This study aimed to explore the effects of different residue return methods on nitrogen fractions, nitrogen mineralization and nitrogen-cycling genes in black soil of Northeast China, and to clear the soil nitrogen supply capacity and the change of soil nitrogen cycling gene community structure under long-term residue return.

Based on the long-term experiment of black soil in Northeast China, the residue incorporated into soil (RI) and the residue covered on soil surface (RC) under monoculture maize were selected, with residue removed as control (CK). Nitrogen content in soil fractions were measured, soil nitrogen mineralization incubation was conducted by using leaching incubation at intervals, and fluorescence quantitative PCR (qPCR) was used to determine the copy number of nitrogen-cycling genes in soil.

After 8-year experiment, compared with CK, RC significantly increased the content of particulate organic nitrogen (PON)(0.21 g·kg^-1) and mineral-associated organic nitrogen (MAON) (0.27 g·kg^-1) in surface (0-5 cm) soil, whereas RI only increased the content of MAON (0.13 g·kg^-1) in soil (P<0.05). Residue return (RI and RC) markedly increased the microbial biomass nitrogen (MBN) in soil by 1.4-2.8 times (P<0.05), the RI had higher content of ammonium nitrogen (NH₄⁺) and dissolved organic nitrogen (DON), while the RC had the lowest content of nitrate nitrogen (NO₃^-). In comparison with CK, residue return significantly enhanced soil nitrogen mineralization amount by 25.3%-83.2% (P<0.05), taking the descending order of RC>RI>CK. Residue return remarkably increased the potential of soil nitrogen mineralization (N₀) and mineralization rate constant (k) (P<0.05) by using a first-order reaction kinetics model, both showing the highest values under RC, with N₀ and k reached 199.8 mg·kg^-1 and 0.31 mg·kg^-1·d^-1, respectively. Random forest analysis indicated that PON, MBN, and NO₃^- had greater impacts on N₀. In addition, the abundance of nifH, AOB and nirS genes under residue return were enhanced and the abundance of AOA and nirK genes under residue return were declined in comparison with residue removed (P<0.05), which indicated that residue return could change the structure of soil nitrogen-cycling genes communities. Redundancy analysis (RDA) result showed that the changes of soil microbial community structure were affected by different nitrogen fractions under different residue return methods.

Long-term residue covered on soil surface had the highest organic nitrogen content and nitrogen mineralization potential in soil. It was beneficial to improve soil nitrogen pools and to ensure the supply of nitrogen required for plant growth, which provided greater possibility for reducing the application of chemical nitrogen fertilizer in cropland in black soil of Northeast China.

Conservation tillage is an important measure for restoring and enhancing soil fertility, and soil organic carbon (SOC) plays a crucial role in maintaining soil fertility, supporting crop growth, and protecting soil environment. However, there is currently a lack of long-term monitoring platforms for conservation tillage in China, so using modeling methods can help study SOC dynamics under long-term conservation tillage.

A long-term tillage experiment was established in the black soil region in 2001 with three treatments: no-tillage (NT), moldboard plow (MP), and ridge-tillage (RT). The structures and parameters of process-based models (RothC, AMG model) and statistical models (MLPNN model) were optimized. The changes in SOC under long-term conservation tillage were simulated and compared. The effectiveness of different models in simulating and predicting the SOC dynamics under conservation tillage was evaluated, and the long-term response and influencing factors of SOC in black soil in Northeast China to conservation tillage were revealed.

After optimizing parameters for carbon pool allocation, errors of the RothC and AMG models were significantly reduced. During the first 11 years (2001-2012) of conservation tillage, there was no significant difference in the simulation of SOC between RothC and AMG models, indicating that the structural complexity of process models does not have significant impacts on the simulation results for relatively short term. The simulation results of the statistical model MLPNN were similar to process models, proving the application of statistical models in small-scale regions. Over the next 100 years, RothC and AMG models predicted similar trends in SOC changes, but AMG model significantly overestimated the increase in SOC stocks, which may be attributed to SOC saturation and the influence of tillage practices. Both RothC and AMG models showed high sensitivity to carbon input, but they responded differently to climate and soil factor changes.

It is necessary to choose appropriate models based on local conditions while using models to simulate SOC in long-term conservation tillage. For short-term prediction of SOC under conservation tillage, a relatively simple AMG modelcan be used, while for long-term prediction, a more complex RothC model can be used. Under specific conditions, statistical models show similar effects to process models in simulating soil organic carbon at a small-scale regions, such as plots and fields.