Urban–Agricultural–Ecological Interactions and Land Surface Temperature—A Spatiotemporal Study of the Middle Yangtze River Region

Abstract

The land use dynamics of urban, agricultural, and ecological spaces are critical factors influencing land surface temperature (LST); however, the existing methods for describing the spatial carriers of land surface temperature evolution face issues such as granularity effects and projection sensitivity, which hinder effective comparisons across different regions and categories, thus limiting the progress of current research. This study introduces a quadtree-based spatial framework to achieve unified measurements of scale and fragmentation across Urban–Agricultural–Ecological spaces, with an empirical analysis of the Middle Yangtze River Region. Results show that between 2000 and 2020, urban and agricultural spaces expanded while ecological spaces declined, with all three types becoming increasingly fragmented. Urban agglomeration and expansion significantly elevated LST; agricultural spaces exerted relatively limited effects; and ecological fragmentation generated localized cooling but weakened core regulatory functions, ultimately leading to warming within ecological spaces themselves. This study proposes a robust method for spatial identification and fragmentation quantification, revealing the dual role of scale and morphology in regulating regional thermal environments and underscoring the importance of balanced Urban–Agricultural–Ecological configurations for climate-adaptive land use planning.

1. Introduction

Land surface temperature (LST) variation is one of the central topics in global climate change research and serves as a critical indicator of major environmental issues such as global warming and the greenhouse effect [1,2]. LST is influenced by the evolution of land cover, which is shaped by human land use activities, including urbanization, agricultural production, and ecological conservation [2,3,4]. In the context of climate change mitigation, controlling the scale of urban and agricultural land while optimizing the spatial structure of different land types has become a key shared objective in global land use planning and management [5,6].

Under China’s national territorial spatial planning framework, urban, agricultural, and ecological spaces are defined as mutually exclusive functional categories that together exhaust the national territory [7]. We adopt this classification to align our analysis with the planning system of the study area and to enhance the planning relevance of our findings. In this framework, urban and agricultural spaces are dominated by human activities, whereas ecological space primarily fulfills ecological functions and is governed by non-human processes. This human-versus-non-human delineation provides a useful basis for examining the spatial heterogeneity of LST evolution, allowing us to separate LST changes driven mainly by human factors from those controlled predominantly by biophysical processes. It also clarifies why the three spaces do not overlap and how their functional attributes correspond to distinguishable LST signatures [8,9].

Although land surface temperature is primarily governed by climatic conditions such as radiation, precipitation, and wind, human activities also play a significant role. Consequently, the organization and utilization of land that supports these activities are important contributing factors to land surface temperature [10,11]. Previous studies have shown that urban areas, characterized by heat emissions from transportation, industry, and residential activities, along with the accumulation of impervious surfaces, contribute substantially to surface warming. The dense network of roads, buildings, and other human-made structures in cities often exacerbates the heat island effect, making urban areas significantly warmer than surrounding rural areas. Studies have also highlighted that the intensity of LST variation in urban settings is not uniform but varies depending on factors such as the scale of urbanization, land use distribution, and the spatial configuration of the built environment. Furthermore, urban morphology, such as the presence of parks or green spaces, can mediate some of the warming effects by providing shade and enhancing evapotranspiration processes [12,13]. By contrast, ecological spaces dominated by natural vegetation and water bodies reduce local land surface temperatures through evapotranspiration and serve as buffers and stabilizers in the thermal dynamics of both the areas themselves and their surrounding environment. These cooling effects are especially pronounced in large, contiguous natural reserves, where a high degree of spatial continuity ensures maximum cooling potential. However, fragmentation of these ecological areas reduces their ability to regulate temperature effectively [2,3,4]. Agricultural spaces exhibit more complex thermal effects: their seasonal nature leads to periodic temperature fluctuations, with distinct temperature patterns arising from crop type, cultivation practices, and seasonal variations. While natural vegetation in agricultural lands can lower temperatures, modern agricultural practices—such as the use of machinery, irrigation systems, and fertilizers—can also contribute to local heat emissions. Moreover, the spatial configuration of agricultural areas, such as crop rotation and field arrangements, influences LST dynamics [14,15].

Furthermore, research has revealed that the spatial morphology of these three functional types is an important factor in LST dynamics [16,17]. A dispersed and fragmented urban layout can alleviate the heat island effect, while high-density agglomeration may intensify local warming. Conversely, excessive fragmentation of ecological spaces weakens their overall cooling capacity, and large-scale monoculture agriculture may lead to heightened temperature variability. Therefore, a comprehensive understanding of LST variation requires considering the scale, spatial distribution, and morphology of these functional spaces, highlighting the interactions between them as key drivers of regional climate dynamics [18,19].

