Abstract:

Based on daily mean temperature observations from 18 meteorological stations in Renhuai, from 2014 to 2023 and Digital Elevation Model (DEM) data, 1km×1km daily mean temperature grid dataset was constructed by Light gradient boosting machine (LightGBM) to impute missing station data and Random forest regression kriging (RFK) spatial interpolation. The spatio−temporal evolution characteristics of mean temperature at the Twenty−four solar terms scale (solar−term mean temperature) was analyzed, while its relationship with the timing of sauce−aroma baijiu production processes, aiming to clarify the characteristics of solar−term mean temperature changes in Renhuai under the background of climate warming, thereby providing scientific reference for optimizing local sauce−aroma baijiu brewing technology and enhancing climate adaptation. The results showed that: (1) the LightGBM algorithm demonstrated high reliability in imputing missing station data and effectively captured the distribution characteristics and variation processes of the true values. The gridded temperature data interpolated using the RFK algorithm exhibited superior applicability and stability in the topographically complex terrain of Renhuai. (2) From 2014 to 2023, the annual and four seasonal−type solar terms mean temperatures in Renhuai exhibited an overall spatial pattern of "warmer river valleys and cooler mountainous areas". While the Chishui river valley had higher solar−term mean temperatures than the hill−valley transition zone with smaller interannual variability. Among the Twenty−four solar terms, the highest average temperature occurred during Major Heat (26.3℃), and the lowest during Minor Cold (5.3℃). Spring−type solar terms showed an overall warming trend. The earliest solar term reaching the phenological threshold of spring−type mean temperature advanced from the Spring Equinox to Hibernator Awakening, while the latest solar term departing from the summer−type mean temperature phenological threshold was delayed from Heat Withdraws to White Dew. This indicated that the phenology in Renhuai exhibited "a warming feature of earlier warming and delayed cooling". (3) At Maotai station, the days with daily mean temperatures between 20−30℃ accounted for 37.7% of the year, providing suitable thermal conditions for Daqu fermentation in sauce−aroma baijiu production. The timing of the seven-rounds was highly synchronized with the temperature changes across the Twenty−four solar terms: relatively lower daily mean temperatures in the early stages favored the accumulation of flavor compounds, while rising temperatures in the later stages promoted an increase in alcohol content in the liquor body.

Key words: Renhuai city, LightGBM algorithm, Self?organizing maps clustering (SOM), Twenty?four solar terms, Sauce?aroma baijiu