Rice is one of the most important food crops globally, with more than half of the world population fed with rice [1
]. Global rice production is projected to increase from 473 million tonnes in 1990 to at least 781 million tonnes by 2020 [2
]. Paddy fields in China account for 23% of all cultivated land and nearly 20% of the global rice production [3
]. High doses of chemical fertilizers have been used in rice cultivation to increase production in order to meet the increasing demand [4
]. The long-term use of chemical fertilizers, however, acidifies the soil and compromises the sustainability of paddy production [5
]. The excessive use of fertilizers also increases the risk of pollution [6
] and may generate nutrient imbalances in soils and crops, particularly between nitrogen (N) and phosphorus (P) [7
]. The use of green fertilizers such as farmyard manure [8
] and crop straw [9
] has been strongly promoted in recent years as substitutes for, or to reduce the use of, industrial fertilizers in an effort to develop a more sustainable rice production. The return of crop straw has particularly been promoted, because straw is an economical and important source of organic matter and nutrients [10
]. The biotic and abiotic decomposition of straw cellulose and hemicellulose releases N, P, and potassium [11
]. The application of straw can also increase soil carbon (C) storage [12
] to help mitigate global climate change [13
Straw is currently returned to farmland soil after harvesting, but this practice of soil fertilization and amendment can be problematic. If rice straw decomposes slowly, its residual presence can impede the growth of rice shoots after transplantation to paddies [14
] and can affect the total production of the paddies. Optimal strategies for straw application must thus be determined for sustaining nutrient supplies for rice growth and for reducing the inhibition of rice seedlings [15
]. Various practices of straw application and incorporation into the soil have been tested, including incorporation with tillage at different depths [16
], various proportions of straw incorporation with conventional fertilizers [17
], and straw incorporation combined with water-management strategies [10
]. These practices, however, have not accounted for the different properties of paddy ridges and ditches.
Plant growth in Chinese wetlands is generally N limited [18
]. Nutrient limitation is especially significant in paddy fields, likely because the periodic inundation of the soil limits the access of plants to soil nutrients by the effects of anoxia on root growth [19
], the reduction of mineralization rates [20
], and the increase in leaching, particularly of N [21
]. Rice and vegetable crops are commonly rotated in many provinces of China, such as Fujian, Jiangxi, and Zhejiang. Studying the relationship between crop rotation and the return of rice straw is thus very important. The cultivation of a vegetable crop after a late rice harvest usually requires the construction of a ridge and ditch structure on the soil surface, which potentially generates two different microenvironments where rice straw can be applied. The decomposition of the straw and the release of nutrients will thus likely differ between these microenvironments due to potential differences in soil temperature and salinity, but these differences are not yet known. The elucidation of these unknowns could substantially improve rice production with sustainable practices. In southern China, the common practice consists of growing one crop in each of three growing seasons, including two successive rice crops (early and late) followed by a vegetable (lettuce) crop, with intervening periods of drainage [22
]. This management is applied to 56% of the 0.3 million km2
of China rice croplands [3
]. This management system with rotation systems is also widely used in all South Asia, mainly in sites with a pronounced dry season. Thus, a change in rice straw management to improve fertilization has great potential for enhancing rice production.
This study determined (i) the changes in the C:N:P stoichiometry and mass of the residual rice straw and their relationships with other soil variables (temperature, pH, and salinity) during decomposition in the ridges and ditches, and (ii) the capacity of the straw applied at the beginning of the crop rotational cycle to release N and P in the ridges and ditches during rice growth.
4. Materials and Methods
4.1. Study Site
All field experiments were performed in the Wufeng Agronomy Field of the Fujian Academy of Agricultural Sciences (26.1° N, 119.3° E; Figure 7
) in subtropical southeastern China. This field was managed following the common practice of growing one crop in each of three growing seasons, including two successive rice crops (early and late) followed by a vegetable (lettuce) crop, with intervening periods of drainage [22
]. The early rice crop is grown from March-April to June-July, the late rice crop is grown from July-August to November-December, and the vegetable crop is grown from November–December to February-March. During this last period ditches are flooded during some periods, whereas ridges are never flooded. Moreover, during this period plastic sheeting is used to cover the ridges but not the ditches. This practice increases the temperature and reduces the weed impact, thus improving the vegetable growth. Chemical fertilizer (N, P2
, and K2
O at a rate of 200, 158, and 141 kg ha−1
, respectively) was applied once to the vegetable crop on 17 December 2011. Fertilizer was applied to dry soil. The other two additional fertilizations were applied to flooded water. Chemical fertilizers were applied in three splits with different nutrient loadings using a mix of complete fertilizer (N:P2
O = 16%:16%:16%, Keda Fertilizer Co., Ltd., Jingzhou, China) and urea fertilizers (46% N). Chemical fertilizer (including N, P2
, and K2
O at a rate of 95, 70, and 70 kg ha−1
, respectively) was applied to the rice crops before transplantation, at the tillering stage, and at the panicle-formation stage. The three dates of fertilization were 8 April 2012, 20 April 2012, and 12 June 2012. The field was ploughed to a depth of 15 cm with a moldboard plow. The plough dates for the vegetable crop and the early rice crop were 10 December 2011 and 8 April 2012. The rice variety was Hesheng 10, which is the rice variety commonly cultivated in southeast China, and the lettuce variety was Kexing 5, also frequently used in southeast China. The spacing between individual plants was 14 × 28 and 40 × 60 cm, respectively.