Nevertheless, current research remains predominantly focused on urban areas and their heat island effects, with insufficient attention paid to the thermal roles of agricultural and ecological spaces. Land use functions are mutually exclusive; the expansion of one functional space inevitably comes at the expense of others, and changes in the spatial configuration (e.g., agglomeration or fragmentation) of one type often trigger corresponding adjustments in the others. This interplay results in complex, multi-functional spatial responses in LST variation. However, the synergistic effects of urban, agricultural, and ecological spaces on LST—across both scale and morphology—remain underexplored [17,20]. The majority of LST research has been conducted at the urban scale, which limits its relevance for understanding thermal dynamics in non-urban areas, especially given that urban land represents less than 1% of Earth’s land surface. For instance, while fragmenting impervious surfaces to disrupt heat islands is a widely accepted cooling strategy in cities, this approach may inadvertently disrupt the pre-existing “cold island” patterns in ecological spaces when viewed from a holistic perspective of functional spatial synergy [17,20]. Therefore, spatial optimization strategies derived solely from an urban perspective require further validation regarding their effectiveness in regulating LST at regional or even global scales.

We argue that existing studies suffer from two main limitations. First, most previous research divides the surface space based on zoning logic dominated by land use types. However, the spatial activities carried by the land have a more direct impact on surface temperature and climate. Therefore, more attention should be given to zoning logic based on spatial functions. Yet, defining mutually exclusive urban–agricultural–ecological spaces across scales while maintaining consistency and preserving intra-class heterogeneity remains a challenge [21,22]. Second, some studies employ object-oriented and morphological segmentation methods. For instance, morphological spatial pattern analysis (MSPA) identifies spatial elements such as core areas, edges, and connectivity corridors based on the morphological characteristics of land features. Other studies adopt climate-oriented classification systems, such as the local climate zone (LCZ) framework, which classifies areas according to underlying surface characteristics and thermal environmental attributes [23,24]. These different approaches provide rich technical reserves and diversified perspectives for identifying and structuring urban spaces. However, when applied across different spatial scales, these methods are often affected by scale effects, boundary effects, and information loss, which limit comprehensive evaluation of spatial effects [25,26]. Moreover, morphological metrics such as landscape pattern indices can effectively characterize spatial morphology within a given spatiotemporal range, but their values are often sensitive to pixel size, projection, parameter settings, and category definitions. Consequently, they are not well-suited for cross-scale comparisons at the urban–agricultural–ecological spaces level—comparisons that are crucial for systematically understanding the spatial mechanisms influencing LST [27,28,29].

To address these gaps, this study proposes a quadtree-based analytical framework that converts traditional land use data into three mutually exclusive, space-filling functional types. By quantifying fragmentation through quadtree hierarchical depth, we enable unified measurements of LST, area, and fragmentation [30,31]. Focusing on the Middle Yangtze River region (Hubei, Hunan, and Jiangxi provinces)—one of China’s hottest climate zones—we systematically evaluated the evolution trajectories of these three spatial types over a twenty-year period and the mechanisms by which they affect LST, offering methodological insights for constructing climate-friendly spatial governance models [32,33,34].

2. Materials and Methods

2.1. Study Area and Data Sources

The Middle Reaches of the Yangtze River (MRYR) primarily encompass the provinces of Hunan, Hubei, and Jiangxi, located in the transitional zone between China’s second and third topographic terraces, characterized by complex terrain and covering a total area of 564,600 km² (Figure 1). Characterized by a subtropical monsoon climate, the region experiences hot, rainy summers and mild, dry winters. The provincial capitals—Changsha, Wuhan, and Nanchang—as well as Chongqing, are collectively known as China’s “Four Furnaces” due to their extreme summer heat [35,36]. The local climate significantly impacts human comfort, and in recent years, frequent extreme summer temperatures have even threatened public health, underscoring the urgent need for thermal environment optimization and regulatory measures in MRYR.

Meanwhile, with the steady advancement of China’s strategies, such as the Rise of Central China and the Yangtze River Economic Belt, MRYR has emerged as a vital economic growth hub. Rapid land use changes have profoundly influenced regional land surface temperature (LST) dynamics [37,38]. Urban expansion has encroached upon croplands and natural environments, altering the regional energy balance, intensifying urban heat island effects, and fragmenting agricultural and ecological spaces. These changes pose new challenges to the regulatory capacity of the regional thermal environment [39,40]. Given these factors, selecting MRYR as the study area to investigate the systematic evolution of the urban–agricultural–ecological space structure and its relationship with LST is highly typical and of practical significance.

The research primarily utilizes the Global Land Cover with Fine Classification System at a 30 m resolution (GLC_FCS30) and MODIS LST data (

Table 1. Data and Descriptions.

Category	Content	Resolution	Year	Source
Land use/cover	GLC_FCS30	30 m	2000, 2010, 2020	https://zenodo.org/records/4280923 (accessed on 1 March 2024).
Land Surface Temperature	MOD11A2	1 km	2000, 2010, 2020	https://ladsweb.modaps.eosdis.nasa.gov/search/ (accessed on 1 March 2024).

). Studies have shown that the overall accuracy of the GLC_FCS30 product ranges from 82% to 90%, with the validation accuracy in China typically exceeding 88%. The MOD11A2 LST product has been calibrated and validated by multiple researchers, showing a mean bias within 1 °C, and has been widely applied in regional-scale studies of land surface thermal environments [41,42,43]. Moreover, since this study employs quadtree aggregation and spatial averaging processes, local errors are partially smoothed during aggregation, thus ensuring the stability of the overall trend analysis results. Meanwhile, this study focuses on the average thermal differences among functional spatial zones and their long-term variation trends, rather than on short-term meteorological processes. Annual averaging effectively reduces noise at daily and seasonal scales, making the results of spatial aggregation more robust [44,45].

