The temporal trajectories and driving forces of wildfire regimes differing between wildland-cropland and wildland-urban interfaces on the Central Yunnan Plateau

The temporal trajectories and driving forces of wildfire regimes differing between wildland-cropland and wildland-urban interfaces on the Central Yunnan Plateau | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article The temporal trajectories and driving forces of wildfire regimes differing between wildland-cropland and wildland-urban interfaces on the Central Yunnan Plateau Yang Lin, Mingjian Xiahou, Jiesheng Rao, Caifang Luo, Tao Yang, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9559449/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background As human activities continue to extend into wildland areas, wildfire risk in wildland-urban interfaces (WUI) has drawn considerable attention. By contrast, the wildland-cropland interface (WCI) has been less examined, and few studies have compared the spatiotemporal patterns and driving forces of wildfire between these two types of interfaces. Based on Landsat data from 1990 to 2023, we compared fire activity and its drivers in WCI and WUI landscapes on the Central Yunnan Plateau (CYP), China, at multiple spatial scales. Results The results indicate a clear turning point around 2010. Before 2010, burned area increased in both interfaces. After 2010, both burned area in the WCI and its share of total burned area declined markedly, whereas the share of the WUI remained stable, after 2010. Concurrently, the fire regime shifted from winter-spring dominance toward greater summer extension, with spring burned area in WUI increasing after 2000. Meteorological factors dominated fire variability in both interfaces: temperature primarily controlled fire occurrence, whereas relative humidity governed burned-area percentage. Beyond this shared climatic background, fire occurrence in the WCI was more strongly associated with population density, whereas fire occurrence in the WUI was more constrained by topographic relief. Burned-area percentage in the WCI exhibited a climate-dominated pattern, whereas that in the WUI was jointly influenced by coniferous forest proportion and population density. Conclusions By comparing WCI and WUI fires over more than three decades, this study shows that the two interfaces have distinct fire regimes and require differentiated management. After 2010, the regional focus of wildfire activity shifted increasingly from WCI landscapes toward the WUI, where fuel continuity and population exposure jointly shaped burned-area percentage. Fire management should therefore place greater emphasis on agricultural fire regulation in the WCI and fuel management in the WUI. Wildland-urban interface Wildland-cropland interface Fire occurrence Spatiotemporal dynamics Driving mechanisms Management Insights Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction Wildfire is a major disturbance in terrestrial ecosystems and one of the most widespread natural hazards. It reshapes vegetation structure, alters biodiversity patterns (Bowman et al., 2009 ; He et al., 2019 ), and strongly affects energy budget and elements cycling on the ground from local to global scales (Archibald et al., 2018 ). Meanwhile, the growing threats that wildfire poses to environmental quality and human health have attracted increasing attention (Radeloff et al., 2023 ; Zhao et al., 2025b ). Under intensifying climate change and human disturbance, fire regime shifting in various ecosystems are becoming increasingly evident across regional to global scales (Shen et al., 2025 ). Over the past three decades, although some studies have reported a decline in global total burned area (Andela et al., 2017 ), the economic losses and ecological threats caused by increasingly frequent extreme wildfire events have continued to rise, affecting ever larger populations (Cunningham et al., 2024 ). This trend is driven in part by global warming and the intensification of regional extreme droughts, which have increased fluctuations in vegetation biomass and fuel flammability (Abatzoglou and Williams, 2016 ; Senande-Rivera et al., 2022 ), and also to a large extent by the spatial expansion of human activities in multiple forms, including housing, farming, public infrastructure, and tourism facilities (Radeloff et al., 2018 ). The spread of population and assets into vegetated areas has made fire impacts more likely to concentrate at the interface between natural vegetation and human land use, thereby exacerbating disaster consequences (Schug et al., 2023 ; Zamanialaei et al., 2025 ). As a result, the wildland-urban interface (WUI) has increasingly become a focal area in wildfire research and management (Haight et al., 2004 ; Chen et al., 2024 ). For example, by combining a global WUI time-series map for 1985–2020 with burned area and population distribution data for 2001–2020, Chen et al. ( 2024 ) found that about 12.54% of global WUI areas, representing approximately 10.11 million people, were located within a 4800 m wildfire threat buffer. Carlson et al. ( 2025 ) further showed that the proportion of wildfire-exposed buildings that were destroyed increased from about 10% in 2002–2012 to around 32% in 2013–2022, indicating that wildfire events are becoming more destructive in addition to exposing more infrastructure. Recent national and regional assessments within China have advanced our understanding of wildfire risks in the traditional WUI (Gong et al., 2024 ; Li et al., 2025 ). However, the WUI framework alone may not adequately capture wildfire risk in many landscapes with long agricultural histories, including China, India, Southeast Asian countries, as well as the Brazilian Cerrado, where urban peripheries are often dominated by managed ecosystems such as croplands and plantations, instead of clustering houses (Vadrevu et al., 2019 ; Xu and You, 2023 ). Here, the transition between human activities and wildlands occurs mainly along the margins of these managed landscapes. We refer to this configuration as the wildland-cropland interface (WCI). The interweaving of croplands, orchards, tea and rubber plantations, nurseries, secondary forests, shrublands, and grasslands creates a mosaic landscape that is also prone to frequent wildfires (Segura-Garcia et al., 2025 ). More broadly, agricultural expansion and land-use intensification have reshaped fire regimes in many managed landscapes over recent decades. Land clearing, plantation development, and agricultural burning have increased human influences on fire activity and altered patterns of fire occurrence and spread (Archibald et al., 2012 ; Ribeiro et al., 2024 ). In these mosaics, human-related ignition sources and fuel management may produce fire probabilities and burned-area responses that differ from those observed in WUI landscapes (Ying et al., 2021 ; Gelabert et al., 2025 ). However, targeted research on wildfire dynamics and driving mechanisms in WCI remains scarce, and the similarities and differences in how human activities shape fire processes across WUI and WCI landscapes remain poorly understood. Comparing wildfires in WCI and WUI is meaningful not only to explore their formats of spatial distribution and temporal dynamics of anthropogenic burning, but also to disentangle the driving forces in natural and social-cultural aspects of these agriculture-intensive regions. Against this background, we use the fire-prone Central Yunnan Plateau as a case study to compare wildfire dynamics in WUI and WCI landscapes, relate them to regional fire trends, and identify differences in the mechanisms governing fire occurrence and burned-area percentage. The mountainous region of Southwest China is one of the major wildfire hotspots in China and forms part of the broader high-fire belt of the Indochina Peninsula, with Yunnan Province serving as a core region of high wildfire activity in Southwest China (Ying et al., 2021 ; Zhang et al., 2025 ), and the Central Yunnan Plateau being the most fire-prone area within the province (Su et al., 2015 ). This region has the highest population density and agricultural activity in Yunnan and has also experienced rapid urbanization in recent decades (Wang et al., 2021 ; Li et al., 2024a ). Urban expansion, denser road networks, and changing agroforestry management practices have increased the contact boundaries between natural vegetation and both developed land and cropland. As a result, WUI and WCI landscapes coexist in the region and show increasingly complex spatial configurations (Li et al., 2024b ). Although previous studies have examined forest fires in Yunnan, most have focused on regional-scale assessments or specific fire events (Han et al., 2015 ; Wu et al., 2021 ). Research on wildfire dynamics within landscape interfaces is still limited. In particular, the fire activity patterns in WUI and WCI landscapes and the different effects of anthropogenic and natural drivers remain poorly understood (Wang et al., 2023 ; Deng et al., 2025 ). A comparative analysis of wildfire evolution in WUI and WCI interfaces on the Central Yunnan Plateau can therefore help clarify how human and environmental factors contribute to fire risk in these heterogeneous landscapes. Such knowledge is important for developing differentiated vegetation management strategies and targeted wildfire mitigation policies. Accordingly, this study examines long-term wildfire dynamics within WUI and WCI landscape interfaces on the Central Yunnan Plateau. We address three questions: 1) How did burned area vary interannually and spatially across the Central Yunnan Plateau, the WCI, and the WUI during 1990–2023? 2) How do wildfire seasonality and distributions along topographic and vegetation gradients differ between WCI and WUI landscapes? 3) What similarities and differences characterize the dominant drivers of fire occurrence and burned-area percentage in these two interface types? 2. Materials and methods 2.1 Study area The Central Yunnan Plateau (CYP) is located in central Yunnan Province, China. It includes Kunming, Dali, Yuxi, Qujing, the Chuxiong Yi Autonomous Prefecture, and the northern part of the Honghe Hani and Yi Autonomous Prefecture, including Mengzi, Gejiu, Jianshui, Kaiyuan, Mile, Luxi, and Shiping. The region covers about 111400 km 2 (Fig. 1 ), accounting for roughly 30% of the total area of Yunnan Province. As a core part of the Yunnan-Guizhou Plateau, the CYP has an average elevation of about 2000m and a typical plateau mountain-basin mosaic landscape. The climate belongs to the subtropical plateau monsoon type, with a clear dry-wet seasonal pattern. More than 80% of annual precipitation falls in the wet season from June to October. The dry season, from November to the following May, is affected by the southern branch of the subtropical westerly jet. Warm and dry conditions, together with high wind speeds during this period, create a critical fire-weather window. Influenced by intensive anthropogenic activities, the region contains diverse vegetation types. The primary fuel types consist of secondary coniferous forests dominated by Pinus yunnanensis , semi-moist evergreen broad-leaved forests, and xerophytic shrub-grasslands, all of which exhibit significant spatial heterogeneity in their horizontal and vertical structures (Zhang et al., 2025 ). This heterogeneous fuel matrix, coupled with seasonal drought, fosters frequent wildfire activity, while the complex topographic gradients provide a highly variable environment for fire propagation (Ying et al., 2021 ; Qin et al., 2024 ). 2.2 Data sources and processing WUI and WCI data : We used 30m land-cover data for the CYP from 1990 to 2023 from the China Land Cover Dataset (CLCD) (Yang and Huang, 2021 ). Based on the Google Earth Engine (GEE) platform, we mapped the spatial and temporal distributions of the Wildland-Urban Interface (WUI) and the Wildland-Cropland Interface (WCI) using buffer analysis. The WUI is usually defined as the area where human infrastructure and built-up land meet or intermingle with natural vegetation (Chen et al., 2024 ). For the delineation of the WUI, we established a 2400 m geographic buffer around urban built-up land and extracted its intersection with natural vegetation (including forests, grasslands, and shrublands) (Carlson et al., 2022 ). Correspondingly, the WCI is defined as the transition zone where farmlands and plantations directly interface with natural vegetation. Following previous work on agricultural fire dynamics and forest edge effects (Cochrane, 2001 ; Broadbent et al., 2008 ), and with support from spatial sensitivity tests, we used a 500 m buffer around croplands to delineate the WCI. Sensitivity tests using 250 m and 1000 m thresholds showed highly consistent temporal patterns of fire dynamics, supporting the 500 m baseline (Fig. S1 ). We then applied connectivity analysis to remove small fragmented patches and improve the spatial continuity of the interface zones (Guo et al., 2024 ). Wildfire activity data : We used burn area and burned-area percentage to estimate wildfire activity in the CYP. Burned area was obtained from a monthly Landsat 30 m burned-area product for 1990–2023 developed by Xiahou & Shen(2025), which was clipped to the spatial extent of the CYP. This product was generated by training a Random Forest model with filtered active fire detections from MODIS and VIIRS and pre- and post-fire spectral changes from Landsat imagery. Burned pixels were identified using a probability threshold of 0.90. Water bodies, built-up land, and other non-vegetated surfaces were masked to improve mapping accuracy. To calculate burned-area percentage, we divided the CYP into 1 × 1 km grid cells and calculated the proportion of burned area within each grid cell. Climate data : Monthly gridded datasets of precipitation, mean temperature, maximum temperature, and relative humidity at a 30m resolution for 1990–2023 were generated using the Thin Plate Spline interpolation method (Hutchinson, 1995 ). The interpolation used observations from 124 meteorological stations in and around the CYP, with digital elevation model (DEM) data included as the main topographic covariate. Elevation data came from the SRTM dataset ( http://srtm.csi.cgiar.org ), with a spatial resolution of 3 arc-seconds to facilitate interpolation correction (Xiahou et al., 2024 ). We aggregated all meteorological variables into annual dry-season indicators from November to April of the following year to represent long-term variation in fire-season climate. Human activity data : The human activity variables used in this study included population density (POP), gross domestic product (GDP), and the Human Footprint Index (HFI). GDP and POP data were obtained from the Resource and Environmental Science and Data Center of the Chinese Academy of Sciences ( https://www.resdc.cn/ ) as annual 1 km gridded datasets. HFI data were taken from Mu et al. ( 2022 ) as annual grids with a 1 km spatial resolution. We also calculated building density from the CLCD dataset as the proportion of built-up land within each 1 km grid cell. Landcover map We developed a decadal vegetation classification dataset for the CYP at 30m resolution for 1990–2023. Following the Ecosystem List of Yunnan Province (2018 Edition), we selected 15066 high-quality samples from more than 25000 land-cover and land-use plots surveyed between 2018 and 2020. The classification used Landsat spectral indices from visible, near-infrared, and shortwave infrared bands, including the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Normalized Difference Built-up Index (NDBI), and Land Surface Water Index (LSWI). Topographic factors from ASTER GDEM and climate data were also included. We used a 7:3 ratio for training and validation and applied a Random Forest algorithm together with a sample transfer learning strategy. The transfer procedure used spectral similarity analysis to propagate high-confidence samples from the baseline year to historical images. Stable samples were identified by comparing reference and target images with Euclidean distance and spectral angle distance. This procedure helped optimize the models and maintain consistency across the multi-temporal classification results. These steps produced the preliminary long-term classification map (Pan and Yang, 2010 ; Tuia et al., 2016 ; Hamrouni et al., 2021 ). We then refined the maps using forest inventory data from the Yunnan Academy of Forest Inventory and Planning, including information on plantations, the distribution of Pinus yunnanensis and Pinus armandii , and semi-moist evergreen broad-leaved forests, together with potential vegetation maps and expert knowledge. The final vegetation maps had an average overall accuracy of 0.72, and the accuracy for vegetation subtypes was 0.69 (Xiahou et al., 2025 ). To ensure spatial consistency across multiple sources, all spatial data were reprojected to the Southern Albers Equal Area Conic projection and resampled to a uniform 30m resolution. 