Estimating net primary production and annual plant carbon inputs, and modelling future changes in soil carbon stocks in arable farmlands of northern Japan

Soil C sequestration in croplands is deemed to be one of the most promising greenhouse gas mitigation options for Japan's agriculture. In this context, changes in soil C stocks in northern Japan's arable farming area over the period of 1971–2010, specifically in the region's typical Andosol (volcanic ash-derived) and non-Andosol soils, were simulated using soil-type-specific versions of the Rothamsted carbon model (RothC). The models were then used to predict the effects, over the period of 2011–2050, of three potential management scenarios: (i) baseline: maintenance of present crop residue returns and green manure crops, as well as composted cattle manure C inputs (24–34 Mg ha−1 yr−1 applied on 3–55% of arable land according to crop), (ii) cattle manure: all arable fields receive 20 Mg ha−1 yr−1 of composted cattle manure, increased C inputs from crop residues and present C inputs from green manure are assumed, and (iii) minimum input: all above-ground crop residues removed, no green manure crop, no cattle manure applied. Above- and below-ground residue biomass C inputs contributed by 8 major crops, and oats employed as a green manure crop, were drawn from yield statistics recorded at the township level and crop-specific allometric relationships (e.g. ratio of above-ground residue biomass to harvested biomass on a dry weight basis). Estimated crop net primary production (NPP) ranged from 1.60 Mg C ha−1 yr−1 for adzuki bean to 8.75 Mg C ha−1 yr−1 for silage corn. For the whole region (143 × 103 ha), overall NPP was estimated at 952 ± 60 Gg C yr−1 (6.66 ± 0.42 Mg C ha−1 yr−1). Plant C inputs to the soil also varied widely amongst the crops, ranging from 0.50 Mg C ha−1 yr−1 for potato to 3.26 Mg C ha−1 yr−1 for winter wheat. Annual plant C inputs to the soil were estimated at 360 ± 45 Gg C yr−1 (2.52 ± 0.32 Mg C ha−1 yr−1), representing 38% of the cropland NPP. The RothC simulations suggest that the region's soil C stock (0–30 cm horizon), across all soils, has decreased from 13.96 Tg C (107.5 Mg C ha−1 yr−1) in 1970 to 12.46 Tg C (96.0 Mg C ha−1 yr−1) in 2010. For the baseline, cattle manure and minimum input scenarios, soil C stocks of 12.13, 13.27 and 9.82 Tg C, respectively, were projected for 2050. Over the period of 2011–2050, compared to the baseline scenario, soil C was sequestered (+0.219 Mg C ha−1 yr−1) by enhanced cattle manure application, but was lost (−0.445 Mg C ha−1 yr−1) under the minimum input scenario. The effect of variations of input data (monthly mean temperature, monthly precipitation, plant C inputs and cattle manure C inputs) on the uncertainty of model outputs for each scenario was assessed using a Monte Carlo approach. Taking into account the uncertainty (standard deviation as % of the mean) for the model's outputs for 2050 (5.1–6.1%), it is clear that the minimum input scenario would lead to a rapid decrease in soil C stocks for arable farmlands in northern Japan.

► We estimated plant C inputs, using yield statistics and allometric relationships.
► Plant C inputs in croplands of northern Japan were 2.52 Mg C ha−1 y−1 (38% of NPP).
► We simulated changes in soil C stocks with the estimated plant C inputs with RothC.
► Use of the plant C inputs allowed more accurate simulations of soil C stocks.
► Uncertainty analysis showed the minimum input scenario would cause a large C loss.