Based on an enhanced “LULC-to-Urban–Agricultural–Ecological” reclassification scheme, the 30 m-resolution land use/land cover (LULC) maps for 2000, 2010, and 2020 were transformed into functional spatial categories. Specifically, impervious surfaces were reclassified as potential urban space, croplands as potential agricultural space, and all other land types (e.g., forests, wetlands, grasslands, water bodies) were grouped as potential ecological space. These reclassified maps provide the foundational input for the quadtree-based spatial structure analysis.

2.2. Fragmentation Assessment Method for Urban–Agricultural–Ecological Division Based on Quadtree Algorithm

Currently, research on thermal environments mainly focuses on conceptual entities such as cities and rural areas, whereas classifications based solely on land use types fail to meet this need. Although traditional methods use high-precision grid data, they struggle to distinguish features such as nurseries within farmland, urban parks, small facilities scattered across agricultural areas, and fragmented farmland within ecological spaces [46,47]. As a result, products like LUCC, while faithfully representing surface characteristics, tend to blur the functional differences among land parcels with similar surface features, thereby reducing the accuracy of functional assessments for urban, agricultural, and ecological spaces [48]. Accordingly, this study proposes an urban–agricultural–Ecological space division method integrating structural and morphological characteristics based on the quadtree algorithm (Figure 2). Through multi-scale hierarchical construction and attribute assignment, it facilitates precise measurement of the structural scale and morphological features of urban, agricultural, and ecological spaces [49,50]. Compared with conventional methods, this approach introduces an information entropy-based decision rule, in addition to single-attribute functional division, analyzing intra-spatial homogeneity to effectively identify and optimize fragmented spaces and complex boundary areas [49,50]. This ensures the rationality and scientific rigor of the division of urban–agricultural–ecological space, thereby enabling accurate monitoring of land surface temperatures across different spatial types.

2.2.1. Quadtree Algorithm-Based Urban–Agricultural–Ecological Space Base Map Construction

To establish the foundational base map for the spatial structure and morphological patterns of urban–agricultural–ecological spaces, this study designs the following spatial identification process and implements it using Python (v.3.13):

(1) The land use data of the study area is reclassified into three categories—Urban (U), Agricultural (A), and Ecological (E) spaces—to form the initial base map of the urban–agricultural–ecological spaces, which serves as the input data for the quadtree algorithm. The D value is defined as the spatial fragmentation degree, representing the number of times a grid has been subdivided. A higher division depth corresponds to smaller grid areas and a greater degree of spatial fragmentation. In the initial stage, the study area is divided into square grids with a side length (l) = 16,000 m (D_l = 0). The grids are further subdivided at scales of ‘l’ = 8000 m, 4000 m, 2000 m, 1000 m, and 500 m, corresponding to D values of 1, 2, 3, 4, and 5, respectively. The calculation formula is as follows:

$$\begin{array}{c} {\mathrm{D}}_{\mathrm{l}} = {\log }_2 \left({\displaystyle \frac{1}{500}}\right) + 1\end{array}$$

(1)

(2) To determine the dominant spatial attribute of each grid, this study establishes a dual-criteria rule B (Equation (2)) based on area proportion and information entropy (Equation (3)): When the area proportion of any urban–agricultural–ecological space category within a grid reaches or exceeds 75% of the total grid area, condition B is satisfied. The calculation formula is as follows:

$$\begin{array}{c} \mathrm{B} = \mathrm{b}\mathrm{o}\mathrm{o}\mathrm{l} ({\displaystyle \frac{\max \left({\mathrm{C}}_{\mathrm{u}}, { \mathrm{C}}_{\mathrm{a}},{ \mathrm{C}}_{\mathrm{e}}\right) }{{\mathrm{l}}^2}}\ge 0.75 \wedge \mathrm{H} < 1)\end{array}$$

(2)

$$\mathrm{H}=-\sum _{\mathrm{i}\in \{\mathrm{U}, \mathrm{A},\mathrm{E}\}}{\mathrm{p}}_{\mathrm{i}}{\log }_2\left({\mathrm{p}}_{\mathrm{i}}\right), {\mathrm{p}}_{\mathrm{i}}={\mathrm{C}}_{\mathrm{i}}/{\mathrm{l}}^2$$

(3)

where

• C_u, C_a and C_e represent the areas of Urban, Agricultural, and Ecological spaces, respectively;
• H denotes information entropy;
• max ( ) is the function to determine the maximum value;
• bool ( ) is a Boolean function evaluating whether the condition is met.

If B = True, the grid is considered spatially homogeneous and dominated by a single type, and its dominant spatial type T and division depth (D_l) are assigned and stored. If B = False, further subdivision of the grid is assessed, and the process proceeds to the next step.