2.3 Statistical analysis We compared fire-pattern differences between WUI and WCI at multiple scales. The analysis focused on three aspects: spatiotemporal evolution, seasonal characteristics, and driving factors at the prefecture, county, and town levels. This framework was used to examine wildfire characteristics and the mechanisms behind fire activity in the two interface types (Fig. 2 ). Temporal trend analysis To quantify the interannual variability of wildfire activity from 1990 to 2023, we calculated annual burned area and burned-area percentage for three spatial domains: the CYP, the WCI, and the WUI. We used Ordinary Least Squares (OLS) linear regression to evaluate long-term trends in these wildfire indicators. The slope, coefficient of determination (R 2 ), and p -value were used to describe the magnitude and reliability of each trend. Statistical significance was defined at p < 0.05. Hotspot (Getis-Ord Gi*) analysis We used the Getis-Ord Gi* statistic to identify significant spatial clusters of fire frequency from 1990 to 2023. This local spatial autocorrelation method detects clusters of high or low values by calculating a standardized z-score for each spatial unit and its neighborhood relative to the global mean (Getis and Ord, 1992 ; Ord and Getis, 1995 ). We compared Gi* statistic distributions across several grid scales in ArcGIS Pro 3.0.1 to test the stability and significance of the hotspot patterns. Based on this spatial sensitivity analysis, we selected an 8 km grid as the optimal scale for representing regional wildfire clustering in the CYP. Spatial units were classified as hotspots, coldspots, or non-significant areas according to their z -scores and p -values, and were mapped at 90%, 95%, and 99% confidence levels. Variable extraction To identify differences in the drivers of fire occurrence and burned-area percentage, we divided the study area into 1 × 1 km grid cells and constructed two annual fire-response variables. Fire occurrence was defined as whether a grid cell burned in a given year (0/1). Burned-area percentage was defined as the proportion of each grid cell that burned in a given year. Predictor selection followed a climate-topography-fuel-human activity framework, with candidate variables chosen from established wildfire studies (Zhang et al., 2025 ; Zhao et al., 2025a ). Because wildfire activity in the CYP is strongly concentrated in the dry season, we adapted the predictor set to regional fire-season conditions. The selected predictors included four groups. Meteorological variables included dry-season precipitation, dry-season mean temperature, dry-season maximum temperature, and dry-season relative humidity, which represented hydrothermal constraints on fire. Topographic variables included elevation relief and mean slope. Fuel variables included the proportions of coniferous forest and shrubland and pre-fire NDVI, which described vegetation structure. Human activity variables included population density, GDP, road density, the human footprint index, and building density, which represented anthropogenic ignition pressure and land-surface modification. Considering the complex nonlinear relationships and potential interactions among wildfire drivers, we used Random Forest models (Breiman, 2001 ). With the "ranger" package in R (Wright and Ziegler, 2017 ), we built classification models for fire occurrence and evaluated them with AUC, and regression models for burned-area percentage and evaluated them with R 2 . Variable importance and partial dependence plots (PDPs) were used to characterize the nonlinear responses of the main predictors. All analyses were conducted in ArcGIS Pro 3.0.1 and R 4.5.2 (R Core Team, 2025 ). 3. Results 3.1 Spatiotemporal patterns of burned area Overall, burned area in the CYP, WCI, and WUI followed a pattern of increase followed by decline (Fig. 3 ). Regional fire activity peaked in 2010, when burned area in the CYP reached 715 km 2 , or 0.64% of the regional area. After 2010, the annual mean burned area in the WCI was 136.26 km 2 , lower than the 1990s mean of 149.64 km 2 (Fig. 3 b). In contrast, the WUI burned area in the 2010s was 64.55 km 2 , higher than the 1990s mean of 32.88 km 2 (Fig. 3 c). The WCI accounted for most burned area in the CYP over the study period. When 2010 was used as a temporal boundary, the contribution of the two interfaces changed. Before 2010, the burned-area proportions of both WCI and WUI increased, especially in the WUI. After 2010, both proportions declined, but the decline was much stronger in the WCI, while the WUI remained relatively stable (Fig. 3 d). At different administrative levels, wildfire activity showed similar interannual fluctuations, with an increase before 2010 and a decline afterward (Fig. S3). WCI and WUI shared this broad temporal pattern, but differed in burned-area magnitude and proportional contribution (Fig. S4a-e). Total burned area increased from the prefecture level to the town level, and towns contributed the largest share of total burned area (Fig. S4a-c). Burned-area proportions also changed around 2010, but their administrative distribution differed between the two interfaces (Fig. S4d-e). The WUI proportion was higher than the WCI proportion at the prefecture level, whereas the WCI proportion was higher at the county and town levels. County-level units had the highest burned-area proportions in both interface types, indicating stronger wildfire risk in county-level interface landscapes. Fire activity in the CYP, WCI, and WUI showed a strong seasonal pattern (Fig. 4 a). The multi-year monthly distribution was unimodal, with burned area concentrated in the dry season from November to the following April and peaking in March. After May, fire activity dropped sharply and stayed low through summer and autumn. Seasonal trends differed among regions (Fig. 4 b). Summer fires increased across the study area, although their absolute contribution remained small. In the WUI, the seasonal center shifted toward spring. After 2000, the proportion of spring fires in the WUI increased with fluctuations, while the winter contribution decreased. In the WCI, winter and spring remained dominant, and the seasonal structure changed little over the past three decades. Hotspot analysis showed clear spatial differences between the two interfaces (Fig. 5 ). In the WCI, high-frequency fire zones were concentrated mainly in Honghe Prefecture in the southern part of the study area, forming large and continuous clusters. Central and northern areas were mostly non-significant or contained isolated hotspots (Fig. 5 b). In the WUI, hotspots were more dispersed and polycentric, with significant clusters around highly urbanized areas such as Kunming, Yuxi, Qujing, and Honghe (Fig. 5 a). Hotspots in Dali and Chuxiong were more fragmented, but several independent high-frequency fire zones were still present. Overall, WUI fire hotspots were more widely distributed across the CYP. Fire activity also differed among vegetation types and environmental gradients (Fig. 6 , Fig. S5). For each vegetation category, burned area changed in stages around 2010 (Fig. 6 a-d). Before 2010, burned area generally increased across all vegetation types, with especially large increases in shrublands in both the WCI and WUI. After 2010, burned area declined across all vegetation types. The decline in savannas was faster than their earlier increase. Warm-temperate coniferous forests and shrublands still had high peak values in 2010 and maintained large contribution proportions, indicating that they were the main fire-carrying vegetation types in the interface landscapes. topographic gradients also affected fire distribution (Fig. S5). Along the elevation gradient, burned area was concentrated mainly at mid-elevations, with the 1600-2000m range contributing the largest proportion. WCI fires had a higher proportion at low elevations below 1200m, whereas WUI fires were more concentrated at mid-to-high elevations of 1600-2400m. Along the slope gradient, the burned-area proportion increased with slope and reached its maximum on steep slopes above 25 degrees. 3.2 Contrasting drivers of fire occurrence and burned-area percentage Random Forest importance rankings and PDPs showed that fire occurrence in the CYP and both interface types was mainly controlled by climate. Climatic variables accounted for about 44% of total importance across the three spatial domains, and dry-season mean temperature was consistently the most important predictor (Fig. 7 , Fig. S6a-b). PDPs showed that fire occurrence increased strongly when temperature exceeded about 12℃. Secondary controls differed between the two interfaces. Population density ranked higher in the CYP and WCI, with importance values of 9.1% and 8.9%, respectively. Its PDPs showed higher fire probabilities at very low and very high population densities, and lower probabilities at intermediate densities. In the WUI, elevation relief was more important, and fire occurrence increased with terrain complexity. Thus, fire occurrence in the CYP and WCI showed a stronger human influence, whereas WUI fire occurrence was more closely related to topographic conditions. Burned-area percentage was strongly constrained by atmospheric moisture in the CYP, WCI, and WUI (Fig. 8 , Fig. S6c-d). Relative humidity was the most important predictor in all models, with contribution rates of 18.6%, 24.5%, and 14.4% in the CYP, WCI, and WUI, respectively. PDPs showed high burned-area percentage when dry-season relative humidity was below 60%, followed by a rapid decline above this threshold. Although this moisture constraint was shared by all three spatial domains, the secondary controls differed. In the WCI, climatic factors dominated the model, with a cumulative contribution of 57.4%, and temperature ranked just behind humidity. In the WUI, burned-area percentage was more strongly related to fuel and human activity, especially coniferous forest proportion (9.9%) and population density (9.7%). PDPs showed that WUI burned-area percentage increased with coniferous forest cover and rose more sharply in densely populated areas, particularly above about 400 persons/km 2 . These patterns indicate that large fires in the WCI are mainly climate-controlled, whereas large fires in the WUI are linked to moisture, fuel structure, and human disturbance. 4. Discussion 4.1 Spatiotemporal shifts of interface wildfires Our results identify 2010 as a major turning point in wildfire dynamics on the CYP. Before 2010, burned area generally increased despite strong interannual fluctuations, whereas the 2010 peak likely reflected the historic extreme drought that affected Yunnan Province during that year (Li et al., 2019 ). Under such extreme drought conditions, the sharp decline in fuel moisture substantially increased vegetation flammability and facilitated fire spread at the regional scale, highlighting the high sensitivity of wildfire activity to climatic anomalies (Pausas and Ribeiro, 2013 ; Young et al., 2017 ). The overall decline in burned area after 2010, by contrast, points to the effectiveness of policy intervention, particularly the marked improvement in administrative fire prevention following the promulgation of the Forest Fire Prevention Regulations in 2009 and reforms to the forest public security system. Climate data show that precipitation remained low and temperature continued to rise after 2010 in the CYP (Fig. S7), yet total burned area across the region still decreased, further underscoring the effectiveness of policy regulation (Guo et al., 2015 ; Wang et al., 2022 ; Lian et al., 2024 ). This trajectory, in which extremes are triggered by climatic anomalies and subsequent declines are shaped by policy intervention, reflects the complex interplay between natural controls and anthropogenic management in wildfire dynamics. Within the two landscape interfaces, WCI consistently accounted for the larger share of burned area, indicating that ignitions and fuel management associated with agricultural activities remained the main sources of fire in the agroforestry mosaic of the CYP (Ying et al., 2021 ). This interpretation is consistent with studies from other agricultural or managed landscapes, where cropland and other managed fires contribute substantially to total fire activity, and where fires are closely linked to agricultural expansion and field clearing along forest edges (Lin et al., 2014 ; Xu et al., 2021 ). The marked reduction in WCI burned area after 2010 suggests that policy measures directly constrained agricultural fire sources and partly offset climatic risk. In contrast, burned area in the WUI remained relatively stable rather than declining substantially, underscoring the more complex effects of land-use change. Although fire regulation was strengthened, continued urban expansion led to a steady increase in WUI area and a persistent extension of the contact zone between human activities and forest fuels (Fig. S2). As a result, wildfire risk is increasingly shifting toward areas with higher population and asset density, thereby increasing potential societal exposure to wildfire hazards (Radeloff et al., 2018 ; Liu et al., 2025 ). The multi-scale analysis further showed that town units were the primary zones of fire occurrence within the interface landscapes. At this scale, fragmented built-up land increased the contact boundaries between anthropogenic activities and combustible fuels, creating more opportunities for both ignition and fire spread (Liu et al., 2025 ). In addition, the proportion of WUI fires was higher around prefecture-level capitals, whereas WCI fires accounted for a larger share at the county and town levels. This pattern reflects the influence of the urbanization gradient on fire regimes, the higher proportion of WUI fires around prefecture-level cities was associated with urban encroachment on and disturbance of natural vegetation, whereas the higher proportion of WCI fires at county and town levels was mainly linked to traditional agricultural burning. Such cross-scale heterogeneity in fire distribution suggests that spatial fire management should adopt a hierarchical and coordinated strategy. Spatial hotspot analysis revealed further contrasts between the two interface types. WCI fires showed a unipolar clustering pattern centered on Honghe Prefecture, whereas WUI fires exhibited a multi-nodal distribution around cities such as Kunming and Yuxi. This contrast further highlights the role of rapid urbanization in reshaping the spatial pattern of wildfire risk. 4.2 Seasonal evolution and landscape heterogeneity of wildfires in WCI and WUI Seasonal wildfire patterns on the CYP are jointly shaped by monsoon-controlled hydroclimatic seasonality and human activity. During winter and spring, precipitation deficits and low relative humidity rapidly dry fine fuels, creating favorable conditions for fire occurrence (Zhu et al., 2022 ). These conditions coincide with peaks in spring ritual burning and agricultural fire use, which further elevate ignition probability (Liu et al., 2018 ; Qi et al., 2022 ). Collectively, human activities and meteorological conditions determine the seasonal wildfire patterns of the CYP. These drivers are fundamentally consistent with the seasonal causes of fire in other regions. In Mediterranean climate zones, fires are also highly seasonal and often concentrated during summer droughts with high temperatures or at the end of the dry season (Moreira et al., 2020 ; Moreno et al., 2023 ). Notably, our results reveal a significant long-term increase in summer fire activity. This shift may reflect the growing influence of compound heat-drought extremes, which weaken the fire-suppressing effect of the rainy season and extend the seasonal window of flammable conditions (Shi et al., 2024 ). At the same time, increasing spring fire activity in the WUI suggests that urban expansion is moving residential and production activities closer to forest edges, thereby increasing the likelihood of anthropogenic ignitions under high fire-weather conditions (Guo et al., 2024 ). Conversely, the seasonal structure of the WCI has remained relatively stable over the long term, which may be closely related to the periodic inertia of agricultural fire use habits. Fires were concentrated mainly in mid-elevation and steep-slope environments, especially in the WUI, where continuous fuels and high human accessibility frequently coincide (Ye et al., 2017 ). Additionally, the policy of urban expansion toward mountain areas in Yunnan Province has increased the probability of fire at mid to high elevations (Yin and Ding, 2015 ). Steep slopes constitute high incidence environments by accelerating fire spread and increasing the difficulty of suppression. In terms of vegetation types, Warm-temperate coniferous forest, savanna and shrublands constitute the primary carriers of combustion (Zhang et al., 2022 ; Chen et al., 2023 ). Pinus yunnanensis is rich in oils and its litter dehydrates easily, resulting in extremely high flammability. Between 1990 and 2020, the area of warm temperate coniferous forests contracted after 2000, while the area of shrublands exhibited an increasing trend (Fig. S8). This reflects that current forest fire prevention policies have effectively suppressed coniferous forest fires but may have neglected the expanding shrubland areas to some extent. The rising proportion of shrubland fires not only reveals that edge habitats in interfaces are more likely to become ignition points under human interference but also indicates that the lack of targeted management for shrubland fuels has amplified the impact of fire disturbances through edge effects. 