerating forest alone was not associated with variations in coffee yields. However, the presence of young regenerating forest (within a 500 m buffer) was positively related to higher coffee yields when the amount of preserved forest in a 2 km buffer is above a 20% threshold cover. These results further reinforce that regional coffee yields are influenced by changes in biodiversity-mediated ecosystem services, which are explained by the amount of mature forest in the surrounding of coffee fields. We argue that while regenerating fragments may contribute to increased connectivity between remnants of forest fragments and crop fields in landscapes with a minimum amount of forest (20%), older preserved forests (more than 20 years) are essential for sustaining pollinator and pest enemy’s populations. These results highlight the potential time lag of at least 20 years of regenerating forests’ in contributing to the provision of ecosystem services that affect coffee yields (e.g., pollination and pest control). We emphasize the need to implement public policies that promote ecosystem restoration and ensure the permanence of these new forests over time.


We collected data on crop productivity from the Brazilian Institute of Geography and Statistics (IBGE,, corresponding to 1.3 Mha destined to coffee production within 610 municipalities in the Atlantic Forest. Most of the municipalities produce C. arabica (509), and a small proportion cultivates either C. canephora (44) or both species (57). We specified the coffee species planted in each municipality by using a pollinator demand index (PD), which considers the benefits known to draw from pollination weighted by the area destined to each species (Klein et al 2003; González-Chaves et al 2021). We calculated the mean coffee yields (number of 60 kg bags per hectare) from three consecutive years for each municipality, accordingly to the coffee field maps available for each of the five Brazilian States where coffee is produced within the Atlantic Forest. The year of the coffee field maps ranges between 2008 and 2012, as they were independently done by initiatives in each State and brought together by the National Supply Company (CONAB in Portuguese, who shared the data. 

We estimated the age of each native forest pixel surrounding coffee fields using the annual land-use cover maps from the MapBiomas collection 5.0 based on Landsat imagery with a spatial resolution of 30 m from 1985 to 2019 (Souza et al 2020). We developed a forest age map for each of the five states included in the analysis based on the years of available coffee field maps. First, we mapped native forest cover that was present since 1985 (older forest), which is the first year of the Mapbiomas time series. We then identified regenerating forests as forest pixels that were classified either as cropland or pasture for at least three years and remained as forest until the correspondent year from the coffee reference map, considering a minimum of 3 years as forest (Rosa et al 2021). The age of regeneration was calculated as the number of years between the regeneration event and the coffee reference map for each state.

Regenerated forests are still a small fraction of the forest cover in the regions. Most forest in the region have > 30 years (80.6%; Rosa et al. 2021) and the mean age of the forest regenerating after 1989 is around 12 years (Piffer et al., this issue). Thus, we subdivided the Atlantic Forest fragments from our study region into two age groups: those with less than 20 years of age (hereafter young regenerating forest, YRF) and those with 20 or more years, older forest (see Fig. 1d). The 20 years age threshold was established due to limitation in our data, as the Mapbiomas time series started in 1985 and the oldest coffee map available for Espirito Santo state was 2008, thus 20 years is the oldest age available for all municipalities (Rosa et al 2021, Souza et al 2020). Therefore, the forest age is unknown for forest already present before 1985. We did not consider any threshold for younger forest fragments as secondary forests are highly dynamic in Latin America (including the Atlantic Forest) and forest fragments with less than 10 years of age are more commonly cleared (Chazdon et al 2016, Rosa et al 2021). Moreover, biodiversity can partially recover after 20 years of age (Barlow et al 2007, Poorter et al 2016). Our final maps contained three forest features: i) young regenerating forest (YFR) under twenty years of age, ii) forest with more than twenty years of age and iii) overall forest cover disregarding forest age.

We calculated the percentage of young regenerating forest and older forest surrounding of coffee fields by using a moving window analysis for each coffee pixel at 2 km and 500 m buffer radius, with the raster package using R 4.1 (R Development Core Team 2021). The spatial scales were considered based on studies showing that above and below ground biodiversity responds to those scale in human modified landscapes (le Provost et al., 2021), as well as pollination and pest control services benefitting coffee production (Aristizábal and Metzger, 2019; González-Chaves et al., 2020; Librán-Embid et al., 2017; Saturni et al., 2016) . Moreover, regeneration is also known to respond to landscape at similar scales (Crouzeilles and Curran, 2016; Piffer et al., this issue). Coffee yield data was available at the municipality level, a scale at which forest cover has been monitored to reduce deforestation (Koch et al 2019). Thus, apart from calculating mean values of the percentage of forest surrounding coffee fields for each municipality, we also calculate the percentage of each forest feature at the municipality level. Therefore, we evaluated the three forests features, at the three different scales (500 m, 2 km and municipality level). 


São Paulo Research Foundation, Award: 2013/23457-6

São Paulo Research Foundation, Award: 2017/19411-1

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

Conselho Nacional de Desenvolvimento Científico e Tecnológico, Award: 305157/2018-3

Conselho Nacional de Desenvolvimento Científico e Tecnológico, Award: 305484/2017-6

Conselho Nacional de Desenvolvimento Científico e Tecnológico, Award: 421668/2018-0