Validation of FY-4A/AGRI LST and High Temporal–Spatial Resolution Analysis of Surface Heat Resources in Hunan Province, Central China

Abstract

Land surface temperature (LST) is a crucial parameter in climate and ecology, exerting significant influence on meteorological conditions, ecosystems, and human life. LST data sources are diverse, with remote sensing being the prevailing means of acquisition. FY-4A/AGRI offers high-quality LST products for East Asia. We conducted a comprehensive evaluation and refined analysis of surface heat resources in Hunan Province, central China, over a two-year period using the 4 km/1 h resolution product in this study. The results demonstrate that the FY-4A LST product effectively captures surface temperature (R = 0.893), albeit with a relatively high error level (Bias = −6.295 °C; RMSE = 8.58 °C), particularly in capturing high LST values. The performance of this product is superior in the eastern flat terrain area of Hunan Province compared to its performance in the western mountainous region due to environmental conditions causing systematic errors that contribute to instability in detection deviation for this product. Surface heat resources are more abundant in eastern Hunan Province than in mountainous areas located west and southwardly, and the detailed distribution of them at finer scales is mainly influenced by terrain and climate conditions. There is no obvious seasonal difference in the distribution of heat resources except in winter, and rapid urbanization within Chang–Zhu–Tan urban agglomeration over two years has significantly altered the spatial distribution pattern of surface heat resources across Hunan Province. These findings provide a quantitative baseline for assessing FY-4A satellite’s detection capability while serving as a reference for further application of its LST products and establishing foundations for divisional classification and utilization strategies pertaining to surface heat resources within Hunan Province.

1. Introduction

Land surface temperature (LST) is a crucial parameter in surface energy balance, directly influencing the interaction between the surface and atmosphere [1], as well as various activities such as climate and hydrological cycles [2], animal and plant growth [3], human lives [4], and social production [5]. Consequently, the precise monitoring of LST is widely acknowledged as a high-priority parameter by the International Geosphere and Biosphere Program (IGBP) [6], and the determination of this parameter is indispensable for various analyses pertaining to surface heat resource assessment, quantification of urban heat island effects, and evaluation of surface ecological environment conditions.

The mainstream observation methods for LST include in situ measurements, remote sensing techniques, model simulations, etc., each with its own advantages and disadvantages. While data obtained through manual or instrument-based observations are accurate, they require continuous support in terms of manpower and funds for station construction and maintenance. Moreover, due to hardware limitations and ground environment constraints, in situ measurements cannot provide large-scale spatial coverage or high spatial resolution data products. On the other hand, satellite-based remote sensing can easily acquire real-time land surface parameters over extensive areas, but the accuracy of these products exhibits significant variability due to variations in satellite platforms, sensor configurations, spectral bands, retrieval algorithms, and surface environmental conditions. Consequently, the general accuracy performance of international mainstream satellite remote sensing LST products varies considerably, and their applicability to different surface environmental conditions also differs [7]. Continuous enhancement of instrument performance and retrieval techniques, based on an accurate understanding of product accuracy through authenticity testing and assessment, is essential for optimizing these products [8]. Model simulations can generate long-term series of high spatial–temporal resolution LST products; nevertheless, they demand advanced modeling techniques and precise input parameters while often exhibiting significant time delays [9] that fail to meet requirements for time-efficient applications.

A variety of authenticity testing and evaluation methods can be employed in the validation research of remote sensing products, including ground-based validation [10], satellite product inter-comparison [11], radiation-based validation [12], etc. In situ measurements of temperature or radiation provide the most accurate parameters for verification research [13] and facilitate a better understanding of the calculated validation results. However, due to surface heterogeneity and scale mismatch, there may be certain errors in the verification results [14]. Furthermore, satellite product inter-comparison alone does not yield absolute outcomes and is insufficient to conduct a comprehensive evaluation of remote sensing products [15]. It has been observed that not only do different satellite remote sensing products exhibit significant performance differences but also variations in validation data and contrast methods can lead to errors in evaluation results [16]. Validation research based on measured data serves as the most authentic and accurate reflection parameter for land surface temperature, providing reliable results for satellite product evaluations. Nonetheless, obtaining high-quality, long-term series measured data with high-density, regional-scale sites poses challenges due to the expensive nature of in situ observations which also limits relevant research development.

