The authors have declared that no competing interests exist.
Conceived and designed the experiments: NL YL MC. Performed the experiments: NL YL MC. Analyzed the data: NL BL. Contributed reagents/materials/analysis tools: NL YL MC. Wrote the paper: NL YL BL GV MC.
Because industrial agriculture keeps expanding in Southeast Asia at the expense of natural forests and traditional swidden systems, comparing biodiversity and ecosystem services in the traditional forest–swidden agriculture system vs. monocultures is needed to guide decision making on land-use planning. Focusing on tree diversity, soil erosion control, and climate change mitigation through carbon storage, we surveyed vegetation and monitored soil loss in various land-use areas in a northern Bornean agricultural landscape shaped by swidden agriculture, rubber tapping, and logging, where various levels and types of disturbance have created a fine mosaic of vegetation from food crop fields to natural forest. Tree species diversity and ecosystem service production were highest in natural forests. Logged-over forests produced services similar to those of natural forests. Land uses related to the swidden agriculture system largely outperformed oil palm or rubber monocultures in terms of tree species diversity and service production. Natural and logged-over forests should be maintained or managed as integral parts of the swidden system, and landscape multifunctionality should be sustained. Because natural forests host a unique diversity of trees and produce high levels of ecosystem services, targeting carbon stock protection, e.g. through financial mechanisms such as Reducing Emissions from Deforestation and Forest Degradation (REDD+), will synergistically provide benefits for biodiversity and a wide range of other services. However, the way such mechanisms could benefit communities must be carefully evaluated to counter the high opportunity cost of conversion to monocultures that might generate greater income, but would be detrimental to the production of multiple ecosystem services.
Drastic land-use transformations have occurred in the tropical forest landscapes of Southeast Asia in the past decades, leading to the disappearance of natural forests and the replacement of traditional land-use systems with monoculture plantations. On the island of Borneo, the lowland rainforests are at the crossroads of multiple and divergent interests. While these rainforests are hotspots of biological diversity with a high rate of endemism and hold important carbon stocks, they are also a major source of valuable timber, and are situated on lands that are very suitable for conversion to oil palm or other large industrial plantations [
Since the late 1960s, logging has affected most of the lowland forests [
Since negative correlations usually exist between goods and services (e.g. [
Research gaps concerning the effect of land-use changes on the ecosystem services produced by swidden systems have been identified through a systematic review currently under progress that aims to bring unbiased evidence to the debate [
In this study, we address the following question: What level of biodiversity and ecosystem services are found in the different land uses related to the traditional forest–swidden agriculture system? We conducted a case study from a northern Bornean agricultural landscape where we quantitatively estimated the contribution of various land uses to: (1) climate change mitigation through carbon storage in live aboveground biomass and topsoil, (2) tree species diversity, and (3) soil erosion control. The two services were chosen because of their relevance for multiple beneficiaries at different scales (local to regional for soil erosion control, and global for climate change mitigation). Tree species diversity, which we did not consider as an ecosystem service, was chosen because of its cross-cutting and cross-scale nature, as it jointly influences the delivery of goods (e.g. food, raw material, fruit, and timber for local people) and services (e.g. water regulation at the regional scale) [
This study strictly complied with Indonesian laws. Authorizations to carry out research activities were obtained from appropriate sources both at national (Ministry of Forestry and Ministry of State for Research and Technology) and local level (Head of Kapuas Hulu regency). Permissions from local owners to work on their lands were obtained prior to any activity. Sending soil and herbarium samples from the field site to laboratories in Java for further analysis was done after authority approval. We did not collect endangered plants or any animals.
Field work was carried out in the surroundings of Keluin (1°08'57" N, 112°15'37" E), a village located in the district of Batang Lupar, Kapuas Hulu regency, West Kalimantan province, Indonesia (
The study area is located on the island of Borneo (top left panel), in the Indonesian province of West Kalimantan, in the regency of Kapuas Hulu (bottom left panel). In the main panel, plot location and broad land-use types (LUT) are displayed along with former logging roads (red) and rivers (blue). The black square indicates the location of the village.