4.3 Divergence and evolution of driving mechanisms Meteorological factors were the dominant controls on both fire occurrence and burned-area percentage. Temperature and relative humidity ranked first in the occurrence and burned-area percentage models, respectively, and both showed clear nonlinear threshold responses (Badia et al., 2011 ). Fire occurrence increased sharply when mean dry-season temperature exceeded 12℃, whereas burned-area percentage expanded when dry-season relative humidity dropped below 60%. These thresholds are consistent with the basic ecological controls of fire, high temperatures accelerate the drying of fine fuels and litter, thereby facilitating ignition (Matthews, 2014 ), whereas very low atmospheric moisture allows fires to spread over larger areas (Jolly et al., 2015 ). This result is broadly consistent with recent studies on WUI fires in China, which likewise identified meteorological conditions as the dominant controls on WUI fire occurrence and hazard (Gong, 2025 ). However, under a unified climatic background, the WCI and WUI demonstrate differentiated driving mechanisms. Fire occurrence in the WCI is more dependent on population density, reflecting its anthropogenic driving characteristics dominated by high-intensity agricultural burning such as crop residue burning (Lv et al., 2024 ). In contrast, fire occurrence in the WUI shows stronger topographic dependency. The limiting effects of elevation relief and slope indicate that fire occurrence in this region is more constrained by complex terrain. In the WUI, steep terrain facilitates the upslope spread of fire, and the continuous vegetation coverage in such habitats allows ignition sources to develop into fire events of a certain scale more easily under equivalent conditions (Rothermel, 1972 ). The burned-area percentage models further clarified the roles of individual controls. Dry-season relative humidity was the most important limiting factor in both interfaces, indicating a universal atmospheric-moisture constraint on burned-area percentage. Yet the mechanisms diverged beyond this commonality. In the WCI, climatic factors remained overwhelmingly dominant. This likely reflects the fact that WCI fuels consist mainly of crop residues and herbaceous vegetation, which have large surface areas, are highly sensitive to drying, and lose moisture rapidly. Moreover, the open farmland environment lacks forest-canopy shelter, leaving surface fuels directly exposed to the atmosphere and allowing them to become flammable quickly as conditions dry (Nelson, 2001 ). By comparison, the WUI region is significantly enhanced by the dual effects of vegetation properties and human activities. When the proportion of coniferous forest exceeds 20%, the burned-area percentage in the WUI rises sharply, indicating that resinous and spatially continuous coniferous forests are the foundation supporting large-scale forest fire spread in WUI regions (Su et al., 2015 ; Chen et al., 2023 ). At low population densities, burned-area percentage varied only modestly, but once population density exceeded 300 persons/km 2 , burned-area percentage rose rapidly. At low densities, roads and buildings fragment continuous vegetation and provide some physical barriers to fire spread. Once density reaches a higher threshold, however, more frequent human disturbance and ignition sources overwhelm these early barriers, making it easier for locally continuous and intense burning environments to develop (Syphard et al., 2007 ). This also highlights the risk of large-scale fires in high population density WUI areas, which should be regarded as core zones for future forest fire prevention. 4.4 Implications for adaptive fire management The contrasting drivers identified for the WCI and WUI imply that wildfire governance should shift from uniform prevention strategies toward differentiated, interface-specific management. In the WCI, prevention should focus on regulating agricultural fire use along forest-cropland boundaries, especially during busy farming periods when anthropogenic ignitions are most likely. In the WUI, priority should be given to fuel management, including prescribed burning, fuel reduction, and maintenance of defensible buffer zones, to reduce the probability of fire spread into populated areas. Across administrative levels, prefecture-scale planning should integrate WUI fire risk into land-use planning and disaster prevention systems, whereas town level management should emphasize ignition-source control and rapid early response during peak burning seasons. 4.5 Limitations and Prospect This study clarifies the spatiotemporal contrasts and driving mechanisms of wildfires in WCI and WUI landscapes, but several limitations should be acknowledged. First, because of data constraints, we did not incorporate wind speed or real time meteorological conditions, both of which are critical for explaining extreme fire behavior (Werth et al., 2011 ). Second, our monthly-scale analysis cannot fully capture the fine-scale triggering processes of individual fire events. Future work should integrate higher-resolution satellite observations with fire behavior models to resolve these mechanisms more explicitly and improve prediction and management in heterogeneous interface landscapes. 5. Conclusion This study compared wildfire dynamics between the WCI and WUI on the Central Yunnan Plateau and examined their spatiotemporal patterns and driving mechanisms under a common framework of climate, topography, fuel, and human activity. The results show that 2010 was a major turning point. Before 2010, burned area generally increased in both interfaces. After 2010, burned area declined markedly in the WCI, while wildfire risk shifted increasingly toward the WUI. County- and town-level areas were the main locations of fire occurrence. Spatially, WCI fires were concentrated in Honghe Prefecture, whereas WUI fires clustered around expanding urban fringes. Winter and spring remained the main fire seasons, but summer fire activity increased over time, suggesting that wildfire risk is extending into non-traditional seasons. Coniferous forests and shrublands were the main fire-carrying vegetation types. In terms of mechanisms, temperature was the main control on fire occurrence, while dry-season relative humidity was the dominant control on burned-area percentage. Beyond this shared climatic control, WCI fire occurrence was more closely associated with population density, whereas WUI fire occurrence was more constrained by topographic relief. Burned-area percentage in the WCI remained mainly climate-driven, while burned-area percentage in the WUI was increasingly shaped by coniferous forest cover and population density. These findings indicate that WCI and WUI fires represent different interface fire regimes rather than variants of a single process. Future fire prevention should therefore adopt differentiated management for WCI and WUI landscapes, with stronger control of agricultural fire use in the WCI and greater attention to fuel management and population exposure in high-density WUI areas along expanding urban fringes. Declarations Competing interests The authors declare no competing interests. Author Contribution Yang Lin: Writing-review & editing, Writing-original draft, Visualization, Methodology, Data curation. Mingjian Xiahou: Writing-review & editing, Data curation. Jiesheng Rao: Writing-review & editing. Caifang Luo: Writing-review & editing, Methodology. Tao Yang: Writing-review & editing. Zehao Shen: Writing-review and editing, Methodology, Funding acquisition, Conceptualization Acknowledgments This study is supported by the Yunnan Fundamental Research Projects (202302A0370016) Data Availability Data will be made available on reasonable request. References Abatzoglou JT, Williams AP (2016) Impact of anthropogenic climate change on wildfire across western US forests. Proceedings of the National Academy of Sciences 113(42), 11770–11775 Andela N, Morton DC, Giglio L, Chen Y, van der Werf GR, Kasibhatla PS, DeFries RS, Collatz GJ, Hantson S, Kloster S, Bachelet D, Forrest M, Lasslop G, Li F, Mangeon S, Melton JR, Yue C, Randerson JT (2017) A human-driven decline in global burned area. Science 356(6345):1356–1362 Archibald S, Lehmann CER, Belcher CM, Bond WJ, Bradstock RA, Daniau AL, Dexter KG, Forrestel EJ, Greve M, He T, Higgins SI, Hoffmann WA, Lamont BB, McGlinn DJ, Moncrieff GR, Osborne CP, Pausas JG, Price O, Ripley BS, Rogers BM, Schwilk DW, Simon MF, Turetsky MR, Van der Werf GR, Zanne AE (2018) Biological and geophysical feedbacks with fire in the Earth system. Environ Res Lett 13(3):033003 Archibald S, Staver AC, Levin SA (2012) Evolution of human-driven fire regimes in Africa. Proceedings of the National Academy of Sciences 109(3), 847–852 Badia A, Serra P, Modugno S (2011) Identifying dynamics of fire ignition probabilities in two representative Mediterranean wildland-urban interface areas. Appl Geogr 31(3):930–940 Bowman DM, Balch JK, Artaxo P, Bond WJ, Carlson JM, Cochrane MA, D'Antonio CM, Defries RS, Doyle JC, Harrison SP, Johnston FH, Keeley JE, Krawchuk MA, Kull CA, Marston JB, Moritz MA, Prentice IC, Roos CI, Scott AC, Swetnam TW, van der Werf GR, Pyne SJ (2009) Fire in the Earth system. Science 324(5926):481–484 Breiman L (2001) Random forests. Mach Learn 45(1):5–32 Broadbent EN, Asner GP, Keller M, Knapp DE, Oliveira PJC, Silva JN (2008) Forest fragmentation and edge effects from deforestation and selective logging in the Brazilian Amazon. Biol Conserv 141(7):1745–1757 Carlson AR, Hawbaker TJ, Mockrin MH, Radeloff VC, Bair LS, Caggiano MD, Meldrum JR, Alexandre PM, Kramer HA, Steblein PF (2025) Rising rates of wildfire building destruction in the conterminous United States. Proceedings of the National Academy of Sciences 122(51), e2505886122 Carlson AR, Helmers DP, Hawbaker TJ, Mockrin MH, Radeloff VC (2022) The wildland–urban interface in the United States based on 125 million building locations. Ecol Appl 32(5), e2597 Chen B, Wu S, Jin Y, Song Y, Wu C, Venevsky S, Xu B, Webster C, Gong P (2024) Wildfire risk for global wildland–urban interface areas. Nat Sustain 7(4):474–484 Chen F, Si L, Zhao F, Wang M (2023) Volatile oil in pinus yunnanensis potentially contributes to extreme fire behavior. Fire 6(3):113 Cochrane MA (2001) Synergistic interactions between habitat fragmentation and fire in evergreen tropical forests. Conserv Biol 15(6):1515–1521 Cunningham CX, Williamson GJ, Bowman DMJS (2024) Increasing frequency and intensity of the most extreme wildfires on Earth. Nat Ecol Evol 8(8):1420–1425 Deng H, Li D, Cai S, Zhao F, Zhao J (2025) Spatio-temporal dynamics of forest fire occurrence in Yunnan, China from 2001 to 2021 based on MODIS. npj Nat Hazards 2(1):52 Gelabert PJ, Jiménez-Ruano A, Ochoa C, Alcasena F, Sjöström J, Marrs C, Ribeiro LM, Palaiologou P, Bentué Martínez C, Chuvieco E, Vega-García C, Rodrigues M (2025) Assessing human-caused wildfire ignition likelihood across Europe. Nat Hazards Earth Syst Sci 25(11):4713–4729 Getis A, Ord JK (1992) The analysis of spatial association by use of distance statistics. Geographical Anal 24(3):189–206 Gong D (2025) Higher Emissions Scenarios Increase Wildland–Urban Interface Fire Hazard in China. Sustainability 17(16):7409 Gong D, Sun L, Hu T (2024) Characterizing the occurrence of wildland-urban interface fires and their important factors in China. Ecol Ind 165:112179 Guo F, Innes JL, Wang G, Ma X, Sun L, Hu H, Su Z (2015) Historic distribution and driving factors of human-caused fires in the Chinese boreal forest between 1972 and 2005. J Plant Ecol 8(5):480–490 Guo Y, Wang J, Ge Y, Zhou C (2024) Global expansion of wildland-urban interface intensifies human exposure to wildfire risk in the 21st century. Sci Adv 10(45):eado9587 Haight RG, Cleland DT, Hammer RB, Radeloff VC, Rupp TS (2004) Assessing fire risk in the wildland-urban interface. J Forest 102(7):41–48 Hamrouni Y, Paillassa E, Chéret V, Monteil C, Sheeren D (2021) From local to global: A transfer learning-based approach for mapping poplar plantations at national scale using Sentinel-2. ISPRS J Photogrammetry Remote Sens 171:76–100 Han J, Shen Z, Ying L, Li G, Chen A (2015) Early post-fire regeneration of a fire-prone subtropical mixed Yunnan pine forest in Southwest China: Effects of pre-fire vegetation, fire severity and topographic factors. For Ecol Manag 356:31–40 He T, Lamont BB, Pausas JG (2019) Fire as a key driver of Earth's biodiversity. Biol Rev 94(6):1983–2010 Hutchinson MF (1995) Interpolating mean rainfall using thin plate smoothing splines. Int J Geographical Inform Syst 9(4):385–403 Jolly WM, Cochrane MA, Freeborn PH, Holden ZA, Brown TJ, Williamson GJ, Bowman DMJS (2015) Climate-induced variations in global wildfire danger from 1979 to 2013. Nat Commun 6(1):7537 Li C, Lin S, Wang Y, Chen L (2024a) Evaluation of regional integration in urban agglomeration area from the perspective of urban symbiosis. Sustainable Cities Soc 105:105338 Li Q, Li L, Zhang J, He X (2024b) Spatio-temporal characteristics and driving mechanisms of urban expansion in the Central Yunnan Urban Agglomeration. Land 13(9):1496 Li X, Li Y, Chen A, Gao M, Slette IJ, Piao S (2019) The impact of the 2009/2010 drought on vegetation growth and terrestrial carbon balance in Southwest China. Agric For Meteorol 269–270:239–248 Li Y, Pan Z, Gao C, Shi C, Lin S, Huang X (2025) Decoding wildland-urban interface fire regimes in Hong Kong, vol 113. Urban Forestry & Urban Greening, p 129095 Lian C, Xiao C, Feng Z, Ma Q (2024) Accelerating decline of wildfires in China in the 21st century. Front Forests Global Change 6:1252587 Lin H-W, McCarty JL, Wang D, Rogers BM, Morton DC, Collatz GJ, Jin Y, Randerson JT (2014) Management and climate contributions to satellite-derived active fire trends in the contiguous United States. J Geophys Research: Biogeosciences 119(4):645–660 Liu Q, Shan Y, Shu L, Sun P, Du S (2018) Spatial and temporal distribution of forest fire frequency and forest area burnt in Jilin Province, Northeast China. J Forestry Res 29(5):1233–1239 Liu Y, Xu S, Xiao R, Zhao Y (2025) Increasing fragmentation of the wildland–urban interface as it expands across China’s dryland landscape from 1990 to 2020. Landscape Ecol 40(12):231 Lv Q, Yang Z, Chen Z, Li M, Gao B, Yang J, Chen X, Xu B (2024) Crop residue burning in China (2019–2021): Spatiotemporal patterns, environmental impact, and emission dynamics. Environ Sci Ecotechnology 21:100394 Matthews S (2014) Dead fuel moisture research: 1991–2012. Int J Wildland Fire 23(1):78–92 Moreira F, Ascoli D, Safford H, Adams MA, Moreno JM, Pereira JM, Catry FX, Armesto J, Bond W, González ME (2020) Wildfire management in Mediterranean-type regions: paradigm change needed. Environ Res Lett 15(1):011001 Moreno M, Bertolín C, Arlanzón D, Ortiz P, Ortiz R (2023) Climate change, large fires, and cultural landscapes in the mediterranean basin: An analysis in southern Spain. Heliyon 9(6), e16941 Mu H, Li X, Wen Y, Huang J, Du P, Su W, Miao S, Geng M (2022) A global record of annual terrestrial Human Footprint dataset from 2000 to 2018. Sci Data 9(1):176 Nelson RM (2001) Water Relations of Forest Fuels. In: Johnson EA, Miyanishi K (eds) Forest Fires: Behavior and Ecological Effects. Academic, San Diego, CA, pp 79–149 Ord JK, Getis A (1995) Local spatial autocorrelation statistics: distributional issues and an application. Geographical Anal 27(4):286–306 Pan SJ, Yang Q (2010) A survey on transfer learning. IEEE Trans Knowl Data Eng 22(10):1345–1359 Pausas JG, Ribeiro E (2013) The global fire–productivity relationship. Glob Ecol Biogeogr 22(6):728–736 Qi X, Wei M, Wang Z, Jiang T, Wang P, Teng M, Yan Z (2022) Effects of the COVID-19 lockdown in Hubei, China: cessation of incense burning reduces regional landscape fire. Ecol Soc 27(3):5 Qin H, Tan Y, Shen T, Schaefer DA, Chen H, Zhou S, Xu Q, Zhu Y, Cheng J, Zhao G, Xu J (2024) Spatiotemporal trends of atmospheric dryness during 1980–2021 in Yunnan, China. Front Forests Global Change 7:1397028 Core Team R (2025) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing Radeloff