With the advancement of satellite payload technology and continuous refinement of retrieval algorithms, there has been a significant enhancement in the precision and resolution of satellite-based remote sensing LST products. Notably, the latest generation of official geostationary satellite LST products can achieve kilometer-scale spatial resolution and hour-scale temporal resolution, thereby greatly enhancing their applicability. The FY-4A, as the inaugural satellite of China’s second generation geostationary meteorological satellites, is equipped with cutting-edge detection instruments. And the advanced geostationary radiation imager (AGRI), which serves as the primary instrument of the satellite payload, is primarily designed to capture high temporal–spatial resolution images of land surfaces, atmospheres, and cloud targets [17]. With the application of FY-4A products in China’s meteorological business in recent years, technicians generally reflect that the accuracy of the product is still short of that of the international mainstream products. The authority does not provide a quality report for FY-4A/AGRI LST; however, relevant studies indicate that this product exhibits certain detection errors and a general underestimation phenomenon [18,19], with larger detection errors compared to mainstream Landsat and MODIS products [7,20]. The comparison of FY-4A and MODIS LST products in relevant studies also confirms their high correlation, with a significantly increased difference in higher LST during summer and autumn [21]. Therefore, there is a pressing need for further validation research on the FY-4A LST product to establish a quantitative foundation for its performance in local regions, thereby facilitating the enhancement and application of this product.

The traditional analysis of surface heat resources and urban heat island effect primarily relies on in situ observation data. However, due to the significant heterogeneity of land use types and land surface characteristics such as topography, soil texture, and vegetation condition [4,22], LST exhibits rapid spatial and temporal changes [23]. The limited number of in situ observation stations results in low spatial resolution of data products that fail to meet the demand for high-quality analysis (e.g., Hunan Province’s meteorological bureau has only 99 LST observation stations across an area of 211,800 km²). Nevertheless, satellite data provide the exclusive opportunity for measured LST over a large spatial range with sufficiently high temporal–spatial resolution and comprehensive spatially averaged values rather than individual point measurements [1]. Therefore, remote sensing LST products are more suitable for conducting temporal–spatial analysis of surface heat resources on a larger scale [24]. It is essential to authenticate remote sensing products within specific areas to determine their scientific validity as alternative measured data before assessing surface heat resources based on long-term series remote sensing LST products.

This study aims to comprehensively evaluate the FY-4A/AGRI LST product using in situ observation data from Hunan Province in central China while analyzing surface heat resources based on this long-term series remote sensing product. The research findings can serve as a valuable reference for the further promotion and application of FY-4A LST products, while also providing essential technical support for the assessment and zoning of surface heat resources in Hunan Province. Additionally, this study’s work can offer valuable data support for production management, urban construction, and government decision making.

2. Materials and Methods

2.1. Study Area

Hunan Province is located in south-central China, spanning from 24°38′ to 30°08′ N and 108°47′ to 114°15′ E. Encompassing an area of 211,800 km², the province lies within the transitional zone between the Yunnan-Guizhou Plateau and the Jiangnan Hills as well as the Nanling Mountains to the Jianghan Plain. The province is highest in the south and lowest in the north and is surrounded by mountains on three sides (Figure 1). Hunan Province has a variety of topographical and environmental conditions, with hills alternating with river valleys and basins in the central region, and the low-lying Dongting Lake Plain lies in the north. The climate of Hunan is continental, subtropical, humid, and prone to monsoons, with abundant light, heat, and water resources but large intra-annual variations and significant vertical changes. The province has a high forest cover and a good natural environment.

2.2. Data

2.2.1. FY-4A/AGRI LST

FY-4A is a second-generation geostationary meteorological satellite in the Fengyun series of Chinese meteorological satellites. It was launched in December 2016 and successfully deployed in September 2017, being positioned initially at 99.5° E over the equator before subsequently drifting to 104.7° E. FY-4A was the most advanced integrated atmospheric observation satellite of its time. The three-axis stabilized FY-4 series offers full-disc coverage every 15 min or better (compared to 30 min of FY-2) and the option for more rapid regional and mesoscale observation modes. The Advanced Geostationary Radiation Imager (AGRI) on FY-4A, with 14 channels, can be used to improve applications in a wide range of ocean, land, and atmosphere monitoring and in forecasting extreme weather, especially typhoons and thunderstorms [17]. The LST product is retrieved based on the split-window algorithm [25] and uses two TIR bands (10.3–12.5 µm). The sensitivity of the LST product is 0.2 K, the spatial resolution is 4 km, and the temporal resolution is up to 15 min. The product was projected using the normalized projection (NOM) and was downloaded from http://satellite.nsmc.org.cn/ (accessed on 24 March 2022). Data were stored in netCDF format.