(3) Due to the limitation of data precision (30 m), when the grid is further divided to a side length of l = 250 m, no grid can satisfy the condition where the area proportion of any urban–agricultural–ecological spaces reaches 75% and the information entropy is greater than 1. Therefore, grids with a side length smaller than ‘l’ = 500 m are deemed indivisible and are directly assigned the corresponding urban–agricultural–ecological spaces type and fragmentation degree (D_l), which are then stored. If a grid meets the subdivision criteria, it is partitioned into four equal sub-grids, and the area proportion evaluation is repeated for the newly generated grids.

(4) The computation terminates when all grids have been evaluated and no further subdivision is possible. The final output is a comprehensive urban–agricultural–ecological spaces structure and morphological feature map that fully covers the study area.

2.2.2. Spatial Statistics of Urban–Agricultural–Ecological Spaces Fragmentation

To further quantify the morphological characteristics and their changes in the urban–agricultural–ecological spaces in MRYR, this study uses 8000 m uniform grids as basic units to calculate the fragmentation index (D_xi) of each unit at different time periods and its change value (ΔD_xi). These metrics serve as important indicators for evaluating regional land space development and conservation outcomes. The morphological index is defined as the weighted average of the area (S_i) of urban, agricultural, or ecological space within a specific administrative unit in a given year. A higher value indicates greater fragmentation of the corresponding space type in that unit. The calculation formula is as follows:

$${\mathrm{D}}_{\mathrm{x}\mathrm{i}} = \sum _{\mathrm{i}=1}^{\mathrm{n}}({\mathrm{S}}_{\mathrm{i}} \times { \mathrm{D}}_{\mathrm{l}})/\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{S}}_{\mathrm{i}}$$

(4)

Here, the fragmentation change value (ΔD_xi) represents the difference in fragmentation between year p and year q for a given county(city) or district. The calculation formula is as follows:

$$\Delta {\mathrm{D}}_{\mathrm{x}\mathrm{i}} = {\mathrm{D}}_{\mathrm{x}\mathrm{i}\left(\mathrm{q}\right)}-{\mathrm{D}}_{\mathrm{x}\mathrm{i}\left(\mathrm{p}\right)}$$

(5)

When ΔD_xi > 0, it indicates an increase in the fragmentation of a certain space type from year p to year q; conversely, it suggests a reduction in fragmentation. The absolute value of ΔD_xi reflects the magnitude of overall morphological change in the region. The proposed spatial fragmentation index effectively characterizes the morphological features of urban, agricultural, and ecological spaces in MRYR.

A high urban space fragmentation index suggests dispersed urban development, which hinders concentrated resource utilization and efficient organization of urban functions. A high agricultural space fragmentation index reflects severe farmland fragmentation, making large-scale farming conditions difficult to achieve. Similarly, a high ecological space fragmentation index indicates fragmented and isolated ecological patches, undermining the effective functioning of ecosystems. Conversely, lower fragmentation values denote more contiguous and concentrated spatial distributions, leading to higher spatial efficiency.

2.3. Characterization of Land Surface Temperature

To elucidate the spatiotemporal evolution patterns of land surface temperature (LST) and its interactions with the urban–agricultural–ecological spaces structure in MRYR, this study constructs a spatiotemporal data framework for LST based on county-level units and monitoring data. Regression analysis is employed to explore the relationship between LST variations and the evolution of the urban–agricultural–ecological spaces structure.

2.3.1. Spatial Statistical Analysis

Using decadal LST monitoring data from 2000 to 2020, this study conducts spatial statistics on LST with a uniform 8000 m grid. The weighted average LST within the urban–agricultural–ecological spaces structure is extracted to analyze its distribution characteristics and temporal trends, providing a foundation for subsequent spatiotemporal evolution analysis [51,52].

2.3.2. Spatial Econometric Regression Analysis

To systematically examine the relationship between LST changes and the evolution of the urban–agricultural–ecological spaces structure, a “global–local” analytical approach is adopted. First, the global relationship between overall LST variations and changes in the area and shape indices of the urban–agricultural–ecological spaces is analyzed to assess their collective impact on regional LST. Subsequently, the study focuses on LST changes within individual zones, analyzing their correlations with the area and shape indices of all urban–agricultural–ecological spaces to uncover localized influences and interaction mechanisms. This integrated global–local analysis aims to reveal the dominant roles and mutual feedback mechanisms of the urban–agricultural–ecological spaces structure in regional LST dynamics, offering scientific insights for optimizing territorial spatial patterns to enhance thermal environment suitability.

For spatial regression modeling, the selection between a spatial lag model (SLM) and a spatial error model (SEM) is based on the significance of spatial lag effects versus spatial error effects. If the spatial lag effect is statistically more significant, SLM is selected; otherwise, SEM is applied. If neither is significant, ordinary least squares (OLS) regression is used [53,54,55,56].

3. Results

3.1. Spatial Evolution Characteristics of the Urban–Agricultural–Ecological Spaces

Using a quadtree algorithm-based spatial identification method, this study quantitatively analyzes the evolution of urban–agricultural–ecological spaces in MRYR from 2000 to 2020 at ten-year intervals. The results reveal significant spatial transformations over the two decades. The proportional area of urban, agricultural, and ecological spaces shifted from 0.9%:32.9%:66.2% in 2000 to 2%:33.7%:64.3% in 2020. Notably, urban space exhibited the most dramatic change, more than doubling in size over the 20-year period. Agricultural space expanded steadily but with far less intensity compared to urban spatial transformation. In contrast, ecological space experienced a continuous decline (

Table 2. Area (km²) and Spatial Fragmentation of Urban–Agricultural–Ecological Spaces in MRYR.

