This study aims to clarify the spatiotemporal evolution of drought in the Huang-Huai-Hai Plain of China. A systematic investigation was implemented on the long-term drought dynamics using multi-index comparison and three-dimensional event identification. The regional drought characteristics were explored to determine the most suitable drought indicator for the specific plains from 1960 to 2020. Daily meteorological observations were taken from 102 stations, together with the regional soil properties. Four drought indices were calculated monthly: Palmer Drought Severity Index (PDSI), Self-Calibrating Palmer Drought Severity Index (scPDSI), Standardized Precipitation Index (SPI), and Standardized Precipitation Evapotranspiration Index (SPEI). A three-dimensional dynamic drought identification was employed to merge spatially adjacent drought grids over successive months, thereby constructing continuous spatiotemporal drought volumes. Each identified event was used to extract the duration, cumulative severity, mean intensity, maximum affected area, and centroid migration distance. The performance of four indices was evaluated to compare the identified events with documented historical drought records and disaster statistics. The most suitable index was selected for spatiotemporal analysis. Among four indices, scPDSI shared the highest applicability for the Huang-Huai-Hai Plain, as it most consistently matched historical drought occurrences and regional hydroclimatic conditions. Furthermore, 89 independent drought events were identified over the 61-year period using scPDSI. The longest event persisted for 13 months, while the maximum cumulative severity reached 6.3×10⁴. Temporally, drought events were concentrated during 1960–2002, accompanied by relatively large maximum affected areas and high peak intensities. The frequency, peak intensity and duration of drought events declined markedly after 2003. These indicators weakened markedly after 2003, indicating the regional wetting. Severe drought episodes were concentrated during three major periods: 1964-1968, 1977-1983, and 1993-2002. The maximum affected area reached 6.9×10⁵ km²during extreme cases, accounting for 88.4% of the total study area. Spatially, there was outstanding heterogeneity. Specifically, the north-central region, particularly the Beijing-Tianjin-Hebei area, was identified as the epicenter of long-duration and high-severity drought events. In contrast, the southern portion of the plain was characterized by the shorter and lower-intensity incidents. Northward-moving droughts were typically large-scale events with the migration distances exceeding 960 km. Conversely, southwestward migration was dominated by localized droughts with displacement distances ranging from 0 to 720 km. Furthermore, the heavy drought events exhibited obvious spatial clustering characteristics, where their centroids gradually shifted from the central plain toward the northern regions over time. In contrast, ordinary drought events displayed multi-source dispersion, with a fan-shaped expansion from the southwest to the northeast. Overall, the three-dimensional dynamic identification can effectively capture the continuity and migration of drought events for the temporal trends and spatial heterogeneity in drought evolution, thus providing for a more integrated representation than conventional static approaches. The post-2003 wetting tendency and directional migration were identified as a significant shift in the regional drought behavior. The scPDSI can be adopted for long-term monitoring in the Huang-Huai-Hai Plain. These findings can provide technical support for regional drought early warning and risk prevention.

In order to improve the flood forecasting accuracy, the real-time online correction of the flood forecasting results by mining the measured water level and discharge data is used to make full use of the information contained in the measured sequences of water level and discharge. A Kalman filter correction technique is proposed based on multi-source forecast residuals. Corresponding rising difference model and autoregressive model were used to construct the multi-source error information source, and then Kalman filtering technology was used to fuse the multi-source error sequences for the real-time correction of flood forecast results. This paper selected the Qiantang River Basin of Zhejiang Province as the study area. The validation results show that the multi-source residual fusion correction technique based on Kalman filtering technology can significantly reduce the flow simulation error and the average relative error is reduced by more than 10%.