2.2.2. In Situ-Measured LST Data

The in situ LST measurements were obtained from the National Meteorological Observation Stations of CMA. The observation station was subject to stringent construction requirements, necessitating the installation of instruments in an observation field characterized by a flat terrain and a soil-based underlying surface with an area of at least 25 m × 25 m. Moreover, the surrounding environmental conditions are also subjected to explicit restrictions. Therefore, the obtained observation data accurately reflect the local-scale surface state and serve as suitable verification benchmarking data. The in situ data were automatically collected using platinum resistance sensors with an accuracy of 0.1 °C. Following quality control processes, the data range was limited to [−80, 80] °C. Hunan Province has a total of 99 stations (Figure 1), and the data were downloaded from https://data.cma.cn/ (accessed on 22 March 2022) at an hourly temporal resolution.

2.3. Methods

2.3.1. Research Methods

The LST measurement unit of each product in the study was standardized as °C, and the time was standardized as UTC. Given that the data quality flags in FY-4A LST product are identified as invalid parameters, no data filtering was performed on the remote sensing product. The following steps (Figure 2) were undertaken to conduct this research:

(1)

Acquiring FY-4A/AGRI LST product from 1 October 2019 0 h to 30 September 2021 23 h and decoding and extracting data specifically for Hunan Province; employing the average method to standardize the temporal–spatial resolution of the products to a uniform scale of 1 h/4 km; conducting nearest-neighbor sampling to match the remote sensing product with in situ measurements by hour; utilizing this matched dataset for evaluation and assessment based on measured data.

(2)

Selecting stations with optimal, worst, and median correlation coefficient values between remote sensing product and in situ measurements based on evaluation results; selecting a dataset covering a one-year period for comparative analysis of time-series between remote sensing products and in situ-measured data.

(3)

Analyzing surface heat resources of Hunan province based on extracted data from FY-4A LST product, encompassing both years’ worth of data.

2.3.2. Performance Indicators of Authenticity Test

The performance of remote sensing-based LST products in Hunan province was evaluated using following error parameters: Pearson correlation coefficient (R), bias, root mean square error (RMSE), and unbiased RMSE (ubRMSE).

$$R=\frac{\mathrm{cov}(RST,IST)}{{\sigma }_{RST}{\sigma }_{IST}}$$

(1)

$$Bias=\frac{1}{m}{\displaystyle \sum _{i=1}^m(RS{T}_i-IS{T}_i)}$$

(2)

$$RMSE=\sqrt{\frac{1}{m}{\displaystyle \sum _{i=1}^m(RS{T}_i-IS{T}_i{)}^2}}$$

(3)

$$ubRMSE=\sqrt{RMS{E}^2-Bia{s}^2}$$

(4)

where RST is the LST of each remote sensing-based product, IST is the in situ-measured LST, cov () is the covariance, and σ is the standard deviation. Of the preceding parameters, greater R, lesser bias, RMSE, and ubRMSE indicate better product performance.

2.3.3. Parameter Classification Method in Analysis—The Natural Breaks (Jenks)

In the spatial analysis of surface heat resources, due to the large differences in LST conditions among different regions of Hunan province, the traditional parameter classification method can not perform a better hierarchical display of data-intensive areas when covering the classification of extreme value areas. To address this limitation, we employ the natural breaks (Jenks) classification method [26] for mapping. This optimization algorithm enables a more detailed representation of spatial information in analysis and mapping by minimizing intra-category data differences and maximizing inter-category data differences during classification. The calculation steps are as follows:

Step 1.

The user selects attribute x for classification and specifies the desired number of classes, k.

Step 2.

The initial class boundaries are established by generating a set of k-1 random or uniform values within the range [min(x), max(x)].