This site was chosen because of the diversity of land uses representative of northern Bornean traditional agricultural systems that were found within a limited perimeter (ca. 2 km W–E by 5 km N–S). Traditionally, the main crop, rice, is cultivated along with other annual crops (such as cassava, maize, etc.) in swiddens from either primary or secondary forest clearing before the plot is abandoned after 1−2 years of cultivation. Rubber seedlings and saplings, which are planted in some crop fields during the cultivation phase or just after plot abandonment, eventually lead to rubber-based secondary forests (also called “jungle rubber” gardens) as a result of plant succession. Past and current uses of land by the local community (swidden agriculture system mixed with rubber gardens) and past logging activities have created a mosaic of vegetation reflecting various disturbance types, ages and intensities in the Keluin area.
We defined seven land-use types in the study area: (1) food crop fields where the main crop, rice (
Two distinct sets of vegetation and erosion plots were established based on a multi-stratified sampling first across land-use type and second across disturbance age for stands regenerating after slash and burn (see
For young and old stands in both secondary regrowth areas and rubber gardens, 12 plots each of 20 × 20 m were randomly selected (for a total surveyed area of 0.48 ha for each of the four land-use types) to capture the variability of vegetation structure and composition within each land-use type (
All trees with diameter at breast height (DBH, 1.3 m) ≥ 5 cm were measured, tagged and mapped, and their height estimated, following standard procedures [
We used three different indices to characterize tree species diversity at the plot level: species richness, Berger–Parker index, and Fisher's α. Species richness is the simplest measure of species diversity but does not take into account community evenness. Conversely, the Berger–Parker index (defined as the inverse of the proportion of individuals of the most common species in the community) depends only on evenness, and is sensitive to the dominance of a few species. Fisher's α is mathematically unrelated to the first two indices, is relatively independent of sample size, and is insensitive to the presence of rare species [
We used the generic Chave et al. allometric equation for tropical forests [
Because of differences in sampling design (inter-plot distance ranged from 20 to 4650 m), we could not use tree diversity values aggregated over the total surveyed area (0.48 ha each for secondary regrowth areas and rubber gardens, 1 ha each for natural and logged-over forest) to compare the different land-use types (see
Silt fences (made from a nonwoven polyester geotextile) were used to measure hillslope erosion. Following guidelines from Robichaud and Brown [
All statistical analyses were done using R 3.1.2 [
A nonmetric multidimensional scaling (NMDS) analysis was performed to illustrate plot similarity in terms of tree species composition (using the metaMDS function; see [
The three species diversity indices were highest in natural forests (
Indices were computed for each 20 × 20 m plot before being averaged by land-use type (over 25 plots for logged-over and natural forest, and 12 otherwise). Mean values with the same letter are not significantly different (Tukey’s HSD test on uncorrected values in case no spatial autocorrelation was detected and corrected values otherwise, p < 0.01).
Older stands (for rubber gardens and secondary regrowth forest) consistently showed higher index values compared with young ones, but the difference was only significant for mean Fisher’s α. For similar time intervals since last disturbance, all the diversity indices were consistently higher in secondary regrowth areas compared with rubber gardens, but the difference was not significant.
In the two dimensions of the NMDS plot (
Each dot (98 in total) represents a 20 × 20 m plot.
Aboveground carbon stock levels were significantly higher in natural forests than in any other land-use types (
Means were computed over 12 to 25 replicates per land-use type for aboveground biomass, and over five replicates per land-use type for topsoil. Aboveground biomass (AGB) was split into four fractions according to tree diameter at breast height (Ø). Mean values with the same letter (lowercase for aboveground biomass, uppercase for topsoil) are not significantly different (Tukey’s HSD test on uncorrected values in case no spatial autocorrelation was detected and corrected values otherwise, p < 0.01).
For all types of land use, the highest proportions of aboveground carbon were contained in trees with DBH = 10−30 cm, while the lowest non-null proportions were contained in trees with DBH = 5−10 cm. Yet, for recently disturbed stands such as those in young secondary regrowth areas, small trees (those with DBH = 5−10 cm) could represent up to ca. 25% of aboveground carbon.
Although natural forests stored on average twice as much carbon in topsoils as other land-use types, significant differences were observed only between natural forests and secondary regrowth areas (both young and old) or old rubber gardens due to high variability within land-use types (for logged-over and natural forests, especially).