Year	Space Area			Space Spatial Fragmentation
	Urban (U)	Agricultural (A)	Ecological (E)	Urban	Agricultural	Ecological
2000	4946.22	185,616.77	374,063.06	4.538	3.767	3.326
2010	8840.24	187,196.52	368,588.57	4.420	3.793	3.330
2020	11,103.85	190,494.25	363,027.96	4.588	3.852	3.350

).

(1) The urban spatial area increased by approximately 6157.63 km², more than doubling over the past two decades. This growth was primarily driven by the conversion of agricultural land into urban areas. The expansion was particularly pronounced in the capital cities of the three provinces and their surrounding regions, where urbanization processes have advanced rapidly. Despite the continuous increase in urban land area, changes in spatial fragmentation reflect an alternating trend of intensification and dispersion in the urbanization process of MRYR. From 2000 to 2010, urban land use became more intensive, and urban functions began to concentrate. However, between 2010 and 2020, major cities experienced outward expansion of urban areas, especially in the peripheral zones of urban agglomerations, where the pace of expansion accelerated and urban spatial patterns tended to become more fragmented (Figure 3). Overall, over the two decades, only the urban spaces of provincial capital cities exhibited concentrated development, while other urban areas predominantly displayed fragmentation in spatial structure.

(2) The agricultural spatial area expanded by 4877.48 km², showing a relatively modest increase. However, spatial transformation within this category was more intense, as agricultural land was encroached upon by urban expansion while simultaneously expanding into ecological spaces. In agriculturally intensive regions such as the Jianghan Plain, the Dongting Lake Plain, and the Poyang Lake Plain, agricultural land area actually showed a declining trend. This decline was mainly attributed to urban sprawl, which led to land use conversions, such as arable land being repurposed for construction or left fallow. In contrast, a slight increase in agricultural area was observed in the peripheral regions of the Jianghan Plain and in the upper river valley regions of the Poyang Lake area, largely due to comprehensive land consolidation projects and increased demand for economic crops. In terms of spatial morphology, agricultural land in MRYR demonstrated an overall trend toward fragmentation (Figure 4). Only a few areas, such as southern Hunan and the border regions between Jiangxi and Hubei, showed signs of concentrated agricultural development. However, these regions are located in hilly or mountainous terrains, which are generally unsuitable for large-scale, intensive agricultural development.

(3) The ecological space area decreased by 11,035.1 km², experiencing the most significant loss among the three land use types. This reduction was mainly caused by encroachment from both urban and agricultural expansion. Over the past two decades, the degree of fragmentation in ecological space has generally increased throughout MRYR (Figure 5). While spatial configurations in most regions remained relatively stable, ecological areas around cities experienced varying degrees of fragmentation, indicating growing pressure from anthropogenic land transformation.

3.2. Evolution Characteristics of Land Surface Temperature

Between 2000 and 2020, the southern part of MRYR exhibited relatively high land surface temperatures (LST), with a noticeable trend of northward expansion in high-temperature zones (Figure 6). In contrast, the northern areas experienced comparatively lower temperatures, especially in the western region of Hubei Province, where a temperature decrease was observed. Notably, the year 2010 saw a general temperature drop across the region, attributable to the cooling effects of the La Niña event. By 2020, regions with well-implemented ecological protection measures, such as the Poyang Lake wetlands, had significantly lower LSTs than the surrounding areas. Over the two-decade period, the LSTs of urban–agricultural–ecological spaces in MRYR underwent varied changes, displaying evident spatial heterogeneity. Overall, a gradual northward shift of high-temperature zones and a slow cooling trend in the western low-temperature zones were observed. Furthermore, the temperature gap between the hottest and coldest areas widened over time (

Table 3. The Overall LST (°C) and the LST of the Urban–Agricultural–Ecological Spaces in MRYR.

Year	Overall	Urban	Agricultural	Ecological
2000	22.41	23.72	23.35	22.22
2010	21.36	22.87	22.28	21.16
2020	22.33	24.06	23.38	22.03

).

(1) LSTs in urban areas showed a consistent upward trend, particularly in the core cities of the region such as Wuhan, Changsha, and Nanchang (Figure 7). The Wuhan metropolitan area and the Yangtze River Delta urban agglomeration, as key urban clusters, exhibited pronounced urban heat island effects and experienced significant LST increases. Urban expansion in these regions not only elevated local temperatures but also intensified LST variations in adjacent areas. By 2020, urban temperatures in provincial capitals such as Wuhan had reached as high as 28 °C, considerably higher than those in agricultural and ecological spaces.