Step 3.

The mean values of each initial class are calculated, and the sum of squared deviations from these means is computed. The total sum of squared deviations (TSSD) is recorded.

Step 4.

The individual values in each class are then systematically assigned to adjacent classes by adjusting the class boundaries, aiming to minimize the TSSD. This iterative process concludes when the improvement in TSSD falls below a predefined threshold, signifying minimal within-class variance and maximal between-class variance. However, it is important to note that achieving true optimization is not guaranteed. Optionally, the entire process can be repeated from Step 1 or 2, allowing for the comparison of TSSD values.

3. Results and Discussion

3.1. Evaluation of FY-4A LST Using In Situ Measurement

The hourly matched LST dataset of FY-4A/AGRI and in situ measurement of Hunan Province from 1 October 2019 0 h to 30 September 2021 23 h had a total of 5.394 × 105 data quantity after data preprocessing. Comparative analysis shows (Figure 3) that the FY-4A product captured changes in surface temperature for Hunan Province well (R = 0.893), but that it generally underestimated LST (Bias = −6.295 °C) and had some deviation from in situ measurement (RMSE = 8.58 °C; ubRMSE = 5.842 °C), for which ubRMSE was significantly lower than RMSE, but still with a relatively high error value, which could also indicates that the FY-4A LST product was greatly affected by systematic error and random error at the same time. Compared with relevant research, the error level of the FY-4A LST product was higher than that of similar advanced Himawari imager (AHI) from Himawari-8 [18].

The strip with higher brightness in Figure 3 is the center of density of the scatterplot; the trend of its central zone changing with an increase in temperature shows that when LST was low (≤25 °C), the accuracy of FY-4A LST was better and stable, and the center of density of its scatterplot was around the y = x line. But as the temperature increased (>25 °C), the deviation from observation gradually increased, and the underestimation of LST became greater, which may be one reason for its larger overall error (Bias = −6.295 °C; RMSE = 8.58 °C). Moreover, there were also some outliers in the product, which means that even when LST was low (the measured LST was 15–25 °C), the FY-4A/AGRI instrument was unstable in detection.

The performance indexes of in situ-measured LST from 99 stations in Hunan Province with the remote sensing LST product in located FY-4A grids were calculated, respectively, using the matched dataset. From the spatial distribution map (Figure 4) of these indexes, it can be seen that:

• The R value of eastern stations in Hunan Province was generally higher than that of western; for the eastern region, the R value in the northeast stations was higher than that in the southeast. The distribution result of R value shows that the accuracy of satellite remote sensing LST may be closely related with the terrain. The west and south of Hunan are mountainous, while the central region is mostly plain terrain, conducive to satellite remote sensing. The lower R value of some stations in northeastern Hunan may be related to the Dongting Lake, which is the second largest freshwater lake in China.
• The overall deviation of stations in eastern Hunan is smaller than that in western Hunan (the bias is closer to 0 and RMSE value is smaller); the random error in the detection has no obviously changing trend under various environmental conditions. The distribution of error parameters shows that the impact of terrain is mostly manifested in the high systematic error, and the accuracy of the remote sensing product in the mountainous area has more impact factors as well as more complex affect mechanisms than the plain area [27]. However, the instrument capability of FY-4A/AGRI and the retrieval algorithm of LST failed to filter out the impact of complex terrain well, resulting in the significantly higher systematic error level in mountainous areas. From this point of view, the eastern region of Hunan Province is more conducive to remote sensing detection. The distribution of ubRMSE shows that the random error has no spatial distribution characteristics in Hunan Province, the remote sensing detection accuracy is no longer greatly affected by topographic factors after removing the systematic error, but the accuracy of FY-4A LST on a water body is still relatively low.