The soil of the paddy field was moist, poorly drained, and had a sand:silt:clay content of 28:60:12 [28
]. The bulk density of the soil prior to this study was 1.1 g cm−3
. The soil pH (1:5 with H2
O) was 6.5, and the concentrations of organic carbon, total N, and total P were 18.1, 1.2, and 1.1 g kg−1
, respectively [28
]. The water level was maintained at 5–7 cm above the soil surface in the rice crops before the late tillering stage, and then drained for the control of non-productive tillering [38
]. After about one week, the paddy field was re-flooded, kept alternately wet and dry, and then drained again two weeks before rice harvest.
4.2. Experimental Design
The rice straw used in the experiment was collected from the late rice crop. The straw-decomposition experiment used nylon-mesh bags [39
]. Each bag was 20 × 20 cm with a pore size of 1 mm and contained 13 g of straw, and the bags were placed on top of the soil. The experiment began on 17 December 2011 during the vegetable crop season. The field contained two microhabitats during this period, the ditches and ridges, which provided the two treatments of this decomposition experiment (Figure A1
), with three replicates each. The ridges were 40 cm apart with heights and widths of 15 cm, which is typical for this area. Straw samples were collected 10, 30, and 60 days after straw application during the vegetable crop (17 December 2011 to 8 March 2012); 90 days after straw application during the first fallow period (8 March to 11 April 2012); 120, 150, and 180 days after straw application during the early rice crop (11 April to 13 July 2012); and 210 days after straw application during the second fallow period (13 July to 31 July 2012). The experiment thus consisted of two treatments (habitats) × eight sampling times × three replicates = 48 sample bags.
4.3. Sample Collection and Analysis
Three samples (one from each replicate) were randomly collected from each treatment on each sampling date. The litter from each nylon bag was gently washed with water and subsequently oven-dried to a constant mass (65 °C for 24–36 h) and weighed. These dried and cleaned samples were then finely ground in a ball mill. The C and N concentrations of the dried litter were determined using a Vario EL III Elemental Analyzer (Elemental Scientific Instruments, Hanau, Germany). The P concentration of the litter was measured using the molybdate-blue reaction [40
] with a UV-2450 spectrophotometer (Shimadzu Scientific Instruments, Kyoto, Japan).
Soil salinity (mS cm−1), pH, and temperature were measured in situ on each sampling date at a depth of 20 cm. Soil pH and temperature were measured with a pH/temperature meter (IQ Scientific Instruments, Carlsbad, CA, USA), and soil salinity was measured using a 2265FS EC Meter (Spectrum Technologies Inc., Paxinos, PA, USA).
4.4. Statistical Analyses
We analyzed the changes in elemental composition and stoichiometry during litter decomposition in the two habitats (ridges and ditches) at the various sampling times (after 10, 30, 60, 90, 120, 150, 180, and 210 days). Litter C, N, and P concentrations; C:N, C:P, and N:P ratios; and C, N, and P remaining (% of initial respective amount) during the studied period of litter decomposition were the dependent variables, while habitat (ridges and ditches) represented the fixed independent factor with repeated measures along time of sampling and plots as random factors. We used the “nlme” [41
] R package with the “lme” function. We chose the best model for each dependent variable using Akaike information criteria. We used the MuMIn [42
] R package in the mixed models to estimate the percentage of variance explained by the model.
We also studied the effect of time by the crop period (vegetable: from 0 to 60 days, fallow: 90 days, rice crop: from 120 to 180 days, fallow: 210 days). Pearson correlation analyses identified the relationships among the rate of litter decomposition and nutrient release with C, N, and P concentrations; C:N, C:P, and N:P ratios; and soil factors. These univariate analyses were performed using SPSS 13.0 (SPSS Inc., Chicago, IL, USA).
We also performed multivariate statistical analyses by using general discriminant analysis (GDA) to determine the overall differences between ridges and ditches in the changes of total soil C, N, and P concentrations; soil C:N, C:P, and N:P ratios; straw mass; residual C, N, and P concentrations; and soil salinity, pH, and temperature during straw decomposition. We also assessed the component of the variance due to the sampling day as an independent categorical variable. GDA is thus an appropriate tool for identifying the variables most responsible for the differences among groups while controlling the component of the variance due to other categorical variables. This analysis used the Squared Mahalanobis Distance statistic that depends on the Euclidean distance in the model between two sets of samples; if the sets were closer or less different, the squared Mahalanobis distance would be lower, and if the sets were more distant or more different, and the squared Mahalanobis distance would be higher [43
]. Soil C:N, C:P, and N:P ratios were calculated as mass ratios. GDA was performed using Statistica 6.0 (StatSoft, Inc., Tulsa, OK, USA).
We used structural equation modeling (SEM) to analyze the factors explaining the maximum variability of the biomass; residual straw C, N, and P concentrations; soil C, N, and P concentrations; and soil C:N, C:P, and N:P ratios throughout the study period as functions of the habitat and the other soil traits. This analysis provides information on the direct, indirect, and total effects of the variables. We fit the models using the sem R package [44
] and acquired the minimally adequate model using the Akaike information criterion. Standard errors in addition to the significance levels of the direct, indirect, and total effects were calculated by bootstrapping (1200 repetitions).