(2) Changes in LST in agricultural areas were relatively minor and consistently lower than those in urban spaces (Figure 8). The cooling effects of crop and soil evaporation contributed to a more stable thermal environment. Specifically, agricultural spaces such as the Poyang Lake Plain and Dongting Lake Plain maintained relatively stable temperatures over the period. However, LST changes were more pronounced in the transitional zones between agricultural and urban regions.

(3) LSTs in ecological areas showed a general decreasing trend (Figure 9). These areas, including wetland and forest reserves in MRYR, benefited from high vegetation coverage and abundant water bodies, which contributed to effective cooling. Temperatures in these zones remained consistently lower than in urban and agricultural spaces. Nevertheless, ongoing urbanization in surrounding regions led to encroachment upon ecological spaces, diminishing their cooling capacity. In particular, regions with weaker ecological protection, such as the Mufu Mountains near the Hubei-Jiangxi border and the Luoxiao Mountains near the Hunan-Jiangxi border, experienced rising LSTs. Conversely, areas such as the Wuling Mountains on the western border of Hunan Province, where strict ecological protection and reforestation policies were enforced, saw enhanced regulation of LST and a gradual temperature decline.

3.3. Systematic Relationship Between Territorial Spatial Evolution and Land Surface Temperature Changes

To thoroughly investigate the relationship between land surface temperature (LST) variations and the evolution of the urban–agricultural–ecological space structure in the MRYR from 2000 to 2020, this study employs a spatial regression model. The cross-conversion areas (

Table 4. Cross-Conversion Areas (km²) Among the Urban–Agricultural–Ecological Spaces.

Time	U→A	U→E	A→U	A→E	E→U	E→A
2000–2020	580	83	7274	4962	2036	18,507

) and fragmentation indices of the three spatial categories serve as independent variables, while both overall and zonal LSTs are used as dependent variables (Figure 10). The aim is to quantify the correlations between spatial transformations and LST changes, thereby elucidating the driving mechanisms behind surface temperature variations.

3.3.1. Driving Mechanisms of Land Surface Temperature Changes

Given the high statistical significance of both spatial lag and spatial error effects in the spatial regression analysis, and considering that the robust LM-lag test statistic (505) exceeds the robust LM-error statistic (327), the spatial lag model was adopted as the primary econometric model.

The analysis reveals a significant negative correlation between the fragmentation of urban space and LST (Figure 11), indicating that increased fragmentation of urban land may enhance surface ventilation or reduce heat accumulation in contiguous built-up areas, thereby mitigating local warming (

Table 5. Results of Spatial Lag Analysis (Independent Variable: LST).

	Spatial Fragmentation			Cross-Conversion Areas(km2)
	U	A	E	U→A	U→E	A→U	A→E	E→U	E→A
Coefficient	−0.20	0.02	−0.06	2.46	−6.74	4.06	−3.87	3.49	1.52
p-value	0.00	0.37	0.05	0.01	0.06	0.00	0.00	0.00	0.00

). This finding aligns with conclusions from urban climatology studies, which suggest that moderate decentralization of urban spatial layouts can help improve thermal conditions. Similarly, ecological space fragmentation exhibits a cooling effect, although weaker. This suggests that while fragmentation reduces the core area’s climate regulation capacity, more active energy exchange between fragmented ecological edges and urban built-up areas may create localized cold islands. In contrast, the impact of agricultural space fragmentation did not pass the significance test, implying that agricultural land maintains relatively uniform thermal surface characteristics. In particular, in this region, where paddy fields and irrigated drylands dominate, fragmentation does not markedly alter fundamental thermal exchange processes.

Regression results for land use conversions further reveal systematic human-induced impacts on LST. Notably, mutual conversions between urban and agricultural spaces significantly intensify surface warming. However, the warming coefficient for agricultural-to-urban transitions is approximately 1.6 times that of urban-to-agricultural transitions. This asymmetry underscores the more severe thermal consequences of urbanization. Particularly, the conversion of ecological land to urban land demonstrates a significantly greater warming effect than conversion to agricultural land, highlighting the severe threat that urban encroachment poses to regional thermal environments. Conversely, the conversion of agricultural land into ecological space results in a significant cooling effect. The direct conversion of urban land to ecological land did not yield statistically significant regression coefficients, potentially due to insufficient sample size during the study period.

3.3.2. Driving Mechanisms of LST Changes in the Urban–Agricultural–Ecological Spaces

(1)

Spatial Morphology

The degree of fragmentation in urban and ecological spaces is significantly negatively correlated with LST, indicating that a moderate level of spatial dispersion helps mitigate the urban heat island effect (

Table 6. Results of Spatial Lag Analysis.