In the aforementioned validation analysis, FY-4A LST exhibited a high overall error level and an unexpectedly significant underestimation compared to in situ measurements at high LSTs, surpassing MODIS [20], Landsat [7], and Himawari-8 [18] counterparts in terms of error level. The phenomenon under investigation may be attributed to various factors. Firstly, the product retrieval algorithm plays a significant role. Previous studies have confirmed that the Ulivieri (1985) algorithm can effectively retrieve FY-4A LST with an RMSE of 3 K, demonstrating comparable accuracy to MODIS data [21]. Additionally, the National Oceanic and Atmospheric Administration Joint Polar Satellite System enterprise algorithm has shown promising results in retrieving FY-4A LST, achieving an RMSE within 3 K when compared to in situ-measured land surface emissivity-derived LST values [28]. However, the local split-window algorithm yielded a higher RMSE of 6 K for retrieved FY-4A LST, similar to observations from certain stations in central Hunan province (Figure 4). Nevertheless, after applying optimization algorithms, the RMSE of this particular product was reduced to 3.45 K [29], further highlighting the substantial impact of retrieval algorithms on improving LST product accuracy. Therefore, the high product error observed in Figure 3 and Figure 4 may be attributed to the inadequate suitability of the retrieval algorithm utilized for the FY-4A official LST product.

Other potential factors contributing to this include elevated water vapor content within Hunan Province and undulating topography across most regions. It has been confirmed with relevant research that negative values below 4.0 K primarily result from cloud pollution or cirrus clouds on LST retrieval pixels [21]; thus, Hunan’s cloudy climate might indirectly contribute to increased errors as well. Furthermore, it is noteworthy that the validation of FY-4A LST in the aforementioned studies primarily focuses on broad, flat, and uniform areas [21,28,29], including the Heihe River Basin [30], these areas predominantly exhibit arid climatic conditions with lower atmospheric water vapor content compared to the humid climate in Hunan Province. Satellite remote sensing retrieval technology tends to yield more precise land surface information in such regions. Consequently, these two factors may contribute to superior validation results reported by other studies than those obtained in this particular study. Moreover, it should be acknowledged that potential instability of the TIR band within the experimental AGRI instrument [31] equipped on FY-4A could also influence the accuracy of its LST products. Furthermore, despite employing a retrieval algorithm similar to Geostationary Operational Environmental Satellites (GOES)-R products, FY-4A LST exhibits larger errors [32] indicating a performance gap between this test satellite’s payload and mainstream international instruments.

Nevertheless, it is important to note that the verification data utilized solely comprise point-scale in situ observations, which may not fully represent the true LST values for FY-4A’s 4 km resolution pixels. This limitation could potentially introduce inaccuracies in product validation results. Alternatively, employing in situ-measured infrared radiation data as a benchmark might be more preferable; however, acquiring observed radiation data over a large area poses greater challenges compared to obtaining LST data. Nonetheless, despite inherent challenges associated with scale mismatch during remote sensing product verification processes leading to deviations in results, verification based on in situ-measured data still provides valuable reference outcomes.

3.2. Time-Series Analysis between FY-4A LST and In Situ-Measured Data

We selected three stations with the highest, median, and lowest R values between FY-4A LST and in situ measurements, and obtained one year of data (from 1 October 2020 0 h to 30 September 2021 23 h) for conducting time series analysis.

Through comparative analysis of time series data diagrams (Figure 5) from the three stations/grids, the following results were obtained: (1) The FY-4A/AGRI LST product accurately captures the fluctuation trend of LST time series but generally underestimates and has limited ability in capturing high LST values. (2) Comparing R values, there are greater differences in bias and RMSE between FY-4A LST and in situ measurements among stations. For instance, at Dongan Station, FY-4A LST was underestimated by an average of more than 8 °C, and bias was not considered when selecting these stations; thus, there may be other stations/grids with even greater underestimation of LST. (3) Compared to other error indicators, ubRMSE values remained relatively stable among stations indicating that random errors in FY-4A LST were consistent across Hunan Province while systematic errors were responsible for variations in detection deviation due to environmental conditions.

Based on the above analysis results, we manually supplemented the overall averaged bias value within Hunan Province (6.295 °C) for FY-4A LST products from the three grids and conducted another time series analysis. By comparing polylines representing two sets of LST time series data as well as changes in error parameters shown in Figure 6, it is evident that this method significantly reduces deviations observed in FY-4A LST products. Both bias and RMSE values noticeably decreased; particularly noteworthy is that RMSE values almost equaled ubRMSE at Huanghua station which exhibited the optimal performance and Guidong station which showed the worst performance. This further confirms that unstable detection deviation observed in FY-4A LST primarily stems from systematic errors influenced by environmental conditions which can be mitigated with the inclusion of a bias value.