The greater the aboveground carbon stocks, the greater the species richness in all land-use types, with the noticeable exception of natural forests for which the relationship is negative (cf. regression lines;
Each dot (98 in total) represents a 20 × 20 m plot. Horizontal and vertical dashed lines represent median values of species richness (n = 25) and carbon stocks in aboveground biomass (68 Mg C ha-1), respectively. Regression lines (along with standard error) are computed independently for each land-use type. A second-order regression model (best-fit significant model selected among polynomial models with degrees 0 to 3) over the whole data set is also displayed (in black).
Annual soil loss values ranged from 0.8 to 2.2 g m–2 yr–1, with individual plot values varying from 0.5 to 2.7 g m-2 yr-1 (see
Data are averaged over the monitoring period (June 2012 to September 2013) and over the different replicates for each land-use type. Values from three replicates (one in young secondary regrowth area, one in young rubber garden, one in logged-over forest) were discarded because they were abnormally high (> two times mean value of the corresponding land-use type). Mean values with the same letter are not significantly different (Tukey’s HSD test on uncorrected values in case no spatial autocorrelation was detected and corrected values otherwise, p < 0.01).
As expected, service production was highest in natural forests (
Indicators are normalized so that the minimum possible value of an indicator is at the center of the radial plot and the maximum observed values are on the outer circles (for the service of soil erosion control, the indicator is the inverse of the measured soil loss).Service indicators: Ct = carbon stocks in topsoil; Cb = carbon stocks in aboveground biomass; Sr = tree species richness; Fi = Fisher’s α; Bp = Berger–Parker index; Ec = soil erosion control.
Aboveground carbon stocks in human-modified land-use types did not reach half those of natural forests. Logged-over forests, agroforests and secondary regrowth areas had topsoil carbon stocks 40–60% lower than those of natural forests. This contrasts with results from Kessler
Regarding soil erosion control, even if soil loss was lowest in natural forests and differed significantly between some pairs of land-use types, low absolute values of annual soil loss (2−3 orders of magnitude lower than the tolerable soil erosion rate [
Our results on tree species diversity show that some areas of human-modified land-use types host high tree species richness (e.g. logged-over forests and old secondary regrowth forest). This has already been shown for various taxa and is a widely accepted phenomenon [
Overall, in our study site, natural forests produce the highest levels of services across the landscape and this would lend support to land management strategies that promote their strict protection [
In addition to the need to preserve natural forests, another challenge is to maximize service production in human-modified land-use areas. With the noticeable exception of tree species richness and tree species composition, our data did not show significant differences in service production between logged-over forests, old rubber gardens and old secondary regrowth forests. Tree species richness is high in logged-over forests (one plot can simultaneously host a mixture of old growth and secondary species), but in such forests, aboveground carbon stocks are low. In contrast, aboveground carbon stocks can be large in some old rubber gardens and secondary regrowth forests, but tree species richness is low. Beyond the diversity of products from these land-use types, there is a complementarity in services produced by such human-modified landscapes. Despite continuous changes in land use and management, the diversification of human activities—farming, rubber tapping, and logging—ensures the sustained delivery of multiple services.
As anticipated, comparing our results with those in the literature emphasized that a mosaic of land uses produces far more services than do rubber or oil palm monocultures: biodiversity is greater [
Our results support the finding that logged-over forests in the study area are better than monocultures in terms of biodiversity conservation and ecosystem service production. Hopefully, the strong case made by this study and many others (e.g. [
We found that carbon storage and tree diversity increased along a successional recovery of the forest after initial clearance: levels of service production were lowest in food crop fields, intermediate in young rubber gardens and secondary regrowth areas, and highest in old rubber gardens and secondary regrowth forests. More striking was no trend for carbon storage in topsoil and the slow recovery of different ecosystem services, either in old rubber gardens or secondary regrowth forests. Mean time since last disturbance for our old secondary regrowth forest plots was 47 years, and yet species richness was still significantly lower than for natural forests. Similarly, we found low similarity in tree species composition between old secondary regrowth forest and natural forest plots.
From a meta-analysis of the recovery of plant biodiversity and carbon stocks in secondary forests, 50 years are enough for species richness to reach natural forest levels, but with only a very low proportion of native forest species (mean value: 26%), even in old stands [
We found topsoil carbon stocks for food crop fields to be half those of natural forest. Our results are consistent with those of a meta-analysis that also showed that soil carbon stocks will eventually fully recover as croplands are allowed to revert to secondary forests [
One study that was also carried out in West Kalimantan found that an increasing number of cycles of cultivation and forest regrowth did not lead to total phosphorus decline, but had detrimental consequences for aboveground carbon sequestration [
In such a rapidly transforming traditional rural landscape in northern Bornean, natural forests host highly unique tree species diversity, have the lowest erosion rate, and store significantly more carbon (in aboveground biomass and topsoil) than do any other land-use type. Logged-over forests provide services similar to natural forest, except for soil erosion control, which is jeopardized by the presence of the remaining decaying road network that leads to soil loss at the landscape level.