VC, Helmers DP, Kramer HA, Mockrin MH, Alexandre PM, Bar-Massada A, Butsic V, Hawbaker TJ, Martinuzzi S, Syphard AD, Stewart SI (2018) Rapid growth of the US wildland-urban interface raises wildfire risk. Proceedings of the National Academy of Sciences 115(13), 3314–3319 Radeloff VC, Mockrin MH, Helmers D, Carlson A, Hawbaker TJ, Martinuzzi S, Schug F, Alexandre PM, Kramer HA, Pidgeon AM (2023) Rising wildfire risk to houses in the United States, especially in grasslands and shrublands. Science 382(6671):702–707 Ribeiro AFS, Santos L, Randerson JT, Uribe MR, Alencar AAC, Macedo MN, Morton DC, Zscheischler J, Silvestrini RA, Rattis L, Seneviratne SI, Brando PM (2024) The time since land-use transition drives changes in fire activity in the Amazon-Cerrado region. Commun Earth Environ 5(1):96 Rothermel RC (1972) A mathematical model for predicting fire spread in wildland fuels. U.S. Department of Agriculture, Intermountain Forest and Range Experiment Station, Ogden, UT, p 40 Schug F, Bar-Massada A, Carlson AR, Cox H, Hawbaker TJ, Helmers D, Hostert P, Kaim D, Kasraee NK, Martinuzzi S, Mockrin MH, Pfoch KA, Radeloff VC (2023) The global wildland–urban interface. Nature 621(7977):94–99 Segura-Garcia C, Alencar A, Arruda S, Bauman VL, Silva D, Conciani W, Oliveras Menor DE (2025) I., The fire regimes of the Cerrado and their changes through time. Philosophical Transactions of the Royal Society B: Biological Sciences 380(1924), 20230460 Senande-Rivera M, Insua-Costa D, Miguez-Macho G (2022) Spatial and temporal expansion of global wildland fire activity in response to climate change. Nat Commun 13(1):1208 Shen Z, Giljohann K, Liu Z, Pausas J, Rogers B (2025) Novel wildfire regimes under climate change and human activity: patterns, driving mechanisms and ecological impacts. Philosophical Transactions of the Royal Society B: Biological Sciences 380(1924), 20230446 Shi Z, Jiang D, Wang Y (2024) Spatiotemporal dependence of compound drought–heatwave and fire activity in China. Weather Clim Extremes 45:100695 Su W-H, Shi Z, Zhou R, Zhao Y-J, Zhang G-F (2015) The role of fire in the Central Yunnan Plateau ecosystem, southwestern China. For Ecol Manag 356:22–30 Syphard AD, Radeloff VC, Keeley JE, Hawbaker TJ, Clayton MK, Stewart SI, Hammer RB (2007) Human influence on California fire regimes. Ecol Appl 17(5):1388–1402 Tuia D, Persello C, Bruzzone L (2016) Domain Adaptation for the Classification of Remote Sensing Data: An Overview of Recent Advances. IEEE Geoscience Remote Sens Magazine 4(2):41–57 Vadrevu KP, Lasko K, Giglio L, Schroeder W, Biswas S, Justice C (2019) Trends in Vegetation fires in South and Southeast Asian Countries. Sci Rep 9(1):7422 Wang C, Wu D, Shen Z, Peng M, Ou X (2021) How do physical and social factors affect urban landscape patterns in intermountain basins in Southwest China? Landscape Ecol 36(7):1893–1911 Wang H, Jin B, Zhang K, Aktar S, Song Z (2022) Effectiveness in mitigating forest fire ignition sources: A statistical study based on fire occurrence data in China. Fire 5(6):215 Wang W, Zhao F, Wang Y, Huang X, Ye J (2023) Seasonal differences in the spatial patterns of wildfire drivers and susceptibility in the southwest mountains of China. Sci Total Environ 869:161782 Werth PA, Potter BE, Clements CB, Finney MA, Goodrick SL, Alexander ME, Cruz MG, Forthofer JA, McAllister SS (2011) Synthesis of knowledge of extreme fire behavior: volume I for fire managers. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station Wright MN, Ziegler A (2017) ranger: A Fast Implementation of Random Forests for High Dimensional Data in C + + and R. J Stat Softw 77(1):1–17 Wu C, Xu W, Xiao C, Wang Q, Yuan H, Dong J, Huang S, Xiong Y (2021) Dynamic change of recovery ratios and influencing factors of typical post-fire burn areas in central Yunnan Province. Resour Sci 43(12):2465–2474 [in Chinese] Xiahou M, Liu Y, Yang T, Shen Z (2024) Estimating potential vegetation distribution and restoration in a biodiversity hotspot region under future climate change. J Geog Sci 34(11):2128–2144 Xiahou M, Peng M, Shen Z, Wen Q, Wang C, Liu Y, Zhang Q, Peng L, Yu C, Ou X, Fang J (2025) Vegetation mapping of Yunnan Province by integrating remote sensing, field observations, and models. Sci China Earth Sci 68(3):836–849 Xu C, You C (2023) Agricultural expansion dominates rapid increases in cropland fires in Asia. Environ Int 179:108189 Xu W, Liu Y, Veraverbeke S, Wu W, Dong Y, Lu W (2021) Active Fire Dynamics in the Amazon: New Perspectives From High-Resolution Satellite Observations. Geophys Res Lett 48(20), e2021GL093789. Yang J, Huang X (2021) Earth Syst Sci Data 13(8):3907–3925The 30 m annual land cover dataset and its dynamics in China from 1990 to 2019 Ye J, Wu M, Deng Z, Xu S, Zhou R, Clarke KC (2017) Modeling the spatial patterns of human wildfire ignition in Yunnan province, China. Appl Geogr 89:150–162 Yin J, Ding W (2015) Analysis and application of ‘‘Degree’’in the mountain town process in Yunnan province. Geographical Sci Res 4(3):144–150 [in Chinese] Ying L, Cheng H, Shen Z, Guan P, Luo C, Peng X (2021) Relative humidity and agricultural activities dominate wildfire ignitions in Yunnan, Southwest China: Patterns, thresholds, and implications. Agric For Meteorol 307:108540 Young AM, Higuera PE, Duffy PA, Hu FS (2017) Climatic thresholds shape northern high-latitude fire regimes and imply vulnerability to future climate change. Ecography 40(5):606–617 Zamanialaei M, San Martin D, Theodori M, Purnomo DMJ, Tohidi A, Lautenberger C, Qin Y, Trouvé A, Gollner M (2025) Fire risk to structures in California’s Wildland-Urban Interface. Nat Commun 16(1):8041 Zhang H, Wang W, Ban Q (2025) Seasonal forest fire risk and key drivers in Yunnan Province: a machine learning approach. npj Nat Hazards 2(1):59 Zhang W, Yan X, Wang Q, Long T, Li X, Pu J, Ding Z (2022) Effects of prescribed burning on fire behavior of surface fuel in Pinus yunnanensis forest land. J Beijing Forestry Univ 44(5):69–76 [in Chinese] Zhao H, Zhang Z, Zhen S, Wang X, Yin Y (2025a) Multifactorial interactions contribute to contrasting wildfire trends at mid–high latitudes of the Northern Hemisphere. Agric For Meteorol 367:110507 Zhao J, Zheng B, Ciais P, Chen Y, Gasser T, Canadell JG, Zhang L, Zhang Q (2025b) Global warming amplifies wildfire health burden and reshapes inequality. Nature 647(8091):928–934 Zhu Z, Deng X, Zhao F, Li S, Wang L (2022) How environmental factors affect forest fire occurrence in Yunnan forest region. Forests 13(9):1392 Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterials.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9559449","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":636355315,"identity":"52dd8685-4ac9-4612-959a-f60e33ff36ba","order_by":0,"name":"Yang Lin","email":"","orcid":"","institution":"Yunnan University","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Lin","suffix":""},{"id":636355316,"identity":"6cb97095-abd4-4d79-b210-76686ceee3ce","order_by":1,"name":"Mingjian Xiahou","email":"","orcid":"","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Mingjian","middleName":"","lastName":"Xiahou","suffix":""},{"id":636355317,"identity":"44900fc3-9302-4b00-8d5f-e71651d787cc","order_by":2,"name":"Jiesheng Rao","email":"","orcid":"","institution":"Yunnan University","correspondingAuthor":false,"prefix":"","firstName":"Jiesheng","middleName":"","lastName":"Rao","suffix":""},{"id":636355318,"identity":"20aeaee3-3f82-4501-bfc2-c88aa1ec1094","order_by":3,"name":"Caifang Luo","email":"","orcid":"","institution":"Yunnan University","correspondingAuthor":false,"prefix":"","firstName":"Caifang","middleName":"","lastName":"Luo","suffix":""},{"id":636355319,"identity":"6a174406-13d0-4b60-ad9c-faa37dd4fac8","order_by":4,"name":"Tao Yang","email":"","orcid":"","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Yang","suffix":""},{"id":636355320,"identity":"667bfefe-20ba-44fc-928c-7ad4a6746298","order_by":5,"name":"Zehao Shen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4klEQVRIiWNgGAWjYDACCQY2hg8MDAkgNmMDiDxAhBbGGSRrYeYhSYv87PZnj23b7uXxz25gezizjUGO70YC4+cCPFoM7hxIN85tKy6WuHOA3XBjG4Ox5I0EZukZ+LRIJByTzm1LSGy4kcAm+bCNIXEDkAFyKm6HzUhsk7YEapkP1VJPUAvDjWQ2aUagFpBKSaDDEgwIaTG4kcYm2XMuodjwRmK74YxzEoYzzzxslsbvsPRnEj/KEvLkbiQfe9hTZiPPdzz54Ge8DkMAxjYGUDTBoocowEa0ylEwCkbBKBhZAAAUkkx7oEr9mQAAAABJRU5ErkJggg==","orcid":"","institution":"Peking University","correspondingAuthor":true,"prefix":"","firstName":"Zehao","middleName":"","lastName":"Shen","suffix":""}],"badges":[],"createdAt":"2026-04-29 02:38:26","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9559449/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9559449/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108789174,"identity":"eeb5a46b-b43e-43e9-a654-c2335392e32b","added_by":"auto","created_at":"2026-05-08 12:04:14","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":26923310,"visible":true,"origin":"","legend":"\u003cp\u003eStudy area and baseline wildfire and land-cover conditions on the Central Yunnan Plateau (CYP), China. (a) Location of the CYP. (b) Wildfire burn frequency during 1990-2023. (c) Land-cover map in 2020. Photographs show representative fire-scar and post-fire landscapes in the region.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-9559449/v1/de8c04051a7e389af84a9a4f.png"},{"id":108789176,"identity":"ce77e004-3cc4-4ce9-87cb-75cb7b7519be","added_by":"auto","created_at":"2026-05-08 12:04:14","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1467638,"visible":true,"origin":"","legend":"\u003cp\u003eFramework for interface delineation and fire-regime analysis. WUI and WCI were extracted by intersecting wildland vegetation with buffered urban land (\u003cem\u003ed\u003c/em\u003e\u003csub\u003e\u003cem\u003eu\u003c/em\u003e\u003c/sub\u003e) and cropland (\u003cem\u003ed\u003c/em\u003e\u003csub\u003e\u003cem\u003ec\u003c/em\u003e\u003c/sub\u003e) layers from time-series land-cover data. Historical fire records from 1990 to 2023 were then used to analyze spatiotemporal patterns, seasonality, and driving mechanisms across prefecture, county, and town scales.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-9559449/v1/1dede83acab4e2992262cf6f.png"},{"id":108789177,"identity":"f41953cc-b15c-40cb-80d0-65f047bb7fd4","added_by":"auto","created_at":"2026-05-08 12:04:14","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2818184,"visible":true,"origin":"","legend":"\u003cp\u003eInterannual burned-area dynamics in the CYP and its interface zones from 1990 to 2023. Annual burned area is shown for (a) the entire CYP, (b) WCI, and (c) WUI. (d) Proportional contribution of WCI and WUI to total burned area. Thin colored lines show annual values, and solid lines with shaded bands show linear regression fits and 95% confidence intervals. The vertical dashed line marks the peak fire year. Asterisks indicate significance: * p \u0026lt; 0.05, ** p \u0026lt; 0.01, and *** p \u0026lt; 0.001\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-9559449/v1/55d490b8642d314fd2210855.png"},{"id":108806920,"identity":"04f82d8e-9a27-4b2d-b982-c1b48288589e","added_by":"auto","created_at":"2026-05-08 15:29:37","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":2591672,"visible":true,"origin":"","legend":"\u003cp\u003eSeasonal patterns and long-term trends of fire activity in the CYP, WCI, and WUI from 1990 to 2023. (a) Mean monthly proportion of burned area; error bars show standard deviations. (b) Interannual variation and linear trends in seasonal burned-area contribution for spring, summer, autumn, and winter. Solid lines show fitted trends, and asterisks indicate significance at \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-9559449/v1/dc6482f485ba10a2dae604fa.png"},{"id":108789179,"identity":"a877dfa1-5e15-4aba-bd50-a5c03401c3ae","added_by":"auto","created_at":"2026-05-08 12:04:14","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":3261745,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial hotspots and coldspots of fire frequency in WUI and WCI from 1990 to 2023. (a) Fire hotspots in WUI. (b) Fire hotspots in WCI. Hotspots and coldspots were identified using the Getis-Ord Gi* statistic at 90%, 95%, and 99% confidence levels.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-9559449/v1/ff16bd459b755d99c570cf9a.png"},{"id":108807549,"identity":"df8c44dd-e774-4c56-8448-768b4b6fdde0","added_by":"auto","created_at":"2026-05-08 15:30:36","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":4035278,"visible":true,"origin":"","legend":"\u003cp\u003eAnnual burned-area trends for major vegetation types in the CYP from 1990 to 2023. Panels show burned-area variation in (a) planted and economic forests, (b) warm-temperate coniferous forests, (c) savannas, and (d) shrublands for the CYP, WCI, and WUI. Dashed lines show linear regression fits. Asterisks indicate significance: * p \u0026lt; 0.05, ** p \u0026lt; 0.01, and *** p \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-9559449/v1/7324b3c5a5deb698ec88c728.png"},{"id":108789181,"identity":"cc59f7ed-3c5a-4821-a4f5-8520131baabb","added_by":"auto","created_at":"2026-05-08 12:04:14","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":2554994,"visible":true,"origin":"","legend":"\u003cp\u003eRandom Forest results for fire occurrence in WCI and WUI. (a, c) Relative importance of predictor variables in WCI and WUI, respectively. (b, d) Partial dependence plots showing nonlinear responses of fire probability to key predictors in WCI and WUI, respectively. Colors in (a) and (c) indicate factor groups: climate, human activity, fuel, and topography.\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-9559449/v1/4c83900b8caf45c7d15ef2a5.png"},{"id":108810380,"identity":"dca3693d-ebea-4c08-9e17-2e5b2351b987","added_by":"auto","created_at":"2026-05-08 15:58:42","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":40979257,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9559449/v1/a396a482-cf7a-49c5-8bb5-ebfd8e436b46.pdf"},{"id":108806784,"identity":"278b7dfe-1f4e-4acd-ad86-ab405871b976","added_by":"auto","created_at":"2026-05-08 15:29:28","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":2308718,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterials.docx","url":"https://assets-eu.researchsquare.com/files/rs-9559449/v1/87baa03b9f310260fcb68fd0.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"The temporal trajectories and driving forces of wildfire regimes differing between wildland-cropland and wildland-urban interfaces on the Central Yunnan Plateau","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eWildfire is a major disturbance in terrestrial ecosystems and one of the most widespread natural hazards. It reshapes vegetation structure, alters biodiversity patterns (Bowman et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; He et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), and strongly affects energy budget and elements cycling on the ground from local to global scales (Archibald et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Meanwhile, the growing threats that wildfire poses to environmental quality and human health have attracted increasing attention (Radeloff et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e2025b\u003c/span\u003e). Under intensifying climate change and human disturbance, fire regime shifting in various ecosystems are becoming increasingly evident across regional to global scales (Shen et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOver the past three decades, although some studies have reported a decline in global total burned area (Andela et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), the economic losses and ecological threats caused by increasingly frequent extreme wildfire events have continued to rise, affecting ever larger populations (Cunningham et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This trend is driven in part by global warming and the intensification of regional extreme droughts, which have increased fluctuations in vegetation biomass and fuel flammability (Abatzoglou and Williams, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Senande-Rivera et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and also to a large extent by the spatial expansion of human activities in multiple forms, including housing, farming, public infrastructure, and tourism facilities (Radeloff et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). The spread of population and assets into vegetated areas has made fire impacts more likely to concentrate at the interface between natural vegetation and human land use, thereby exacerbating disaster consequences (Schug et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Zamanialaei et al., \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). As a result, the wildland-urban interface (WUI) has increasingly become a focal area in wildfire research and management (Haight et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; Chen