3.3. Analysis of Refined Surface Heat Resources in Hunan Province Based on FY-4A LST

Based on two years of FY-4A LST products, the surface heat parameters were calculated for each FY-4A grid, and a regional analysis of surface heat resources in Hunan Province was conducted using mapping. The spatial distribution of average LST in Hunan Province exhibits a strong correlation with topography and urbanization, as depicted in Figure 7A. The mountainous areas in the west and south display lower levels of LST, while the plains and basins in eastern and central Hunan, characterized by urban agglomerations, consistently exhibit higher mean LST values. Notably, the high-value zone (mean value ≥ 13.3 °C) encompasses major urban centers such as Yueyang City, Yiyang City, Changde City to the north; Changsha City, Zhuzhou City, Xiangtan City, Loudi city, and Shaoyang city at the center; as well as Yongzhou City and Chenzhou city to the south. These findings underscore that both topography and urban underlying surfaces play crucial roles in shaping surface heat resources.

The distribution of the highest LST in two consecutive years is significantly influenced by extreme high temperature climates, which are closely associated with latitude. Consequently, the highest LST levels tend to exhibit similarity in both northern and central regions of Hunan province. In contrast, the southern mountainous areas of Hunan experience comparatively lower maximum LST values over the same period (max value ∈ [31.4, 34.7]). Notably, downtown areas in Hengyang City, Yongzhou City, and Chenzhou City located in southern Hunan display consistently elevated LST levels (max value ∈ [42.9, 66.1]), while other low-lying terrain areas predominantly exhibit yellow and orange colors, indicating relatively moderate maximum LSTs (max value ∈ [36.4, 42.9]).

The distribution of the lowest LST in Hunan Province exhibited a complex pattern (Figure 7C). Based on regional divisions, three areas with high minimum LST values were identified over a span of two years: the Chang–Zhu–Tan urban agglomeration, southern basin area, and western mountainous region. The elevated minimum LST values observed in the first two areas can be attributed to their status as economically developed urban agglomerations. However, it is intriguing that the western mountainous region also displayed relatively high minimum LST values, which may be linked to local topography-induced climate conditions; further investigation is warranted to ascertain the exact cause. Conversely, regions with low minimum LST values encompassed the northeast Dongting Lake area and southwest mountainous region, likely influenced by a combination of reduced human activities and topographic factors.

The distribution of the average daily LST range exhibits distinct patterns, with significantly wider ranges observed in the northwestern and southern regions, comparatively narrower ranges near the provincial border areas, and an intermediate average daily range in the flat terrain area located centrally. Notably, the high daily range values in the northwest region aligned well with the topographical trend. Conversely, in the southern region (mean daily range ∈ [12.7, 19.1]), areas of low elevation experienced high daily ranges, whereas areas with high elevation had lower daily ranges (mean daily range ∈ [0.9, 9.8]). The disparity between topography and daily LST range in these two regions can be attributed to different influencing factors: solar radiation primarily affects surface temperature in mountainous areas of northwest Hunan, resulting in higher daytime LST and subsequently, larger LST variations; warm and humid airflow from the south influences temperature conditions in southern mountainous areas causing cloudy weather because of airflow climbs in the mountainous areas that reduce ridge temperatures during daytime but increase valley temperatures, leading to higher LST ranges due to an enhanced warming effect of this airflow. Additionally, the Nanling Mountains in southern Hunan effectively act as a barrier, impeding warm and humid airflow from the south. Simultaneously, the mountainous regions in the northwest remain relatively unaffected by this airflow.

According to the mean LST values distribution across four seasons of FY-4A grids over two years, as shown in Figure 8, it is evident that the western mountainous area consistently exhibits lower LST levels across all seasons. Particularly, the northwest mountainous region displays the lowest LST levels throughout the year within the province. The average distribution of LST in the central region and eastern region remains relatively consistent during spring, summer, and autumn, aligning with the overall average LST distribution depicted in Figure 7A, except for an expanded range of high temperatures during summer. This observation also underscores how LST patterns vary with topographic conditions and urban centers. However, winter exhibits distinct differences in LST distribution compared to other seasons, notably characterized by relatively low LST values in northeastern Hunan and a more uniform LST distribution across western areas. Although one would expect higher temperatures in this northeast region, the same as the overall average condition, winter temperatures have dropped to match those observed in central regions, possibly due to influences from East Asian monsoons during this season. Prevailing winds from northeasterly directions in winter bring dry and cold continental air masses, resulting in lower temperatures on the northeastern plain.

