Objectives A prompt, efficient and accurate monitoring method was constructed for the dynamic monitoring of soil organic carbon (SOC) and soil total nitrogen (STN) in Area 1 of Plot 9 within the Millennium Forest, Xiong’an New Area, to clarify their spatiotemporal distribution characteristics, serve the sustainable forest soil management, and ensure the long-term stable development of the Millennium Forest.

Methods A total of 95 soil samples were collected in Area 1 of Plot 9 within the Millennium Forest, Xiong’an New Area, for determination of soil organic carbon and total N content.Six Landsat 8 OLI images covering the study area from 2018 to 2023 , all acquired in September or October, were collected to extract and calculate band spectral reflectance values and remote sensing indices, then the data were correlated with the measured SOC and STN. The environmental factors were screened using the Feature Screening method to minimum the complexity of the constructed model. Based on the Random Forest algorithm, Based on the Random Forest algorithm, prediction models for SOC and STN contents were separately constructed, including three types of models: band models, band+single remote sensing index models, and band+multiple remote sensing index models. For the SOC prediction model, band 4 and band 6 are used to establish band model. Red Vegetation Index (RI) was incorporated into the single remote sensing index model, while RI, Soil Salinity Index 2 (SI2), and Soil Salinity Index 3 (SI3) were employed for the multi-remote sensing index model. For the STN prediction model, band 4 and band 3 werer used to establish band models. SI3 participated in the establishment of the single remote sensing index model, while the multi-remote sensing index model combined RI, Soil Salinity Index 1 (SI1) and SI3.The model accuracy was comprehensively evaluated using multiple indicators including the coefficient of determination (R²), relative percent deviation (RPD), mean absolute error (MAE), and root mean square error (RMSE). The optimal prediction model was selected and applied to predict the spatiotemporal distribution of SOC and STN.

Results The SOC prediction model constructed using RF-band 4+band 6+RI+SI2+SI3 (R2=0.778, RPD=1.933, MAE=0.719, RMSE=0.829), and the STN prediction model using RF-band 4+band 3+SI3 (R2=0.679, RPD=1.742, MAE=0.047, RMSE=0.054) showed the highest accuracy and the best prediction effect. From 2018 to 2023, the SOC content in the study area showed an overall trend of fluctuating annual increases, with a slight decrease in 2022.The annual average values fluctuated between 7.274 and 8.334 g/kg. The SOC content in the entire forest area is relatively stable at the edge, while there are significant internal variations. The SOC content in the ecological basic forest and the western part of the near-natural forest have always been higher than that in other forest areas.The overall changing trend of STN content is basically consistent with that of SOC content, and its annual average changing range is 0.676-0.772 g/kg.

Conclusions On the basis of the band model for organic carbon and total N, the introduction of remote sensing index factors could improve the prediction accuracy. According to the prediction, the content of SOC and STN in the study area showed a trend of yearly fluctuating increases, with a slight decrease in 2022. Among the five types of landscape functional forests, the ecological basic forest and the western part of the near-natural forest have always been higher than those in other forest areas.