		Spatial Fragmentation			Cross-Conversion Areas (km2)
		U	A	E	U→A	U→E	A→U	A→E	E→U	E→A
U	Coefficient	−0.83	0.03	−0.20	10.39	−5.18	5.14	−1.48	5.49	1.55
	p-value	0.00	0.30	0.00	0.00	0.19	0.00	0.00	0.00	0.00
A	Coefficient	−0.15	−0.02	−0.17	4.34	3.35	3.69	−5.46	2.70	1.54
	p-value	0.00	0.45	0.00	0.00	0.42	0.00	0.00	0.00	0.00
E	Coefficient	−0.11	−0.02	0.14	4.71	−0.83	2.82	−2.72	1.51	1.08
	p-value	0.02	0.55	0.00	0.00	0.83	0.00	0.00	0.00	0.00

). Specifically, ecological land embedded in urban areas as patches or wedges enhances ventilation and forms local ‘cool islands’, improving the urban thermal environment. For agricultural land, the relationship between LST and its own or surrounding spatial morphology is not significant, suggesting that morphological variation has a limited impact on its heat exchange mechanisms. Although the fragmentation of ecological spaces produces a slight regional cooling effect, its magnitude (coefficient = −0.06, p < 0.001) is smaller than that of urban spaces (coefficient = −0.20, p < 0.05). This means that, under otherwise identical conditions, a one-unit increase in urban spatial fragmentation leads to a 0.2 °C decrease in LST, whereas a one-unit increase in ecological spatial fragmentation results in only a 0.06 °C decrease. Interestingly, ecological fragmentation shows a significant positive correlation with its own LST (coefficient = 0.14, p < 0.001), in contrast to its overall regional effect. This suggests that while fragmented ecological land can provide local cooling to adjacent urban or agricultural areas through edge effects, it simultaneously weakens the core zone’s climate-regulating capacity, leading to thermal deterioration within the ecological space itself. Therefore, land use planning should carefully balance concentrated and dispersed layouts of ecological land.

(2)

Spatial Transformation

Urban LST is most sensitive to land use conversions. The transformation of ecological land into urban land exhibits the strongest warming effect (coefficient = 3.49, p < 0.001), indicating that urban expansion into ecological areas substantially increases local surface temperatures. By contrast, conversion of urban land to agricultural land (coefficient = 2.46, p < 0.01) also causes warming, but to a lesser extent, implying that the thermal impacts of urbanization are often difficult to reverse. Agricultural spaces show pronounced asymmetry in thermal response. The conversion of agricultural land to urban land (coefficient = 4.06, p < 0.001) produces a slightly stronger warming effect than the reverse conversion, suggesting that urbanization exerts a persistent influence on agricultural thermal environments. Notably, the transformation of agricultural land into ecological land (coefficient = −3.87, p < 0.001) generates a significant cooling effect. This can be attributed mainly to two factors: (i) ecological land has stronger vegetation coverage and evapotranspiration capacity than agricultural land, and (ii) although crops provide some cooling, the relatively homogeneous surface of farmland is effectively enhanced by the greater heterogeneity and ecological functionality of newly established ecological areas. The LST of ecological spaces responds more directly to land use change: a reduction in ecological land generally causes warming, whereas its expansion results in cooling. This highlights the irreplaceable role of ecological spaces in climate regulation and underscores the importance of protecting and restoring ecological land as a key strategy for mitigating surface warming.

4. Discussion

Based on the quadtree algorithm, this study systematically analyzes the evolution characteristics of urban–agricultural–ecological spaces in MRYR from 2000 to 2020 and explores their relationship with land surface temperature (LST) changes. Traditional views suggest that the expansion of a certain spatial type is usually accompanied by spatial agglomeration effects. However, the results indicate that there is no significant interaction between spatial scale and fragmentation. In the case of the MRYR, only ecological space aligns with this hypothesis—its area decreased while fragmentation increased. In contrast, agricultural space experienced area growth alongside continuous fragmentation, while urban space showed a “U-shaped” pattern: agglomeration between 2000 and 2010, followed by dispersion between 2010 and 2020. This shift reflects a profound transformation in China’s urbanization strategy: in the early 21st century, the expansion of large cities drove the centralized growth of urban space. Since 2010, however, the expansion of numerous small towns fueled by land finance, along with the disorderly sprawl of large cities, has led to the fragmentation of urban space. Coupled with policies on the protection of arable land, this path has resulted in more dispersed agricultural space even as its total area grew, while ecological space has continued to be encroached upon and fragmented. The study demonstrates that a simple and comparable spatial fragmentation metric can effectively capture the evolutionary characteristics of the three spatial types and reveals the trends in the relationship between spatial scale and fragmentation. This study provides a potential avenue for better understanding regional land use behavior and its environmental impacts.

In multi-city and cross-regional studies, there is a consensus that land types and urban morphology affect LST; our analysis of the MRYR region from 2000 to 2020 both verifies and further refines these findings. First, existing research suggests that more dispersed urban forms help suppress land surface warming, whereas high-density, continuous impervious surfaces lead to surface warming; we have confirmed this view through a mechanistic analysis [44,45]. Second, the cooling effect of ecological land is context- and scale-dependent: when ecological patches are embedded in agricultural or urban matrices, they significantly cool the surrounding environment through evapotranspiration and shading. However, in ecologically dominated landscapes, fragmentation weakens the self-cooling effect and can even lead to warming, as evidenced by the “forest-edge warming” phenomenon [57,58]. Third, thermal responses vary across spatial types and climatic contexts; in warm-humid regions, farmland is more prone to adopting an “urban-like” energy partitioning state than forests due to the albedo–evapotranspiration trade-off [34,53]. In line with this, we found that the local warming effect of agricultural-to-urban conversion is weaker than that of ecological-to-urban conversion, with the effect being further reinforced by the fact that agricultural land is more commonly encroached upon by urban expansion. In cases where expected cooling is not observed, the likely cause is the increase in edges and fragmentation, which reduce effective evapotranspiration. In conclusion, differentiated ecological-space management is necessary: on one hand, the protection of large, continuous ecological patches should be strengthened to enhance temperature regulation; on the other hand, ecological elements should be strategically embedded and connected within agricultural and urban fabrics to maximize buffering while avoiding the dual negative impacts of area loss and fragmentation.