The Chang–Zhu–Tan urban agglomeration, located in the central part of Hunan Province, is a national urban agglomeration in China that comprises Changsha City, Zhuzhou City, and Xiangtan City. It represents the most densely populated and economically developed area in Hunan Province. By extracting mean daily LST data from the first day and last day of the two-year dataset for this urban agglomeration area, we can draw relevant conclusions about its development by comparing LST distribution between two diagrams (Figure 9). Although differences in meteorological conditions result in varying LST levels and variation ranges between Figure 9A,B, spatial distribution of LST combined with heat island effect suggests that development within this urban agglomeration tends to be centralized. Junction areas among the three cities are also densely distributed areas with high LST values which have expanded over two years. The temperature difference between core areas at the center of this region’s urban center versus remote rural regions has increased significantly over time. These differences in LST spatial distribution highlight rapid construction within this urban agglomeration during these past two years while reflecting China’s ongoing commitment to developing it, as outlined in their 14th Five-Year Plan.

According to the aforementioned analysis results, the distribution of surface heat resources in Hunan Province based on FY-4A LST can be summarized as follows: 1. The surface heat resources in the flat terrain eastern part of Hunan Province generally exhibit superiority over those in mountainous areas located in the west and the south. The central region and northern region are prone to extreme heat, while the northwestern region and certain parts of the south experience higher daily LST range, which is conducive to crop growth. 2. The distribution pattern of LST remains similar across spring, summer, and autumn seasons in Hunan Province, in which the eastern region demonstrates distinct advantages regarding surface heat resources. Conversely, during winter months, primary heat resources tend to concentrate in the southeast region. 3. Furthermore, changes observed in surface heat resources also highlight rapid development within the Chang–Zhu–Tan urban agglomeration area because of the indicated tendency towards concentration over the two-year period, with an increasing disparity between core and surrounding areas.

4. Conclusions

In this study, we conducted a comprehensive verification of remote sensing products based on the FY-4A/AGRI LST product and in situ-measured LST data from Hunan province over a two-year period. Our analysis determined the ability of FY-4A/AGRI to capture LST and enabled us to carry out high spatio-temporal resolution analysis of surface heat resources using this data product. In summary, our research lead us to draw the following conclusions:

(1)

The FY-4A/AGRI LST product effectively captures surface temperature in Hunan Province; however, it exhibits a high level of error that becomes more pronounced when temperatures rise above 25 °C. The main reason for the unstable detection deviation of FY-4A LST is attributed to systematic errors influenced by environmental conditions, which can be optimized by incorporating a bias value.

(2)

Spatial analysis reveals the better performance of the FY-4A LST product in the eastern part of Hunan Province compared to the western part due to terrain conditions. The flat terrain in eastern Hunan Province contributes to mitigating systematic errors in the products. Time series analysis demonstrates the product’s ability to accurately capture LST fluctuation trends; however, a general underestimation phenomenon persists and its capability to detect high surface temperatures is limited.

(3)

Surface heat resources are generally more abundant in the eastern region than mountainous areas of west and south, with finer distribution divisions primarily driven by terrain and climate conditions. Apart from winter months, there are no significant differences observed in heat resource distribution among other seasons, and the rapid urban agglomeration development within Chang–Zhu–Tan over two years has led to noticeable surface heat resource changes.

The validation of the LST product from the FY-4 series geostationary satellite and its application in meteorological business have been our team’s primary research focus. In the initial stage, we published an article on the comprehensive assessment of accuracy for the FY-4A/AGRI LST product and analyzed the influencing mechanisms of environmental factors [18]. Building upon this foundation, the present study further expands the spatio-temporal analysis of FY-4A product accuracy based on observational data, thereby exploring its potential application in surface heat resources analysis. The detection capability and practical value of this Chinese domestic satellite are confirmed with these efforts. We will continue to dedicate ourselves to this field in future endeavors, and with a particular focus on the newly released FY-4B LST product.