Mechanical weeding enhances ecosystem multifunctionality and profit in industrial oil palm

Oil palm expansion in Southeast Asia, particularly Indonesia, has been associated with substantial deforestation and environmental costs, yet industrial plantations remain economically attractive due to high short-term returns and global demand. Industrial plantations (over 50 ha) achieve ~50% higher productivity than smallholders but rely on high fertilizer and herbicide inputs, potentially impairing ecosystem multifunctionality. The need for sustainable management focuses on optimizing fertilization rates and weeding methods to maintain yields and profitability while minimizing negative impacts on ecosystem functions and biodiversity. Over-fertilization increases nutrient leaching, greenhouse gas emissions, and can reduce soil quality and profitability. Herbicide-based weed control simplifies understory vegetation, reduces habitat complexity, and can promote herbicide-resistant weeds. Mechanical weeding may allow rapid understory recovery, improving nutrient cycling and biodiversity, but could increase labor costs or competition for water. This study presents a multiyear, spatially replicated 2×2 factorial field experiment in a mature industrial oil palm plantation (Jambi, Indonesia) comparing conventional (high fertilization, herbicide) versus reduced management (reduced, compensatory fertilization; mechanical weeding). The hypothesis was that reduced fertilization and mechanical weeding would enhance ecosystem functions and biodiversity while maintaining high productivity and increasing profit.

Background literature highlights that industrial oil palm management commonly relies on high fertilizer and herbicide use to optimize yields, with documented environmental trade-offs including elevated nutrient leaching, greenhouse gas emissions, reductions in soil microbial biomass, soil acidification, altered microbial communities, and biodiversity losses. Excess fertilization can reduce profitability as yields saturate while input costs rise; aligning fertilization with nutrient export via harvest and using organic inputs can improve nutrient retention and profitability. Herbicide control reduces understory diversity, may favor invasive herbicide-resistant species, and degrades habitat complexity affecting vegetation-dwelling biodiversity. Mechanical weeding preserves root systems, enabling rapid vegetation regrowth, higher understory cover and diversity, potentially improving nutrient cycling and habitat for fauna, though potentially at higher labor costs and with perceived risks of plant competition, especially under drought. Prior studies often examined single functions or short-term responses; a holistic, multiyear, replicated assessment of management impacts on multiple ecosystem functions, biodiversity, and economic outcomes in industrial oil palm has been lacking, motivating this experiment.

Study area and design: The experiment was conducted in a state-owned industrial oil palm plantation (PTPN VI) in Jambi, Indonesia (1.719° S, 103.398° E; 73 m asl) on Acrisol soils (sandy clay loam), with palms planted 1998–2002 at 142 palms ha⁻¹ (8 m spacing) and ≥16 years old during 2016–2020. Climate averages: 27.0 ± 0.2 °C; 2,103 ± 445 mm yr⁻¹ rainfall.

Experimental treatments: A full-factorial 2×2 design established November 2016 compared fertilization (conventional vs reduced) and weeding (herbicide vs mechanical). Four treatments: (1) conventional fertilization + herbicide, (2) reduced fertilization + herbicide, (3) conventional fertilization + mechanical weeding, (4) reduced fertilization + mechanical weeding. Each treatment had four 50 m × 50 m plots (16 plots total) randomized in four blocks; measurements were taken in the inner 30 m × 30 m to avoid edge effects. Management zones within plots: palm circle (2 m radius; weeded 4× yr, raked before fertilization), frond piles (every second inter-row), and inter-rows (weeded 2× yr; no fertilizer).

Fertilization: Conventional rates typical for large-scale plantations: 260 kg N – 50 kg P – 220 kg K ha⁻¹ yr⁻¹. Reduced rates compensated nutrients exported via harvest: 136 kg N – 17 kg P – 187 kg K ha⁻¹ yr⁻¹. Fertilizers applied twice yearly (April, October) after palm-circle weeding and raking. Dolomite (426 kg ha⁻¹ yr⁻¹) and micronutrients (Micro-Mag 142 kg ha⁻¹ yr⁻¹ with 0.5% B₂O₃, 0.5% CuO, 0.25% Fe₂O₃, 0.15% ZnO, 0.1% MnO, 18% MgO) were applied in all treatments per common practice.

Weeding: Herbicide treatment used glyphosate: palm circle 1.50 L ha⁻¹ yr⁻¹ in four applications; inter-rows 0.75 L ha⁻¹ yr⁻¹ in two applications. Mechanical weeding used a brush cutter in the same zones and frequencies as herbicide.

