Yellow soils developed in limestone weathering materials are representative on Guizhou Plateau, Southwest China. As one of the most important cultivated soils in Guizhou Province, karst yellow soils are generally thin and can be significantly damaged by mild soil erosion. This work used the structural equation model (SEM) to analyze the influence of various factors on runoff generation and sediment loss based on a long time series (2015–2020) of natural rainfall-runoff data and soil erosion data collected from 18 runoff plots in the karst yellow soil region of Southwest China, which reflects the erosion dynamics under natural conditions. Slope runoff plots are the most popular and efficient way to collect data on runoff generation and soil erosion. The findings show that: (1) There were 139 rainfall-runoff occurrences between 2015 and 2020, with moderate rain (10–25 mm) and heavy rain (25–50 mm) events making up the majority. Due to its high frequency and prolonged duration, heavy rain had the greatest impact on the overall rainfall erosivity R value (R). (2) Mean surface runoff (SR) values ranged from 17.37 mm to 133.90 mm, while mean sediment loss rates (SLR) ranged from 1.36 t·ha⁻¹·a⁻¹ to 23.49 t·ha⁻¹·a⁻¹. SR and SLR can be successfully reduced by the forest floor. Broadleaf forest, coniferous forest, mixed forest, and orchard had mean SR values of 19.33%, 12.97%, 16.10%, and 33.38% of fallow land, respectively, and had mean SLR values of 6.90%, 5.79%, 6.34%, and 12.64% of fallow land, respectively. (3) SR generation at the plot scale was substantially linked with 30-min maximum rainfall intensity (I₃₀), while vegetation coverage (VC) and antecedent soil water content (ASW) showed negative direct inference on SR and runoff sediment concentration (RSC) according to SEM analysis. The VC and ASW showed the highest indirect impact on SLR. This study will serve as a scientific reference for the water and soil erosion management in karst yellow soil region and serve as a scientific guidance for regional land use in Southwest China.

Within the realm of intelligent transportation systems, the precise forecasting of vehicular speed across individual road segments constitutes a fundamental task. This metric serves as a pivotal indicator for evaluating the extent of network congestion and facilitating informed strategic planning. Contemporary methodologies predominantly employ recurrent neural networks (RNNs) to model temporal dependencies, while leveraging graph convolutional networks (GCNs) to capture spatial dependencies within the data. However, these methods fail to integrate temporal and spatial dependencies to establish global dependencies. This study introduces the spatio-temporal kernel graph convolutional network (STK-GCN), a novel framework designed for modeling and forecasting traffic data. Specifically, we devise a spatio-temporal kernel capable of generating both spatial and temporal matrices, which are subsequently utilized within a encoder–decoder architecture to concurrently capture spatio-temporal dependencies. Furthermore, we introduce a novel temporal graph convolution module to enhance temporal data. To demonstrate the efficacy of the proposed STK-GCN, comprehensive experiments were carried out on two real-world traffic datasets, namely METR-LA and PEMS-BAY. The results indicate that our model surpasses existing state-of-the-art methods.