et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). For example, by combining a global WUI time-series map for 1985\u0026ndash;2020 with burned area and population distribution data for 2001\u0026ndash;2020, Chen et al. (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) found that about 12.54% of global WUI areas, representing approximately 10.11\u0026nbsp;million people, were located within a 4800 m wildfire threat buffer. Carlson et al. (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) further showed that the proportion of wildfire-exposed buildings that were destroyed increased from about 10% in 2002\u0026ndash;2012 to around 32% in 2013\u0026ndash;2022, indicating that wildfire events are becoming more destructive in addition to exposing more infrastructure. Recent national and regional assessments within China have advanced our understanding of wildfire risks in the traditional WUI (Gong et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eHowever, the WUI framework alone may not adequately capture wildfire risk in many landscapes with long agricultural histories, including China, India, Southeast Asian countries, as well as the Brazilian Cerrado, where urban peripheries are often dominated by managed ecosystems such as croplands and plantations, instead of clustering houses (Vadrevu et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Xu and You, \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Here, the transition between human activities and wildlands occurs mainly along the margins of these managed landscapes. We refer to this configuration as the wildland-cropland interface (WCI). The interweaving of croplands, orchards, tea and rubber plantations, nurseries, secondary forests, shrublands, and grasslands creates a mosaic landscape that is also prone to frequent wildfires (Segura-Garcia et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). More broadly, agricultural expansion and land-use intensification have reshaped fire regimes in many managed landscapes over recent decades. Land clearing, plantation development, and agricultural burning have increased human influences on fire activity and altered patterns of fire occurrence and spread (Archibald et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Ribeiro et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). In these mosaics, human-related ignition sources and fuel management may produce fire probabilities and burned-area responses that differ from those observed in WUI landscapes (Ying et al., \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Gelabert et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). However, targeted research on wildfire dynamics and driving mechanisms in WCI remains scarce, and the similarities and differences in how human activities shape fire processes across WUI and WCI landscapes remain poorly understood. Comparing wildfires in WCI and WUI is meaningful not only to explore their formats of spatial distribution and temporal dynamics of anthropogenic burning, but also to disentangle the driving forces in natural and social-cultural aspects of these agriculture-intensive regions. Against this background, we use the fire-prone Central Yunnan Plateau as a case study to compare wildfire dynamics in WUI and WCI landscapes, relate them to regional fire trends, and identify differences in the mechanisms governing fire occurrence and burned-area percentage.\u003c/p\u003e \u003cp\u003eThe mountainous region of Southwest China is one of the major wildfire hotspots in China and forms part of the broader high-fire belt of the Indochina Peninsula, with Yunnan Province serving as a core region of high wildfire activity in Southwest China (Ying et al., \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Zhang et al., \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), and the Central Yunnan Plateau being the most fire-prone area within the province (Su et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). This region has the highest population density and agricultural activity in Yunnan and has also experienced rapid urbanization in recent decades (Wang et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2024a\u003c/span\u003e). Urban expansion, denser road networks, and changing agroforestry management practices have increased the contact boundaries between natural vegetation and both developed land and cropland. As a result, WUI and WCI landscapes coexist in the region and show increasingly complex spatial configurations (Li et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024b\u003c/span\u003e). Although previous studies have examined forest fires in Yunnan, most have focused on regional-scale assessments or specific fire events (Han et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Wu et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Research on wildfire dynamics within landscape interfaces is still limited. In particular, the fire activity patterns in WUI and WCI landscapes and the different effects of anthropogenic and natural drivers remain poorly understood (Wang et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Deng et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). A comparative analysis of wildfire evolution in WUI and WCI interfaces on the Central Yunnan Plateau can therefore help clarify how human and environmental factors contribute to fire risk in these heterogeneous landscapes. Such knowledge is important for developing differentiated vegetation management strategies and targeted wildfire mitigation policies.\u003c/p\u003e \u003cp\u003eAccordingly, this study examines long-term wildfire dynamics within WUI and WCI landscape interfaces on the Central Yunnan Plateau. We address three questions: 1) How did burned area vary interannually and spatially across the Central Yunnan Plateau, the WCI, and the WUI during 1990\u0026ndash;2023? 2) How do wildfire seasonality and distributions along topographic and vegetation gradients differ between WCI and WUI landscapes? 3) What similarities and differences characterize the dominant drivers of fire occurrence and burned-area percentage in these two interface types?\u003c/p\u003e"},{"header":"2. Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Study area\u003c/h2\u003e \u003cp\u003eThe Central Yunnan Plateau (CYP) is located in central Yunnan Province, China. It includes Kunming, Dali, Yuxi, Qujing, the Chuxiong Yi Autonomous Prefecture, and the northern part of the Honghe Hani and Yi Autonomous Prefecture, including Mengzi, Gejiu, Jianshui, Kaiyuan, Mile, Luxi, and Shiping. The region covers about 111400 km\u003csup\u003e2\u003c/sup\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), accounting for roughly 30% of the total area of Yunnan Province. As a core part of the Yunnan-Guizhou Plateau, the CYP has an average elevation of about 2000m and a typical plateau mountain-basin mosaic landscape. The climate belongs to the subtropical plateau monsoon type, with a clear dry-wet seasonal pattern. More than 80% of annual precipitation falls in the wet season from June to October. The dry season, from November to the following May, is affected by the southern branch of the subtropical westerly jet. Warm and dry conditions, together with high wind speeds during this period, create a critical fire-weather window. Influenced by intensive anthropogenic activities, the region contains diverse vegetation types. The primary fuel types consist of secondary coniferous forests dominated by \u003cem\u003ePinus yunnanensis\u003c/em\u003e, semi-moist evergreen broad-leaved forests, and xerophytic shrub-grasslands, all of which exhibit significant spatial heterogeneity in their horizontal and vertical structures (Zhang et al., \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). This heterogeneous fuel matrix, coupled with seasonal drought, fosters frequent wildfire activity, while the complex topographic gradients provide a highly variable environment for fire propagation (Ying et al., \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Qin et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Data sources and processing\u003c/h2\u003e \u003cp\u003e \u003cb\u003eWUI and WCI data\u003c/b\u003e:\u003c/p\u003e \u003cp\u003eWe used 30m land-cover data for the CYP from 1990 to 2023 from the China Land Cover Dataset (CLCD) (Yang and Huang, \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Based on the Google Earth Engine (GEE) platform, we mapped the spatial and temporal distributions of the Wildland-Urban Interface (WUI) and the Wildland-Cropland Interface (WCI) using buffer analysis.\u003c/p\u003e \u003cp\u003eThe WUI is usually defined as the area where human infrastructure and built-up land meet or intermingle with natural vegetation (Chen et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). For the delineation of the WUI, we established a 2400 m geographic buffer around urban built-up land and extracted its intersection with natural vegetation (including forests, grasslands, and shrublands) (Carlson et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Correspondingly, the WCI is defined as the transition zone where farmlands and plantations directly interface with natural vegetation. Following previous work on agricultural fire dynamics and forest edge effects (Cochrane, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Broadbent et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), and with support from spatial sensitivity tests, we used a 500 m buffer around croplands to delineate the WCI. Sensitivity tests using 250 m and 1000 m thresholds showed highly consistent temporal patterns of fire dynamics, supporting the 500 m baseline (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). We then applied connectivity analysis to remove small fragmented patches and improve the spatial continuity of the interface zones (Guo et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cb\u003eWildfire activity data\u003c/b\u003e:\u003c/p\u003e \u003cp\u003eWe used burn area and burned-area percentage to estimate wildfire activity in the CYP. Burned area was obtained from a monthly Landsat 30 m burned-area product for 1990\u0026ndash;2023 developed by Xiahou \u0026amp; Shen(2025), which was clipped to the spatial extent of the CYP. This product was generated by training a Random Forest model with filtered active fire detections from MODIS and VIIRS and pre- and post-fire spectral changes from Landsat imagery. Burned pixels were identified using a probability threshold of 0.90. Water bodies, built-up land, and other non-vegetated surfaces were masked to improve mapping accuracy. To calculate burned-area percentage, we divided the CYP into 1 \u0026times; 1 km grid cells and calculated the proportion of burned area within each grid cell.\u003c/p\u003e \u003cp\u003e \u003cb\u003eClimate data\u003c/b\u003e:\u003c/p\u003e \u003cp\u003eMonthly gridded datasets of precipitation, mean temperature, maximum temperature, and relative humidity at a 30m resolution for 1990\u0026ndash;2023 were generated using the Thin Plate Spline interpolation method (Hutchinson, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e1995\u003c/span\u003e). The interpolation used observations from 124 meteorological stations in and around the CYP, with digital elevation model (DEM) data included as the main topographic covariate. Elevation data came from the SRTM dataset (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://srtm.csi.cgiar.org\u003c/span\u003e\u003cspan address=\"http://srtm.csi.cgiar.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), with a spatial resolution of 3 arc-seconds to facilitate interpolation correction (Xiahou et al., \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). We aggregated all meteorological variables into annual dry-season indicators from November to April of the following year to represent long-term variation in fire-season climate.\u003c/p\u003e \u003cp\u003e \u003cb\u003eHuman activity data\u003c/b\u003e:\u003c/p\u003e \u003cp\u003eThe human activity variables used in this study included population density (POP), gross domestic product (GDP), and the Human Footprint Index (HFI). GDP and POP data were obtained from the Resource and Environmental Science and Data Center of the Chinese Academy of Sciences (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.resdc.cn/\u003c/span\u003e\u003cspan address=\"https://www.resdc.cn/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) as annual 1 km gridded datasets. HFI data were taken from Mu et al. (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) as annual grids with a 1 km spatial resolution. We also calculated building density from the CLCD dataset as the proportion of built-up land within each 1 km grid cell.\u003c/p\u003e \u003cp\u003e \u003cb\u003eLandcover map\u003c/b\u003e \u003c/p\u003e \u003cp\u003eWe developed a decadal vegetation classification dataset for the CYP at 30m resolution for 1990\u0026ndash;2023. Following the Ecosystem List of Yunnan Province (2018 Edition), we selected 15066 high-quality samples from more than 25000 land-cover and land-use plots surveyed between 2018 and 2020. The classification used Landsat spectral indices from visible, near-infrared, and shortwave infrared bands, including the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Normalized Difference Built-up Index (NDBI), and Land Surface Water Index (LSWI). Topographic factors from ASTER GDEM and climate data were also included. We used a 7:3 ratio for training and validation and applied a Random Forest algorithm together with a sample transfer learning strategy. The transfer procedure used spectral similarity analysis to propagate high-confidence samples from the baseline year to historical images. Stable samples were identified by comparing reference and target images with Euclidean distance and spectral angle distance. This procedure helped optimize the models and maintain consistency across the multi-temporal classification results. These steps produced the preliminary long-term classification map (Pan and Yang, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Tuia et al., \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Hamrouni et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). We then refined the maps using forest inventory data from the Yunnan Academy of Forest Inventory and Planning, including information on plantations, the distribution of \u003cem\u003ePinus yunnanensis\u003c/em\u003e and \u003cem\u003ePinus armandii\u003c/em\u003e, and semi-moist evergreen broad-leaved forests, together with potential vegetation maps and expert knowledge. The final vegetation maps had an average overall accuracy of 0.72, and the accuracy for vegetation subtypes was 0.69 (Xiahou et al., \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo ensure spatial consistency across multiple sources, all spatial data were reprojected to the Southern Albers Equal Area Conic projection and resampled to a uniform 30m resolution.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Statistical analysis\u003c/h2\u003e \u003cp\u003eWe compared fire-pattern differences between WUI and WCI at multiple scales. The analysis focused on three aspects: spatiotemporal evolution, seasonal characteristics, and driving factors at the prefecture, county, and town levels. This framework was used to examine wildfire characteristics and the mechanisms behind fire activity in the two interface types (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eTemporal trend analysis\u003c/b\u003e \u003c/p\u003e \u003cp\u003eTo quantify the interannual variability of wildfire activity from 1990 to 2023, we calculated annual burned area and burned-area percentage for three spatial domains: the CYP, the WCI, and the WUI. We used Ordinary Least Squares (OLS) linear regression to evaluate long-term trends in these wildfire indicators. The slope, coefficient of determination (R\u003csup\u003e2\u003c/sup\u003e), and \u003cem\u003ep\u003c/em\u003e-value were used to describe the magnitude and reliability of each trend. Statistical significance was defined at p\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003cp\u003e \u003cb\u003eHotspot (Getis-Ord Gi*) analysis\u003c/b\u003e \u003c/p\u003e \u003cp\u003eWe used the Getis-Ord Gi* statistic to identify significant spatial clusters of fire frequency from 1990 to 2023. This local spatial autocorrelation method detects clusters of high or low values by calculating a standardized z-score for each spatial unit and its neighborhood relative to the global mean (Getis and Ord, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e1992\u003c/span\u003e; Ord and Getis, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e1995\u003c/span\u003e). We compared Gi* statistic distributions across several grid scales in ArcGIS Pro 3.0.1 to test the stability and significance of the hotspot patterns. Based on this spatial sensitivity analysis, we selected an 8 km grid as the optimal scale for representing regional wildfire clustering in the CYP. Spatial units were classified as hotspots, coldspots, or non-significant areas according to their \u003cem\u003ez\u003c/em\u003e-scores and \u003cem\u003ep\u003c/em\u003e-values, and were mapped at 90%, 95%, and 99% confidence levels.