Measured outcomes and indicators (2016–2020):

• Ecosystem functions (8), each with multiple indicators aggregated at plot level:

  1. Greenhouse gas (GHG) regulation: net ecosystem productivity (NEP; fruit harvest subtracted), soil organic carbon (SOC, to 50 cm), soil CO₂, CH₄, N₂O fluxes measured monthly Jul 2019–Jun 2020 via static chambers; NEP derived from net ecosystem C exchange components (above- and belowground productivity via oil palm allometry, frond litter inputs, fruit C; heterotrophic respiration from soil CO₂ partitioning and frond litter decomposition rates).
  2. Erosion prevention: understory vegetation cover (annual surveys 2016–2020; five subplots per plot), expressed as ratio to 2016 baseline.
  3. Organic matter decomposition: litterbag mass loss (8-month incubation starting Dec 2016) and soil animal decomposer activity (from soil arthropods).
  4. Soil fertility: gross N mineralization (15N pool dilution), effective cation exchange capacity (ECEC), base saturation, microbial biomass N (top 5 cm; Feb–Mar 2018); area-weighted across management zones.
  5. General pollination potential: abundance of pan-trapped arthropods (yellow pan traps; Nov 2016, Sep 2017, Jun 2018; ratios to 2016) and nectar-feeding bird activity.
  6. Water filtration: leaching losses of major elements from pore water at 1.5 m depth via suction cup lysimeters monthly over 2017–2018; area-weighted across zones.
  7. Plant refugium: percentage cover and richness of invasive understory plants (non-native to Sumatra among annual top-10 dominants) annually 2016–2020; expressed relative to 2016.
  8. Biological control: activities of insectivorous birds and bats, and soil arthropod predator activity.

• Biodiversity (multitrophic taxonomic richness across 7 groups): Understory vascular plant species richness (annual 2016–2020), soil microorganism richness (DNA/RNA co-extraction; 16S rRNA v3–v4; May 2017), soil arthropod order richness (Oct–Nov 2017; heat-gradient extraction), aboveground arthropod order and insect family richness (pan traps 2016–2018; sweep nets and Malaise trap in Jun 2018), insectivorous bird and bat species/morphospecies richness (acoustic surveys, Sep 2017; diet traits from Elton database).

• Economic indicators (2017–2020 cumulative): yield (harvested fresh fruit bunches weighed per palm in inner 30×30 m and scaled to ha), lower fifth quantile yield per palm per plot, yield shortfall probability (share of palms below 75% of average yield; thresholds: cumulative 630 kg per palm; two-year 300 kg per palm per year), material costs (fertilizers, herbicide, gasoline), labor costs (minimum wage basis; recorded 2017 for harvesting, fertilizing, weeding including annual manual removal of Clidemia hirta), total management cost, profit (revenues − total costs), relative gross margin (gross profit share of revenues). Prices and costs were held constant across years to isolate treatment effects; no discounting applied.

Data processing and statistics: Indicator values were z-standardized at plot level; direction inverted for undesirable high values (NEP, N₂O/CH₄ fluxes, element leaching, invasive cover, yield shortfall, management costs). Ecosystem multifunctionality computed via (i) average of ecosystem function means (z) and (ii) threshold approach (number of functions exceeding 10–90% of maximum performance per function; maximum as mean of top three values per indicator). Linear mixed-effects models tested fertilization, weeding, and interaction effects on individual ecosystem functions and biodiversity (indicators and trophic groups as random effects; interactions tested), and on multifunctionality (ecosystem function as random effect). Linear models assessed yield and economic indicators. Analyses used R 4.0.4 with nlme and influence.ME.

