Adaptive irrigation management by Balinese farmers reduces greenhouse gas emissions and increases rice yields

Abstract

The potential for changes in water management regimes to reduce greenhouse gases (GHG) in rice paddies has recently become a major topic of research in Asia, with implications for top-down versus bottom-up management strategies. Flooded rice paddies are a major source of anthropogenic GHG emissions and are responsible for approximately 11% of global anthropogenic methane (CH₄) emissions. However, rice is also the most important food crop for people in low- and lower-middle-income countries. While CH₄ emissions can be reduced by lessening the time the plants are submerged, this can trigger increased emissions of nitrous oxide (N₂O), a more potent GHG. Mitigation options for CH₄ and N₂O are different, and minimizing one gas may increase the emission of the other. Accurate measurement of these gas emissions in rice paddies is difficult, and the results are controversial. We analysed these trade-offs using continuous high-precision measurements in a closed chamber in 2018-2020. Based on the results, we tested a bottom-up adaptive irrigation regime that improves nitrogen uptake by rice plants while reducing combined GHG emissions and nitrogen runoff from paddies to reefs in agricultural drainages. In 2023, we undertook a follow-up study in which farmers obtained higher rice yields with adaptive intermittent irrigation compared to uniformly flooded fields. These results use the polycentric, self-governing capacity of Balinese subaks for continuous adaptation. This article is part of the theme issue 'Climate change adaptation needs a science of culture'.

Keywords: Bali; climate; commons; greenhouse gases; rice.

Figure 1.

Experiments were conducted in Subak Bene, a water user community of 201 farmers and 70.3 ha located in Kecamatan Marga, Tabanan Bali. (a) Sentinel satellite image of some INT fields, 28 March 2023. (b) Experimental set-up with the Picarro gas analyser. The enclosed chamber has an input and output; the output tubing from the chamber is connected to the Picarro sensor and the input is connected to the instrument's vacuum pump, which circulates the air within this system. Gases emitted from the plant and soil are drawn from the chamber and pumped into the G2508 cavity ringdown spectrometer where the gas concentrations are measured every 2 s. Sampled gas exiting the instrument returns to the chamber. (c,d) Water levels and fertilizer additions. Water level remained above ground level for the entire growing cycle in the treatment field FLD both in 2018 and 2020. Negative levels for INT fields were the depth of the moisture horizon. Fertilizer was added to all fields the day prior to transplanting, and on days 14 and 21.

Figure 2.

Net emission rates of CH₄ (a) and N₂O (b) for 2 years in continuously flooded (FLD) and intermittently wetted (INT) experimental plots. Note the ordinate axis for CH₄ (milligrams) is 1000× greater than that for N₂O (micrograms). Only the middle time of the planting cycle was sampled in 2018, shown as lines with symbols. Solid lines are the mean of triplicate plots in 2020; a single value off scale high or low pulls the mean outside the visible points. Sampling days are for 2020, and 2018 data are plotted on the same scale. Site FLD 2 on 2020 day 8 showed the highest net CH₄ rate we measured of 1174 mg CH₄ m⁻² h⁻¹ (off scale). Site INT 1 on 2020 day 21 showed the highest net N₂O rate we measured of 8666 µg N₂O m⁻² h⁻¹ (off scale).

Figure 3.

Detailed time series of 1 day's incubations. Left side: time series of raw data for CH₄ concentration (ppm) from the Picarro gas analyser are shown for triplicate deployments of the sample chamber at two sites, the experimental (INT) and control (FLD). Continuous data records are colour coded: chamber in transit (green); data subject to placement artefact ‘SKIP’ not analysed (black); sample data control plot FLD (blues); sample data experimental treatment plot FLD (browns). Right side: different modes of emission rate are accentuated by swarm dot-plots of 10–15, two-minute averages rates plotted around the overall net rate (in raw units; swarms below are converted to meaningful rates) (electronic supplementary material, op. cit. offers more details).

Figure 4.

CH₄ detailed rates. Swarms—dot plots of 2 min average partial rates for incubation records—show variability of methane flux rates over time in rice fields with two irrigation regimes. An experimental field was intermittently wetted (INT, brown dots are triplicates) and a control field was continuously flooded (FLD, blue dots are triplicates). Net rates are shown by solid lines (from figure 2a) overlaid on each swarm. (a) The 2018 pilot study sampled one site in each treatment field during the middle of the growing season. (b) In 2020, triplicate sites were sampled starting the day after sowing in both fields. Yellow bands group the samples taken on the same day, two in 2018 and six in 2020.

Figure 5.

Summary of experimental differences in the flux of two GHGs over one growing season for a water protocol of intermittent brief flooding (INT, browns) and a control site of continuous flooding (FLD, blues). Solid circles are time-weighted integrals of net rates from figure 2 for triplicate stations in one control and one experimental field in 2020. Open circles are a single site for a different part of the growing season in 2018. N₂O fluxes are expressed as methane equivalents that are an approximate theoretical measure of global warming potential (GWP). Significant reductions in GHG impact were achieved by the experimental INT regime, 85% reduction based on these seasonal integrals. An insignificant increase in N₂O flux was well below the impact of CH₄ reduction.