\u003c/p\u003e \u003cp\u003e \u003cb\u003eVariable extraction\u003c/b\u003e \u003c/p\u003e \u003cp\u003eTo identify differences in the drivers of fire occurrence and burned-area percentage, we divided the study area into 1 \u0026times; 1 km grid cells and constructed two annual fire-response variables. Fire occurrence was defined as whether a grid cell burned in a given year (0/1). Burned-area percentage was defined as the proportion of each grid cell that burned in a given year. Predictor selection followed a climate-topography-fuel-human activity framework, with candidate variables chosen from established wildfire studies (Zhang et al., \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e2025a\u003c/span\u003e). Because wildfire activity in the CYP is strongly concentrated in the dry season, we adapted the predictor set to regional fire-season conditions. The selected predictors included four groups. Meteorological variables included dry-season precipitation, dry-season mean temperature, dry-season maximum temperature, and dry-season relative humidity, which represented hydrothermal constraints on fire. Topographic variables included elevation relief and mean slope. Fuel variables included the proportions of coniferous forest and shrubland and pre-fire NDVI, which described vegetation structure. Human activity variables included population density, GDP, road density, the human footprint index, and building density, which represented anthropogenic ignition pressure and land-surface modification. Considering the complex nonlinear relationships and potential interactions among wildfire drivers, we used Random Forest models (Breiman, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2001\u003c/span\u003e). With the \"ranger\" package in R (Wright and Ziegler, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), we built classification models for fire occurrence and evaluated them with AUC, and regression models for burned-area percentage and evaluated them with R\u003csup\u003e2\u003c/sup\u003e. Variable importance and partial dependence plots (PDPs) were used to characterize the nonlinear responses of the main predictors.\u003c/p\u003e \u003cp\u003eAll analyses were conducted in ArcGIS Pro 3.0.1 and R 4.5.2 (R Core Team, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Spatiotemporal patterns of burned area\u003c/h2\u003e \u003cp\u003eOverall, burned area in the CYP, WCI, and WUI followed a pattern of increase followed by decline (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Regional fire activity peaked in 2010, when burned area in the CYP reached 715 km\u003csup\u003e2\u003c/sup\u003e, or 0.64% of the regional area. After 2010, the annual mean burned area in the WCI was 136.26 km\u003csup\u003e2\u003c/sup\u003e, lower than the 1990s mean of 149.64 km\u003csup\u003e2\u003c/sup\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). In contrast, the WUI burned area in the 2010s was 64.55 km\u003csup\u003e2\u003c/sup\u003e, higher than the 1990s mean of 32.88 km\u003csup\u003e2\u003c/sup\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec). The WCI accounted for most burned area in the CYP over the study period. When 2010 was used as a temporal boundary, the contribution of the two interfaces changed. Before 2010, the burned-area proportions of both WCI and WUI increased, especially in the WUI. After 2010, both proportions declined, but the decline was much stronger in the WCI, while the WUI remained relatively stable (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed).\u003c/p\u003e \u003cp\u003eAt different administrative levels, wildfire activity showed similar interannual fluctuations, with an increase before 2010 and a decline afterward (Fig. S3). WCI and WUI shared this broad temporal pattern, but differed in burned-area magnitude and proportional contribution (Fig. S4a-e). Total burned area increased from the prefecture level to the town level, and towns contributed the largest share of total burned area (Fig. S4a-c). Burned-area proportions also changed around 2010, but their administrative distribution differed between the two interfaces (Fig. S4d-e). The WUI proportion was higher than the WCI proportion at the prefecture level, whereas the WCI proportion was higher at the county and town levels. County-level units had the highest burned-area proportions in both interface types, indicating stronger wildfire risk in county-level interface landscapes.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFire activity in the CYP, WCI, and WUI showed a strong seasonal pattern (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). The multi-year monthly distribution was unimodal, with burned area concentrated in the dry season from November to the following April and peaking in March. After May, fire activity dropped sharply and stayed low through summer and autumn. Seasonal trends differed among regions (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb). Summer fires increased across the study area, although their absolute contribution remained small. In the WUI, the seasonal center shifted toward spring. After 2000, the proportion of spring fires in the WUI increased with fluctuations, while the winter contribution decreased. In the WCI, winter and spring remained dominant, and the seasonal structure changed little over the past three decades.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eHotspot analysis showed clear spatial differences between the two interfaces (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). In the WCI, high-frequency fire zones were concentrated mainly in Honghe Prefecture in the southern part of the study area, forming large and continuous clusters. Central and northern areas were mostly non-significant or contained isolated hotspots (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb). In the WUI, hotspots were more dispersed and polycentric, with significant clusters around highly urbanized areas such as Kunming, Yuxi, Qujing, and Honghe (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea). Hotspots in Dali and Chuxiong were more fragmented, but several independent high-frequency fire zones were still present. Overall, WUI fire hotspots were more widely distributed across the CYP.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFire activity also differed among vegetation types and environmental gradients (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e, Fig. S5). For each vegetation category, burned area changed in stages around 2010 (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea-d). Before 2010, burned area generally increased across all vegetation types, with especially large increases in shrublands in both the WCI and WUI. After 2010, burned area declined across all vegetation types. The decline in savannas was faster than their earlier increase. Warm-temperate coniferous forests and shrublands still had high peak values in 2010 and maintained large contribution proportions, indicating that they were the main fire-carrying vegetation types in the interface landscapes. topographic gradients also affected fire distribution (Fig. S5). Along the elevation gradient, burned area was concentrated mainly at mid-elevations, with the 1600-2000m range contributing the largest proportion. WCI fires had a higher proportion at low elevations below 1200m, whereas WUI fires were more concentrated at mid-to-high elevations of 1600-2400m. Along the slope gradient, the burned-area proportion increased with slope and reached its maximum on steep slopes above 25 degrees.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Contrasting drivers of fire occurrence and burned-area percentage\u003c/h2\u003e \u003cp\u003eRandom Forest importance rankings and PDPs showed that fire occurrence in the CYP and both interface types was mainly controlled by climate. Climatic variables accounted for about 44% of total importance across the three spatial domains, and dry-season mean temperature was consistently the most important predictor (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e, Fig. S6a-b). PDPs showed that fire occurrence increased strongly when temperature exceeded about 12℃. Secondary controls differed between the two interfaces. Population density ranked higher in the CYP and WCI, with importance values of 9.1% and 8.9%, respectively. Its PDPs showed higher fire probabilities at very low and very high population densities, and lower probabilities at intermediate densities. In the WUI, elevation relief was more important, and fire occurrence increased with terrain complexity. Thus, fire occurrence in the CYP and WCI showed a stronger human influence, whereas WUI fire occurrence was more closely related to topographic conditions.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eBurned-area percentage was strongly constrained by atmospheric moisture in the CYP, WCI, and WUI (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e, Fig. S6c-d). Relative humidity was the most important predictor in all models, with contribution rates of 18.6%, 24.5%, and 14.4% in the CYP, WCI, and WUI, respectively. PDPs showed high burned-area percentage when dry-season relative humidity was below 60%, followed by a rapid decline above this threshold. Although this moisture constraint was shared by all three spatial domains, the secondary controls differed. In the WCI, climatic factors dominated the model, with a cumulative contribution of 57.4%, and temperature ranked just behind humidity. In the WUI, burned-area percentage was more strongly related to fuel and human activity, especially coniferous forest proportion (9.9%) and population density (9.7%). PDPs showed that WUI burned-area percentage increased with coniferous forest cover and rose more sharply in densely populated areas, particularly above about 400 persons/km\u003csup\u003e2\u003c/sup\u003e. These patterns indicate that large fires in the WCI are mainly climate-controlled, whereas large fires in the WUI are linked to moisture, fuel structure, and human disturbance.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Spatiotemporal shifts of interface wildfires\u003c/h2\u003e \u003cp\u003eOur results identify 2010 as a major turning point in wildfire dynamics on the CYP. Before 2010, burned area generally increased despite strong interannual fluctuations, whereas the 2010 peak likely reflected the historic extreme drought that affected Yunnan Province during that year (Li et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Under such extreme drought conditions, the sharp decline in fuel moisture substantially increased vegetation flammability and facilitated fire spread at the regional scale, highlighting the high sensitivity of wildfire activity to climatic anomalies (Pausas and Ribeiro, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Young et al., \u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). The overall decline in burned area after 2010, by contrast, points to the effectiveness of policy intervention, particularly the marked improvement in administrative fire prevention following the promulgation of the Forest Fire Prevention Regulations in 2009 and reforms to the forest public security system. Climate data show that precipitation remained low and temperature continued to rise after 2010 in the CYP (Fig. S7), yet total burned area across the region still decreased, further underscoring the effectiveness of policy regulation (Guo et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Lian et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This trajectory, in which extremes are triggered by climatic anomalies and subsequent declines are shaped by policy intervention, reflects the complex interplay between natural controls and anthropogenic management in wildfire dynamics. Within the two landscape interfaces, WCI consistently accounted for the larger share of burned area, indicating that ignitions and fuel management associated with agricultural activities remained the main sources of fire in the agroforestry mosaic of the CYP (Ying et al., \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). This interpretation is consistent with studies from other agricultural or managed landscapes, where cropland and other managed fires contribute substantially to total fire activity, and where fires are closely linked to agricultural expansion and field clearing along forest edges (Lin et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Xu et al., \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The marked reduction in WCI burned area after 2010 suggests that policy measures directly constrained agricultural fire sources and partly offset climatic risk. In contrast, burned area in the WUI remained relatively stable rather than declining substantially, underscoring the more complex effects of land-use change. Although fire regulation was strengthened, continued urban expansion led to a steady increase in WUI area and a persistent extension of the contact zone between human activities and forest fuels (Fig. S2). As a result, wildfire risk is increasingly shifting toward areas with higher population and asset density, thereby increasing potential societal exposure to wildfire hazards (Radeloff et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Liu et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The multi-scale analysis further showed that town units were the primary zones of fire occurrence within the interface landscapes. At this scale, fragmented built-up land increased the contact boundaries between anthropogenic activities and combustible fuels, creating more opportunities for both ignition and fire spread (Liu et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). In addition, the proportion of WUI fires was higher around prefecture-level capitals, whereas WCI fires accounted for a larger share at the county and town levels. This pattern reflects the influence of the urbanization gradient on fire regimes, the higher proportion of WUI fires around prefecture-level cities was associated with urban encroachment on and disturbance of natural vegetation, whereas the higher proportion of WCI fires at county and town levels was mainly linked to traditional agricultural burning. Such cross-scale heterogeneity in fire distribution suggests that spatial fire management should adopt a hierarchical and coordinated strategy.\u003c/p\u003e \u003cp\u003eSpatial hotspot analysis revealed further contrasts between the two interface types. WCI fires showed a unipolar clustering pattern centered on Honghe Prefecture, whereas WUI fires exhibited a multi-nodal distribution around cities such as Kunming and Yuxi. This contrast further highlights the role of rapid urbanization in reshaping the spatial pattern of wildfire risk.