• Ecosystem multifunctionality: Mechanical weeding increased average multifunctionality compared to herbicide (LME weeding effect F1,12=5.98, P=0.03). Threshold multifunctionality showed more functions above 70% and 90% performance under mechanical weeding (P=0.04 for both thresholds), with marginal effects at 50% and 80% thresholds. Individual ecosystem functions did not differ significantly by treatment, but higher multifunctionality under mechanical weeding reflected higher mean z-values in litter decomposition, soil fertility, water filtration, and plant refugium indicators.
• Biodiversity: Multitrophic taxonomic richness was higher with mechanical weeding than herbicide (LME weeding effect F1,12=8.72, P=0.01). Understory plant species richness increased strongly under mechanical weeding across 2017–2020; overall, 33% more understory plant species occurred under reduced management than conventional. Mechanical weeding increased ground cover of dominant species except the invasive Clidemia hirta, whose cover decreased by 55% (P=0.01).
• Yield and stability: Four-year cumulative yield did not differ among treatments; lower fifth quantile yield and yield shortfall probability (below 75% of average) were similar across treatments, indicating no negative effect of mechanical weeding or reduced fertilization on worst-case yields. No delayed yield or profit effects were detected when analyzed separately for 2017–2018 and 2019–2020.
• Costs and profitability: Material costs were 41% lower under reduced fertilization with mechanical weeding than conventional fertilization with herbicide; LM showed strong treatment effects on material cost (fertilization and weeding P<0.01) and total management cost (fertilization and weeding P<0.01). Labor costs were 10% higher under mechanical weeding (LM weeding effect F1,12=5.80, P=0.03). Despite slightly higher labor, profit was 12% higher under mechanical weeding than herbicide (LM weeding effect F1,12=4.84, P=0.05). Relative gross margin increased by 11% under reduced fertilization with mechanical weeding (fertilization and weeding effects P<0.01). Average annual yield under reduced management (29.6 Mg ha⁻¹ yr⁻¹) approached the reported attainable yield for large-scale plantations (30.6 Mg ha⁻¹ yr⁻¹). A previously reported 56% decrease in dissolved N losses under reduced fertilization with mechanical weeding (relative to conventional) supports improved nutrient retention.
• Sensitivity insights: Reduced fertilization outperformed conventional in profit if fertilizer prices doubled; profitability differences were robust to plausible increases in labor costs (herbicide with reduced fertilization surpassing mechanical weeding only at ~1,250% labor cost increase).

The study demonstrates that replacing herbicide with mechanical weeding, combined with reduced, compensatory fertilization, enhances ecosystem multifunctionality and biodiversity without compromising yield or yield stability in mature industrial oil palm on mineral soils. The increased multifunctionality under mechanical weeding, masked when examining single functions, underscores the importance of integrative assessment across functions. Mechanisms likely include rapid understory regrowth preserving root systems, increasing plant cover and diversity, improving microclimate and organic matter inputs, thereby supporting soil processes such as decomposition, nutrient retention, and water filtration, and offering greater plant refugium. Biodiversity gains, notably in understory plants, translate to improved habitat complexity with potential cascading effects across trophic levels. Economically, reduced fertilization substantially lowered material and total management costs, and together with maintained yields, increased profit and relative gross margin despite modest increases in weeding labor. Risk analyses indicated no greater likelihood of low yields under reduced management, even through a dry period (late 2018–early 2019 ENSO). The findings align with sustainability goals by reducing nutrient leaching, water pollution risks, and avoiding herbicide impacts on non-target biota and worker health, supporting certification guidelines (e.g., RSPO). Legacy effects from >16 years of prior conventional management and the relatively short experimental duration may have dampened detectable responses in individual ecosystem indicators, yet the multifunctionality response emerged within four years, suggesting further benefits may accrue over longer time frames. The results support reduced management as a viable, win–win strategy in mature plantations, with potential broader applicability subject to site and plantation age considerations.

This multiyear, replicated field experiment provides evidence that mechanical weeding paired with reduced, compensatory fertilization enhances ecosystem multifunctionality and biodiversity while maintaining high yields and improving profitability in mature industrial oil palm on mineral soils. Key contributions include demonstrating multifunctionality gains that are not apparent in single-function analyses, biodiversity improvements (notably understory plants), and clear cost savings translating to higher profit and relative gross margin. These management adjustments can reduce environmental impacts (nutrient leaching, herbicide exposure) and align with sustainable certification practices. Future research should: (1) extend monitoring to capture long-term ecological responses beyond legacy effects; (2) test generality across soil types, climates, and landscape contexts; (3) evaluate responses in younger plantations with different nutrient dynamics; (4) incorporate market variability (produce and input prices) and labor constraints into economic risk assessments; and (5) quantify broader ecosystem services and potential trade-offs, including pest dynamics and water use under varying climatic conditions.

• Duration and legacy: The experiment spanned four years following ≥16 years of conventional management, likely inducing legacy effects that dampened short-term responses of individual ecosystem indicators.
• Timing of measurements: Some indicators were measured within one to two years of treatment initiation, potentially too early to detect changes.
• Economic assumptions: Revenues and costs were held constant across years, excluding market price fluctuations; profits were not discounted, and only yield-related production risks were considered.
• Labor measurement: Labor times were recorded by researchers observing operations; while biases are expected to be smaller than treatment differences, they cannot be fully excluded.
• Scope and generalizability: Findings pertain to mature plantations on mineral Acrisol soils; applicability to younger plantations or other soil types may differ due to nutrient accumulation dynamics and management needs.
• Resource constraints: Mechanical weeding increases labor hours (≈12 extra man-hours ha⁻¹ yr⁻¹), which may be relevant under labor shortages.