Based on a robust estimation of the fragmentation of urban, agricultural, and ecological spaces, this study offers the following three policy recommendations for land use under the global climate change context: (1) Adopt a principle of cautious development. While full restriction of development is unrealistic amid ongoing urbanization and economic growth, efforts should be made to avoid ineffective and inefficient land expansion and resource waste, thereby mitigating LST increases caused by land use changes. (2) Manage the three space types from a systemic perspective, particularly focusing on the interaction between spatial scale and morphology. By optimizing spatial layout and morphological structure, it is possible to coordinate urban, agricultural, and ecological spaces in a way that offsets the environmental costs of development. (3) Promote a temperature-friendly land development model, including the development of multiple small urban clusters to curb disorderly sprawl, and embedding fragmented ecological land within agricultural and urban spaces. Through sustainable agriculture and urban renewal, the temperature regulation function can be enhanced while avoiding continued encroachment and fragmentation of existing ecological space.

By introducing the quadtree method, this study precisely classifies urban, agricultural, and ecological functional spaces and quantifies their spatial fragmentation, thereby providing a novel analytical perspective and technical pathway for exploring the coupling between spatial patterns and land surface temperature changes at multiple scales. Compared to traditional methods, this approach not only achieves functional identification of land use types but also effectively assesses spatial agglomeration and dispersion. Its advantages are mainly reflected in three aspects: First, it identifies spatial patterns in a structured way, allowing the study to focus on development and protection characteristics while reducing reliance on subjective processes such as remote sensing correction and multi-indicator integration. Second, it clearly delineates spatial boundaries, improving the comparability and stability of statistical analyses across functional spaces. Third, it provides a concise and intuitive fragmentation metric system, enhancing the capacity to model spatiotemporal changes in land use. These advantages help more accurately capture the impacts of human activities on spatial morphology, thus offering technical support for more effective land use planning and management strategies. Ultimately, this contributes to building a climate-friendly spatial structure for land use, serving the broader goal of sustainable development.

Nevertheless, this study still has certain limitations in terms of data reliability and temporal scale. The remote sensing datasets used inevitably contain classification and retrieval errors, which may propagate through the regression process and to some extent weaken the authenticity of the relationship between land use dynamics and land surface temperature. In addition, the interpolation of long-term time series cannot fully eliminate the influence of uneven data quality across different times and regions [59,60]. Therefore, future research will integrate multi-source and higher-resolution remote sensing products to conduct multi-temporal validation and cross-sensor uncertainty analysis. By incorporating long-term LST series to enhance temporal continuity and combining representative validation in regions with reliable data, subsequent studies can improve the robustness of spatial correlation analyses.

5. Conclusions

This study develops a quadtree-based, multi-scale assessment framework for “urban–agricultural–ecological” functional spaces. We partition space by function—placing functional roles at the core of the delineation—to form three mutually exclusive, non-overlapping systems. By coupling hierarchical subdivision with an information entropy criterion, we quantify area, fragmentation, and LST across scale systems, markedly reducing the sensitivity of conventional morphological metrics to pixel size and parameter settings. Within the same measurement framework, we also link the identification of “morphological effects” and “conversion effects,” enabling cross-scale and cross-type comparisons of temperature impacts and providing a reusable tool to optimize climate-friendly spatial governance. Empirically, over the past two decades, core cities in the middle reaches of the Yangtze River have expanded conspicuously outward, encroaching upon agricultural and ecological spaces; urban form has shifted from compactness toward outward sprawl, while agricultural and ecological spaces have become increasingly fragmented. The results show that (i) provided basic functions are safeguarded, moderate decentralization helps mitigate the heat island effect; however, fragmentation of ecological space weakens its temperature regulation capacity and elevates its own LST. (ii) Converting ecological space to urban land produces the strongest local warming; converting agricultural space to urban land also warms the area, whereas conversion from agricultural to ecological space yields substantial cooling.

Accordingly, optimizing the regional thermal environment requires a “concentration–dispersion” balance among the three functional spaces: (i) continuous ecological patches and wedge-shaped ventilation corridors should be embedded to enhance airflow and evapotranspirative cooling, avoiding over-compaction; (ii) the integrity and connectivity of agricultural and ecological spaces should be maintained, constraining excessive fragmentation that erodes climate regulation capacity; (iii) and an “ecology-first” bottom line should be upheld by strictly controlling urban and agricultural encroachment into ecological space, with priority given to orderly ecological restoration. The proposed analytical framework provides a transferable approach for tracking coupled relationships among spatial scale, morphology, and LST, and for evaluating the thermal effects of alternative governance pathways.