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Seasonal evolution and landscape heterogeneity of wildfires in WCI and WUI\u003c/h2\u003e \u003cp\u003eSeasonal wildfire patterns on the CYP are jointly shaped by monsoon-controlled hydroclimatic seasonality and human activity. During winter and spring, precipitation deficits and low relative humidity rapidly dry fine fuels, creating favorable conditions for fire occurrence (Zhu et al., \u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). These conditions coincide with peaks in spring ritual burning and agricultural fire use, which further elevate ignition probability (Liu et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Qi et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Collectively, human activities and meteorological conditions determine the seasonal wildfire patterns of the CYP. These drivers are fundamentally consistent with the seasonal causes of fire in other regions. In Mediterranean climate zones, fires are also highly seasonal and often concentrated during summer droughts with high temperatures or at the end of the dry season (Moreira et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Moreno et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Notably, our results reveal a significant long-term increase in summer fire activity. This shift may reflect the growing influence of compound heat-drought extremes, which weaken the fire-suppressing effect of the rainy season and extend the seasonal window of flammable conditions (Shi et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). At the same time, increasing spring fire activity in the WUI suggests that urban expansion is moving residential and production activities closer to forest edges, thereby increasing the likelihood of anthropogenic ignitions under high fire-weather conditions (Guo et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Conversely, the seasonal structure of the WCI has remained relatively stable over the long term, which may be closely related to the periodic inertia of agricultural fire use habits. Fires were concentrated mainly in mid-elevation and steep-slope environments, especially in the WUI, where continuous fuels and high human accessibility frequently coincide (Ye et al., \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Additionally, the policy of urban expansion toward mountain areas in Yunnan Province has increased the probability of fire at mid to high elevations (Yin and Ding, \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Steep slopes constitute high incidence environments by accelerating fire spread and increasing the difficulty of suppression. In terms of vegetation types, Warm-temperate coniferous forest, savanna and shrublands constitute the primary carriers of combustion (Zhang et al., \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). \u003cem\u003ePinus yunnanensis\u003c/em\u003e is rich in oils and its litter dehydrates easily, resulting in extremely high flammability. Between 1990 and 2020, the area of warm temperate coniferous forests contracted after 2000, while the area of shrublands exhibited an increasing trend (Fig. S8). This reflects that current forest fire prevention policies have effectively suppressed coniferous forest fires but may have neglected the expanding shrubland areas to some extent. The rising proportion of shrubland fires not only reveals that edge habitats in interfaces are more likely to become ignition points under human interference but also indicates that the lack of targeted management for shrubland fuels has amplified the impact of fire disturbances through edge effects.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Divergence and evolution of driving mechanisms\u003c/h2\u003e \u003cp\u003eMeteorological factors were the dominant controls on both fire occurrence and burned-area percentage. Temperature and relative humidity ranked first in the occurrence and burned-area percentage models, respectively, and both showed clear nonlinear threshold responses (Badia et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). Fire occurrence increased sharply when mean dry-season temperature exceeded 12℃, whereas burned-area percentage expanded when dry-season relative humidity dropped below 60%. These thresholds are consistent with the basic ecological controls of fire, high temperatures accelerate the drying of fine fuels and litter, thereby facilitating ignition (Matthews, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), whereas very low atmospheric moisture allows fires to spread over larger areas (Jolly et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). This result is broadly consistent with recent studies on WUI fires in China, which likewise identified meteorological conditions as the dominant controls on WUI fire occurrence and hazard (Gong, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eHowever, under a unified climatic background, the WCI and WUI demonstrate differentiated driving mechanisms. Fire occurrence in the WCI is more dependent on population density, reflecting its anthropogenic driving characteristics dominated by high-intensity agricultural burning such as crop residue burning (Lv et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). In contrast, fire occurrence in the WUI shows stronger topographic dependency. The limiting effects of elevation relief and slope indicate that fire occurrence in this region is more constrained by complex terrain. In the WUI, steep terrain facilitates the upslope spread of fire, and the continuous vegetation coverage in such habitats allows ignition sources to develop into fire events of a certain scale more easily under equivalent conditions (Rothermel, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e1972\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe burned-area percentage models further clarified the roles of individual controls. Dry-season relative humidity was the most important limiting factor in both interfaces, indicating a universal atmospheric-moisture constraint on burned-area percentage. Yet the mechanisms diverged beyond this commonality. In the WCI, climatic factors remained overwhelmingly dominant. This likely reflects the fact that WCI fuels consist mainly of crop residues and herbaceous vegetation, which have large surface areas, are highly sensitive to drying, and lose moisture rapidly. Moreover, the open farmland environment lacks forest-canopy shelter, leaving surface fuels directly exposed to the atmosphere and allowing them to become flammable quickly as conditions dry (Nelson, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2001\u003c/span\u003e). By comparison, the WUI region is significantly enhanced by the dual effects of vegetation properties and human activities. When the proportion of coniferous forest exceeds 20%, the burned-area percentage in the WUI rises sharply, indicating that resinous and spatially continuous coniferous forests are the foundation supporting large-scale forest fire spread in WUI regions (Su et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Chen et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). At low population densities, burned-area percentage varied only modestly, but once population density exceeded 300 persons/km\u003csup\u003e2\u003c/sup\u003e, burned-area percentage rose rapidly. At low densities, roads and buildings fragment continuous vegetation and provide some physical barriers to fire spread. Once density reaches a higher threshold, however, more frequent human disturbance and ignition sources overwhelm these early barriers, making it easier for locally continuous and intense burning environments to develop (Syphard et al., \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). This also highlights the risk of large-scale fires in high population density WUI areas, which should be regarded as core zones for future forest fire prevention.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Implications for adaptive fire management\u003c/h2\u003e \u003cp\u003eThe contrasting drivers identified for the WCI and WUI imply that wildfire governance should shift from uniform prevention strategies toward differentiated, interface-specific management. In the WCI, prevention should focus on regulating agricultural fire use along forest-cropland boundaries, especially during busy farming periods when anthropogenic ignitions are most likely. In the WUI, priority should be given to fuel management, including prescribed burning, fuel reduction, and maintenance of defensible buffer zones, to reduce the probability of fire spread into populated areas. Across administrative levels, prefecture-scale planning should integrate WUI fire risk into land-use planning and disaster prevention systems, whereas town level management should emphasize ignition-source control and rapid early response during peak burning seasons.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e4.5 Limitations and Prospect\u003c/h2\u003e \u003cp\u003eThis study clarifies the spatiotemporal contrasts and driving mechanisms of wildfires in WCI and WUI landscapes, but several limitations should be acknowledged. First, because of data constraints, we did not incorporate wind speed or real time meteorological conditions, both of which are critical for explaining extreme fire behavior (Werth et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). Second, our monthly-scale analysis cannot fully capture the fine-scale triggering processes of individual fire events. Future work should integrate higher-resolution satellite observations with fire behavior models to resolve these mechanisms more explicitly and improve prediction and management in heterogeneous interface landscapes.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThis study compared wildfire dynamics between the WCI and WUI on the Central Yunnan Plateau and examined their spatiotemporal patterns and driving mechanisms under a common framework of climate, topography, fuel, and human activity. The results show that 2010 was a major turning point. Before 2010, burned area generally increased in both interfaces. After 2010, burned area declined markedly in the WCI, while wildfire risk shifted increasingly toward the WUI. County- and town-level areas were the main locations of fire occurrence. Spatially, WCI fires were concentrated in Honghe Prefecture, whereas WUI fires clustered around expanding urban fringes. Winter and spring remained the main fire seasons, but summer fire activity increased over time, suggesting that wildfire risk is extending into non-traditional seasons. Coniferous forests and shrublands were the main fire-carrying vegetation types. In terms of mechanisms, temperature was the main control on fire occurrence, while dry-season relative humidity was the dominant control on burned-area percentage. Beyond this shared climatic control, WCI fire occurrence was more closely associated with population density, whereas WUI fire occurrence was more constrained by topographic relief. Burned-area percentage in the WCI remained mainly climate-driven, while burned-area percentage in the WUI was increasingly shaped by coniferous forest cover and population density. These findings indicate that WCI and WUI fires represent different interface fire regimes rather than variants of a single process. Future fire prevention should therefore adopt differentiated management for WCI and WUI landscapes, with stronger control of agricultural fire use in the WCI and greater attention to fuel management and population exposure in high-density WUI areas along expanding urban fringes.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eCompeting interests\u003c/h2\u003e \u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eYang Lin: Writing-review \u0026amp; editing, Writing-original draft, Visualization, Methodology, Data curation. Mingjian Xiahou: Writing-review \u0026amp; editing, Data curation. Jiesheng Rao: Writing-review \u0026amp; editing. Caifang Luo: Writing-review \u0026amp; editing, Methodology. Tao Yang: Writing-review \u0026amp; editing. Zehao Shen: Writing-review and editing, Methodology, Funding acquisition, Conceptualization\u003c/p\u003e\u003ch2\u003eAcknowledgments\u003c/h2\u003e \u003cp\u003eThis study is supported by the Yunnan Fundamental Research Projects (202302A0370016)\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData will be made available on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbatzoglou JT, Williams AP (2016) Impact of anthropogenic climate change on wildfire across western US forests. Proceedings of the National Academy of Sciences 113(42), 11770\u0026ndash;11775\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndela N, Morton DC, Giglio L, Chen Y, van der Werf GR, Kasibhatla PS, DeFries RS, Collatz GJ, Hantson S, Kloster S, Bachelet D, Forrest M, Lasslop G, Li F, Mangeon S, Melton JR, Yue C, Randerson JT (2017) A human-driven decline in global burned area. Science 356(6345):1356\u0026ndash;1362\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArchibald S, Lehmann CER, Belcher CM, Bond WJ, Bradstock RA, Daniau AL, Dexter KG, Forrestel EJ, Greve M, He T, Higgins SI, Hoffmann WA, Lamont BB, McGlinn DJ, Moncrieff GR, Osborne CP, Pausas JG, Price O, Ripley BS, Rogers BM, Schwilk DW, Simon MF, Turetsky MR, Van der Werf GR, Zanne AE (2018) Biological and geophysical feedbacks with fire in the Earth system. Environ Res Lett 13(3):033003\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArchibald S, Staver AC, Levin SA (2012) Evolution of human-driven fire regimes in Africa. Proceedings of the National Academy of Sciences 109(3), 847\u0026ndash;852\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBadia A, Serra P, Modugno S (2011) Identifying dynamics of fire ignition probabilities in two representative Mediterranean wildland-urban interface areas. Appl Geogr 31(3):930\u0026ndash;940\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBowman DM, Balch JK, Artaxo P, Bond WJ, Carlson JM, Cochrane MA, D'Antonio CM, Defries RS, Doyle JC, Harrison SP, Johnston FH, Keeley JE, Krawchuk MA, Kull CA, Marston JB, Moritz MA, Prentice IC, Roos CI, Scott AC, Swetnam TW, van der Werf GR, Pyne SJ (2009) Fire in the Earth system. Science 324(5926):481\u0026ndash;484\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBreiman L (2001) Random forests. Mach Learn 45(1):5\u0026ndash;32\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBroadbent EN, Asner GP, Keller M, Knapp DE, Oliveira PJC, Silva JN (2008) Forest fragmentation and edge effects from deforestation and selective logging in the Brazilian Amazon. Biol Conserv 141(7):1745\u0026ndash;1757\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarlson AR, Hawbaker TJ, Mockrin MH, Radeloff VC, Bair LS, Caggiano MD, Meldrum JR, Alexandre PM, Kramer HA, Steblein PF (2025) Rising rates of wildfire building destruction in the conterminous United States. Proceedings of the National Academy of Sciences 122(51), e2505886122\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarlson AR, Helmers DP, Hawbaker TJ, Mockrin MH, Radeloff VC (2022) The wildland\u0026ndash;urban interface in the United States based on 125 million building locations. Ecol Appl 32(5), e2597\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen B, Wu S, Jin Y, Song Y, Wu C, Venevsky S, Xu B, Webster C, Gong P (2024) Wildfire risk for global wildland\u0026ndash;urban interface areas. Nat Sustain 7(4):474\u0026ndash;484\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen F, Si L, Zhao F, Wang M (2023) Volatile oil in pinus yunnanensis potentially contributes to extreme fire behavior. Fire 6(3):113\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCochrane MA (2001) Synergistic interactions between habitat fragmentation and fire in evergreen tropical forests. Conserv Biol 15(6):1515\u0026ndash;1521\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCunningham CX, Williamson GJ, Bowman DMJS (2024) Increasing frequency and intensity of the most extreme wildfires on Earth. Nat Ecol Evol 8(8):1420\u0026ndash;1425\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeng H, Li D, Cai S, Zhao F, Zhao J (2025) Spatio-temporal dynamics of forest fire occurrence in Yunnan, China from 2001 to 2021 based on MODIS. npj Nat Hazards 2(1):52\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGelabert PJ, Jim\u0026eacute;nez-Ruano A, Ochoa C, Alcasena F, Sj\u0026ouml;str\u0026ouml;m J, Marrs C, Ribeiro LM, Palaiologou P, Bentu\u0026eacute; Mart\u0026iacute;nez C, Chuvieco E, Vega-Garc\u0026iacute;a C, Rodrigues M (2025) Assessing human-caused wildfire ignition likelihood across Europe. Nat Hazards Earth Syst Sci 25(11):4713\u0026ndash;4729\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGetis A, Ord JK (1992) The analysis of spatial association by use of distance statistics. Geographical Anal 24(3):189\u0026ndash;206\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGong D (2025) Higher Emissions Scenarios Increase Wildland\u0026ndash;Urban Interface Fire Hazard in China. Sustainability 17(16):7409\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGong D, Sun L, Hu T (2024) Characterizing the occurrence of wildland-urban interface fires and their important factors in China. Ecol Ind 165:112179\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo F, Innes JL, Wang G, Ma X, Sun L, Hu H, Su Z (2015) Historic distribution and driving factors of human-caused fires in the Chinese boreal forest between 1972 and 2005. J Plant Ecol 8(5):480\u0026ndash;490\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo Y, Wang J, Ge Y, Zhou C (2024) Global expansion of wildland-urban interface intensifies human exposure to wildfire risk in the 21st century. Sci Adv 10(45):eado9587\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaight RG, Cleland DT, Hammer RB, Radeloff VC, Rupp TS (2004) Assessing fire risk in the wildland-urban interface. J Forest 102(7):41\u0026ndash;48\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHamrouni Y, Paillassa E, Ch\u0026eacute;ret V, Monteil C, Sheeren D (2021) From local to global: A transfer learning-based approach for mapping poplar plantations at national scale using Sentinel-2. ISPRS J Photogrammetry Remote Sens 171:76\u0026ndash;100\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan J, Shen Z, Ying L, Li G, Chen A (2015) Early post-fire regeneration of a fire-prone subtropical mixed Yunnan pine forest in Southwest China: Effects of pre-fire vegetation, fire severity and topographic factors. For Ecol Manag 356:31\u0026ndash;40\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHe T, Lamont BB, Pausas JG (2019) Fire as a key driver of Earth's biodiversity. Biol Rev 94(6):1983\u0026ndash;2010\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHutchinson MF (1995) Interpolating mean rainfall using thin plate smoothing splines. Int J Geographical Inform Syst 9(4):385\u0026ndash;403\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJolly WM, Cochrane MA, Freeborn PH, Holden ZA, Brown TJ, Williamson GJ, Bowman DMJS (2015) Climate-induced variations in global wildfire danger from 1979 to 2013. Nat Commun 6(1):7537\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi C, Lin S, Wang Y, Chen L (2024a) Evaluation of regional integration in urban agglomeration area from the perspective of urban symbiosis. Sustainable Cities Soc 105:105338\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi Q, Li L, Zhang J, He X (2024b) Spatio-temporal characteristics and driving mechanisms of urban expansion in the Central Yunnan Urban Agglomeration. Land 13(9):1496\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi X, Li Y, Chen A, Gao M, Slette IJ, Piao S (2019) The impact of the 2009/2010 drought on vegetation growth and terrestrial carbon balance in Southwest China. Agric For Meteorol 269\u0026ndash;270:239\u0026ndash;248\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi Y, Pan Z, Gao C, Shi C, Lin S, Huang X (2025) Decoding wildland-urban interface fire regimes in Hong Kong, vol 113. Urban Forestry \u0026amp; Urban Greening, p 129095\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLian C, Xiao C, Feng Z, Ma Q (2024) Accelerating decline of wildfires in China in the 21st century. Front Forests Global Change 6:1252587\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin H-W, McCarty JL, Wang D, Rogers BM, Morton DC, Collatz GJ, Jin Y, Randerson JT (2014) Management and climate contributions to satellite-derived active fire trends in the contiguous United States. J Geophys Research: Biogeosciences 119(4):645\u0026ndash;660\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Q, Shan Y, Shu L, Sun P, Du S (2018) Spatial and temporal distribution of forest fire frequency and forest area burnt in Jilin Province, Northeast China. J Forestry Res 29(5):1233\u0026ndash;1239\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Y, Xu S, Xiao R, Zhao Y (2025) Increasing fragmentation of the wildland\u0026ndash;urban interface as it expands across China\u0026rsquo;s dryland landscape from 1990 to 2020. Landscape Ecol 40(12):231\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLv Q, Yang Z, Chen Z, Li M, Gao B, Yang J, Chen X, Xu B (2024) Crop residue burning in China (2019\u0026ndash;2021): Spatiotemporal patterns, environmental impact, and emission dynamics. Environ Sci Ecotechnology 21:100394\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMatthews S (2014) Dead fuel moisture research: 1991\u0026ndash;2012. Int J Wildland Fire 23(1):78\u0026ndash;92\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoreira F, Ascoli D, Safford H, Adams MA, Moreno JM, Pereira JM, Catry FX, Armesto J, Bond W, Gonz\u0026aacute;lez ME (2020) Wildfire management in Mediterranean-type regions: paradigm change needed. Environ Res Lett 15(1):011001\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoreno M, Bertol\u0026iacute;n C, Arlanz\u0026oacute;n D, Ortiz P, Ortiz R (2023) Climate change, large fires, and cultural landscapes in the mediterranean basin: An analysis in southern Spain. Heliyon 9(6), e16941\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMu H, Li X, Wen Y, Huang J, Du P, Su W, Miao S, Geng M (2022) A global record of annual terrestrial Human Footprint dataset from 2000 to 2018. Sci Data 9(1):176\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNelson RM (2001) Water Relations of Forest Fuels. In: Johnson EA, Miyanishi K (eds) Forest Fires: Behavior and Ecological Effects. Academic, San Diego, CA, pp 79\u0026ndash;149\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOrd JK, Getis A (1995) Local spatial autocorrelation statistics: distributional issues and an application. Geographical Anal 27(4):286\u0026ndash;306\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePan SJ, Yang Q (2010) A survey on transfer learning. IEEE Trans Knowl Data Eng 22(10):1345\u0026ndash;1359\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePausas JG, Ribeiro E (2013) The global fire\u0026ndash;productivity relationship. Glob Ecol Biogeogr 22(6):728\u0026ndash;736\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQi X, Wei M, Wang Z, Jiang T, Wang P, Teng M, Yan Z (2022) Effects of the COVID-19 lockdown in Hubei, China: cessation of incense burning reduces regional landscape fire. Ecol Soc 27(3):5\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQin H, Tan Y, Shen T, Schaefer DA, Chen H, Zhou S, Xu Q, Zhu Y, Cheng J, Zhao G, Xu J (2024) Spatiotemporal trends of atmospheric dryness during 1980\u0026ndash;2021 in Yunnan, China. Front Forests Global Change 7:1397028\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCore Team R (2025) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRadeloff VC, Helmers DP, Kramer HA, Mockrin MH, Alexandre PM, Bar-Massada A, Butsic V, Hawbaker TJ, Martinuzzi S, Syphard AD, Stewart SI (2018) Rapid growth of the US wildland-urban interface raises wildfire risk. Proceedings of the National Academy of Sciences 115(13), 3314\u0026ndash;3319\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRadeloff VC, Mockrin MH, Helmers D, Carlson A, Hawbaker TJ, Martinuzzi S, Schug F, Alexandre PM, Kramer HA, Pidgeon AM (2023) Rising wildfire risk to houses in the United States, especially in grasslands and shrublands. Science 382(6671):702\u0026ndash;707\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRibeiro AFS, Santos L, Randerson JT, Uribe MR, Alencar AAC, Macedo MN, Morton DC, Zscheischler J, Silvestrini RA, Rattis L, Seneviratne SI, Brando PM (2024) The time since land-use transition drives changes in fire activity in the Amazon-Cerrado region. Commun Earth Environ 5(1):96\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRothermel RC (1972) A mathematical model for predicting fire spread in wildland fuels. U.S. Department of Agriculture, Intermountain Forest and Range Experiment Station, Ogden, UT, p 40\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchug F, Bar-Massada A, Carlson AR, Cox H, Hawbaker TJ, Helmers D, Hostert P, Kaim D, Kasraee NK, Martinuzzi S, Mockrin MH, Pfoch KA, Radeloff VC (2023) The global wildland\u0026ndash;urban interface. Nature 621(7977):94\u0026ndash;99\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSegura-Garcia C, Alencar A, Arruda S, Bauman VL, Silva D, Conciani W, Oliveras Menor DE (2025) I., The fire regimes of the Cerrado and their changes through time. Philosophical Transactions of the Royal Society B: Biological Sciences 380(1924), 20230460\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSenande-Rivera M, Insua-Costa D, Miguez-Macho G (2022) Spatial and temporal expansion of global wildland fire activity in response to climate change. Nat Commun 13(1):1208\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShen Z, Giljohann K, Liu Z, Pausas J, Rogers B (2025) Novel wildfire regimes under climate change and human activity: patterns, driving mechanisms and ecological impacts. Philosophical Transactions of the Royal Society B: Biological Sciences 380(1924), 20230446\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShi Z, Jiang D, Wang Y (2024) Spatiotemporal dependence of compound drought\u0026ndash;heatwave and fire activity in China. Weather Clim Extremes 45:100695\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSu W-H, Shi Z, Zhou R, Zhao Y-J, Zhang G-F (2015) The role of fire in the Central Yunnan Plateau ecosystem, southwestern China. For Ecol Manag 356:22\u0026ndash;30\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSyphard AD, Radeloff VC, Keeley JE, Hawbaker TJ, Clayton MK, Stewart SI, Hammer RB (2007) Human influence on California fire regimes. Ecol Appl 17(5):1388\u0026ndash;1402\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTuia D, Persello C, Bruzzone L (2016) Domain Adaptation for the Classification of Remote Sensing Data: An Overview of Recent Advances. IEEE Geoscience Remote Sens Magazine 4(2):41\u0026ndash;57\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVadrevu KP, Lasko K, Giglio L, Schroeder W, Biswas S, Justice C (2019) Trends in Vegetation fires in South and Southeast Asian Countries. Sci Rep 9(1):7422\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang C, Wu D, Shen Z, Peng M, Ou X (2021) How do physical and social factors affect urban landscape patterns in intermountain basins in Southwest China? Landscape Ecol 36(7):1893\u0026ndash;1911\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang H, Jin B, Zhang K, Aktar S, Song Z (2022) Effectiveness in mitigating forest fire ignition sources: A statistical study based on fire occurrence data in China. Fire 5(6):215\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang W, Zhao F, Wang Y, Huang X, Ye J (2023) Seasonal differences in the spatial patterns of wildfire drivers and susceptibility in the southwest mountains of China. Sci Total Environ 869:161782\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWerth PA, Potter BE, Clements CB, Finney MA, Goodrick SL, Alexander ME, Cruz MG, Forthofer JA, McAllister SS (2011) Synthesis of knowledge of extreme fire behavior: volume I for fire managers. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWright MN, Ziegler A (2017) ranger: A Fast Implementation of Random Forests for High Dimensional Data in C\u0026thinsp;+\u0026thinsp;+\u0026thinsp;and R. J Stat Softw 77(1):1\u0026ndash;17\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu C, Xu W, Xiao C, Wang Q, Yuan H, Dong J, Huang S, Xiong Y (2021) Dynamic change of recovery ratios and influencing factors of typical post-fire burn areas in central Yunnan Province. Resour Sci 43(12):2465\u0026ndash;2474 [in Chinese]\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiahou M, Liu Y, Yang T, Shen Z (2024) Estimating potential vegetation distribution and restoration in a biodiversity hotspot region under future climate change. J Geog Sci 34(11):2128\u0026ndash;2144\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiahou M, Peng M, Shen Z, Wen Q, Wang C, Liu Y, Zhang Q, Peng L, Yu C, Ou X, Fang J (2025) Vegetation mapping of Yunnan Province by integrating remote sensing, field observations, and models. Sci China Earth Sci 68(3):836\u0026ndash;849\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu C, You C (2023) Agricultural expansion dominates rapid increases in cropland fires in Asia. Environ Int 179:108189\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu W, Liu Y, Veraverbeke S, Wu W, Dong Y, Lu W (2021) Active Fire Dynamics in the Amazon: New Perspectives From High-Resolution Satellite Observations. Geophys Res Lett 48(20), e2021GL093789.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang J, Huang X (2021) Earth Syst Sci Data 13(8):3907\u0026ndash;3925The 30 m annual land cover dataset and its dynamics in China from 1990 to 2019\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYe J, Wu M, Deng Z, Xu S, Zhou R, Clarke KC (2017) Modeling the spatial patterns of human wildfire ignition in Yunnan province, China. Appl Geogr 89:150\u0026ndash;162\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYin J, Ding W (2015) Analysis and application of \u0026lsquo;\u0026lsquo;Degree\u0026rsquo;\u0026rsquo;in the mountain town process in Yunnan province. Geographical Sci Res 4(3):144\u0026ndash;150 [in Chinese]\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYing L, Cheng H, Shen Z, Guan P, Luo C, Peng X (2021) Relative humidity and agricultural activities dominate wildfire ignitions in Yunnan, Southwest China: Patterns, thresholds, and implications. Agric For Meteorol 307:108540\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYoung AM, Higuera PE, Duffy PA, Hu FS (2017) Climatic thresholds shape northern high-latitude fire regimes and imply vulnerability to future climate change. Ecography 40(5):606\u0026ndash;617\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZamanialaei M, San Martin D, Theodori M, Purnomo DMJ, Tohidi A, Lautenberger C, Qin Y, Trouv\u0026eacute; A, Gollner M (2025) Fire risk to structures in California\u0026rsquo;s Wildland-Urban Interface. Nat Commun 16(1):8041\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang H, Wang W, Ban Q (2025) Seasonal forest fire risk and key drivers in Yunnan Province: a machine learning approach. npj Nat Hazards 2(1):59\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang W, Yan X, Wang Q, Long T, Li X, Pu J, Ding Z (2022) Effects of prescribed burning on fire behavior of surface fuel in Pinus yunnanensis forest land. J Beijing Forestry Univ 44(5):69\u0026ndash;76 [in Chinese]\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao H, Zhang Z, Zhen S, Wang X, Yin Y (2025a) Multifactorial interactions contribute to contrasting wildfire trends at mid\u0026ndash;high latitudes of the Northern Hemisphere. Agric For Meteorol 367:110507\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao J, Zheng B, Ciais P, Chen Y, Gasser T, Canadell JG, Zhang L, Zhang Q (2025b) Global warming amplifies wildfire health burden and reshapes inequality. Nature 647(8091):928\u0026ndash;934\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhu Z, Deng X, Zhao F, Li S, Wang L (2022) How environmental factors affect forest fire occurrence in Yunnan forest region. Forests 13(9):1392\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Wildland-urban interface, Wildland-cropland interface, Fire occurrence, Spatiotemporal dynamics, Driving mechanisms, Management Insights","lastPublishedDoi":"10.21203/rs.3.rs-9559449/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9559449/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eAs human activities continue to extend into wildland areas, wildfire risk in wildland-urban interfaces (WUI) has drawn considerable attention. By contrast, the wildland-cropland interface (WCI) has been less examined, and few studies have compared the spatiotemporal patterns and driving forces of wildfire between these two types of interfaces. Based on Landsat data from 1990 to 2023, we compared fire activity and its drivers in WCI and WUI landscapes on the Central Yunnan Plateau (CYP), China, at multiple spatial scales.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe results indicate a clear turning point around 2010. Before 2010, burned area increased in both interfaces. After 2010, both burned area in the WCI and its share of total burned area declined markedly, whereas the share of the WUI remained stable, after 2010. Concurrently, the fire regime shifted from winter-spring dominance toward greater summer extension, with spring burned area in WUI increasing after 2000. Meteorological factors dominated fire variability in both interfaces: temperature primarily controlled fire occurrence, whereas relative humidity governed burned-area percentage. Beyond this shared climatic background, fire occurrence in the WCI was more strongly associated with population density, whereas fire occurrence in the WUI was more constrained by topographic relief. Burned-area percentage in the WCI exhibited a climate-dominated pattern, whereas that in the WUI was jointly influenced by coniferous forest proportion and population density.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eBy comparing WCI and WUI fires over more than three decades, this study shows that the two interfaces have distinct fire regimes and require differentiated management. After 2010, the regional focus of wildfire activity shifted increasingly from WCI landscapes toward the WUI, where fuel continuity and population exposure jointly shaped burned-area percentage. Fire management should therefore place greater emphasis on agricultural fire regulation in the WCI and fuel management in the WUI.\u003c/p\u003e","manuscriptTitle":"The temporal trajectories and driving forces of wildfire regimes differing between wildland-cropland and wildland-urban interfaces on the Central Yunnan Plateau","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-08 12:04:06","doi":"10.21203/rs.3.rs-9559449/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b6c70164-0cf5-44be-b3e3-828d605342d4","owner":[],"postedDate":"May 8th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"225223074118384330054043032339639485277","date":"2026-05-09T16:51:31+00:00","index":17,"fulltext":""},{"type":"reviewersInvited","content":"12","date":"2026-04-30T09:47:02+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-29T09:36:32+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-29T09:36:28+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-08T12:04:06+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-08 12:04:06","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9559449","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9559449","identity":"rs-9559449","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}