CIAHS-Data.xls

In September 2023, the Ministry of Agriculture and Rural Affairs of the People’s Republic of China (www.moa.gov.cn, accessed on 20 December 2024) announced the seventh batch of CIAHS. As of this release, a total of 188 CIAHS sites (2013–2023) have been identified. Based on their primary functions, we categorized the heritages systems into four types: Planting systems (76 sites, 40.43%), focused on crop cultivation, farmland landscapes, and mixed farming; forestry and fruit systems (92 sites, 48.94%), centered on forestry and fruit planting; animal husbandry systems (15 sites, 7.98%), emphasizing livestock breeding; fisheries systems (5 sites, 2.66%), dedicated to aquaculture. At the national scale, CIAHS sites were represented as point elements using geographic coordinates, and a spatial distribution map was generated (Fig. 1).Additionally, we sourced the following datasets: (1) The raster data for annual average temperature (1960–2023) and annual average precipitation (1960–2023) from the Resource and Environmental Science Data Platform (www.resdc.cn, accessed on 20 December 2024); (2) Gridded data on GDP and population density from the same source; (3) Elevation and slope data from the Geospatial Data Cloud (www.gscloud.cn, accessed on 20 December 2024); (4) Vector data (rivers, roads, and administrative boundaries) from the National Catalogue Service for Geographic Information (www.webmap.cn, accessed on 20 December 2024).The spatial resolution of all raster data, including temperature, precipitation, elevation, slope, economic density, and population density, was adjusted to 1 km. Subsequently, a circular buffer zone with a 20-km radius (a scale commonly used in cultural heritage landscape studies to represent the core area of influence36) was created centered around each heritage site, and the mean value of all rasters within the buffer zone was computed as the indicator value for the corresponding site. Government response reflects the level of governmental focus on agricultural heritage systems. To quantify this, we collected government work reports, policy documents, administrative bulletins, and crawled news from official government websites at all levels. Using text analysis techniques, we counted the frequency of mentions of each heritage site​​ and weighted according to ​​the presumed influence and resource allocation capacity of the administrative hierarchy​​ (provincial weight 0.5, city 0.3, county 0.2). The weighted results were then normalized to derive the government response index. Residents’ attention indicates the degree of local residents’ focus on agricultural heritage systems. This was calculated by ​​crawling the volume of topic discussions about the heritage site​​ on social media platforms (e.g., Douyin, Rednote, Weibo) combined with ​​the number of travel bookings for the heritage site​​ on online travel agencies (e.g., Ctrip, Meituan, Qunar). These aggregated results were normalized to obtain the residents’ attention index. To meet Geodetector’s requirements, all factors underwent discretization. For this purpose, we employed the Natural Breaks classification method to reclassify factor values. This method identifies inherent natural grouping points within the data through the Jenks optimization algorithm, maximizing between-class differences while minimizing within-class differences37. This principle is fully consistent with the theoretical foundation of Geodetector: spatial stratified heterogeneity (where within-stratum variance is smaller than between-stratum variance). Furthermore, empirical analysis of our study confirmed that the Natural Breaks method achieved higher accuracy compared to alternative like Equal Interval and Quantile classification. The factor reclassification results are presented in Table 2.​ In the table, the values 1–6 correspond to the recategorized values resulting from indicator reclassification.