All land uses related to the swidden agriculture system largely outperform oil palm or rubber monocultures in terms of tree diversity, carbon storage, and soil erosion control. Natural and logged-over forests should be maintained or managed as an integral part of the swidden system, and landscape multifunctionality should be sustained as a safety net against the price volatility of traded goods (e.g. rubber, palm oil, timber, tengkawang oil), upon which the economy of monocrop systems is much more dependent.
Because of the congruence of services in natural forest, protection of their carbon stocks, for example through financial mechanisms such as REDD+, will synergistically benefit biodiversity and a wide range of other services provided to communities in this area. However, how such mechanisms could benefit communities must be carefully evaluated to counter the high opportunity cost of conversion to monocultures; these may generate greater income, but may also be more detrimental to the production of multiple ecosystem services.
Ecosystem service recovery time following initial slash-and-burn practices on the vegetation is longer in the study area than has been reported in the literature for similar study situations. As rotation length appears to be a key factor in the sustainability of swidden systems, it is critical to understand the socio-cultural and economic drivers of the reduction in rotation length and the potential feedback of this reduction on social–ecological systems. In the rapidly transforming socio-environmental context of this region, questions remain about the long-term persistence of the swidden agriculture system.
Rarefaction level was the number of individuals surveyed in old rubber gardens (n = 449). Twelve plots (0.48 ha in total) were surveyed for fallows and rubber gardens (both young and old), and 25 plots (1 ha in total) for logged-over and natural forest.
(TIFF)
No vegetation plot was selected in food crop fields under the assumption that tree diversity and aboveground carbon would be null. The first plot size dimension is the dimension along the slope.
(XLSX)
Longitude and latitude of each plot are provided in decimal degrees (WSG84 datum). Plot mean diameter at breast height (DBH), height, wood specific gravity (WSG) and aboveground carbon (AGC) are also provided.
(XLSX)
Longitude and latitude of each plot are provided in decimal degrees (WSG84 datum). Plot annual soil loss (ASL) and topsoil carbon (TSC) are also provided. ASL values in grey-tinted cells were discarded for analyses.
(XLSX)
We used original indicator values when the distribution was normal and log10-transformed values otherwise. The vegetation set did not include food crop field plots under the assumption that tree diversity and aboveground carbon would be null. Depicted values are either test statistics (for Moran’s I, Lagrange Multiplier and the selected model) or model coefficients, and are presented along with information on statistical significance. For the Berger-Parker index, despite some spatial auto-correlation, the spatial dependence coefficient of the spatial error model was not significant. We therefore used a regular linear model and acknowledge that results for this indicator might be slightly biased due to spatial auto-correlation.
(XLSX)
Data from plots under the same land-use type (12 for young and old rubber gardens and fallows and 25 for logged-over and natural forests) were pooled together. IUCN conservation status is provided for each species (CR = Critically Endangered; EN = Endangered; VU = Vulnerable; LR/cd = Lower Risk: Conservation Dependent; LR/nt = Lower Risk: Near Threatened; DD = Data Deficient; LC or LR/lc = Least Concern). In order to be conservative, any species identified to the genus level only (e.g.
(XLSX)
We thank all the villagers from Keluin for their hospitality, availability, and precious help in data collection. We are especially grateful to Bapak Lasa, Bapak Bandan and Bapak Tinggi for sharing their thorough knowledge on plant denomination and use. We sincerely acknowledge Bapak Engkamat for his unfailing help and support in carrying out field work. We deeply thank Imam Basuki for his advices on soil erosion monitoring and help in the field in setting erosion plots. We express gratitude to Bapak Ismail Rachman from the Herbarium Bogoriense for voucher identification. We acknowledge Ervan Rutishauser for his advice on statistical analysis. We are also grateful to Mike Lawes and two anonymous reviewers for their insightful comments and suggestions on earlier versions of the manuscript.