A method for generating hourly high-resolution air temperature remote sensing data based on a diurnal temperature variation model

Abstract

This invention discloses a method for generating hourly high-resolution remote sensing temperature data based on a diurnal temperature variation model. The method includes: using daily maximum and minimum temperatures retrieved from 1km resolution remote sensing, applying extreme value constraints and amplitude correction to the diurnal temperature variation curve fitted based on reanalysis data; simultaneously extracting short-term high-frequency fluctuation components from the reanalysis data for temporal compensation; and finally generating spatially continuous and temporally reasonable hourly high-resolution temperature data. This invention fully integrates the high spatial resolution advantage of remote sensing data, the physical diurnal variation pattern of the DTC model, and the hourly perturbation characteristics of reanalysis data. While improving the spatial refinement of temperature, it ensures the accuracy of intraday variation processes, providing a high-precision data source for regional climate research and urban thermal environment monitoring.

Description

Technical Field

[0001] This invention relates to the field of remote sensing temperature generation and reconstruction technology, specifically to a method for generating hourly high-resolution remote sensing temperature based on a diurnal temperature variation model. Background Technology

[0002] Temperature is a key meteorological element characterizing the energy balance and climate change on the Earth's surface. High spatial resolution temperature data has significant application value in climate change research, urban heat island effect analysis, agricultural weather forecasting, ecological environment monitoring, and energy assessment. Meanwhile, continuous hourly temperature data provides solid data support for high-precision research and operational applications such as refined thermal environment simulation, temporal analysis of vegetation physiological activities, and temporal estimation of building energy consumption. It is the core foundation for achieving accurate inversion and dynamic monitoring of the multi-scale, full-process surface thermal environment.

[0003] Currently, the main sources of hourly temperature data fall into two categories: First, meteorological station observation data, which provides highly accurate hourly temperature observation information and serves as a fundamental data source for climate change research. However, the number of meteorological stations is insufficient, and their spatial distribution is relatively sparse, making it difficult to accurately depict the spatial distribution pattern of temperature at the regional scale. Furthermore, the observation data is point-scale information, making it difficult to directly apply to gridded and regionalized climate studies. Second, reanalysis meteorological data, which is generated by fusing numerical weather prediction models with multi-source observation data. This type of data is a global meteorological data product, such as the ERA5 reanalysis data released by the European Centre for Medium-Range Weather Forecasts (ECMWF). It provides global hourly temperature, humidity, wind speed, and other meteorological variables, and has the advantages of strong temporal continuity and wide spatial coverage. However, its spatial resolution is relatively low (generally 0.1°, i.e., 10km, or even coarser), making it difficult to accurately depict spatial differences in temperature in complex terrain and urban areas. Furthermore, the reanalysis data exhibits a systematic spatial smoothing effect during the assimilation process. This effect directly leads to a systematic underestimation of the daily maximum temperature and a systematic overestimation of the daily minimum temperature, resulting in a significant compression of the daily temperature variation amplitude and an inability to accurately reflect the actual characteristics of the Earth's surface.

[0004] While satellite remote sensing technology can generate continuous spatial temperature fields using methods such as the temperature-vegetation index, regression statistics, and machine learning, geostationary satellite remote sensing data with high temporal resolution suffers from low spatial resolution, and its disk imaging mode leads to severe geometric and radiometric distortions at image edges. Commonly used polar-orbiting remote sensing satellites with high spatial resolution can only perform observations twice a day, thus only generating continuous daily maximum and minimum temperatures, and cannot directly support hourly temperature reconstruction. Among existing technologies related to remote sensing generation of air temperature, there is a method for estimating daily minimum air temperature with full coverage. Based on XGBoost, which integrates remote sensing index sets, altitude information, and surface temperature information, it can obtain hourly air temperature data with a spatial resolution of 1km. However, this type of technology relies entirely on remote sensing and surface parameters for statistical modeling. The generated hourly air temperature is only a statistical fitting result in the spatial dimension, lacking a reasonable intraday warming and cooling time sequence structure and physical rationality. It is prone to problems such as disordered temperature change trends, misalignment of day and night extreme values, and distortion of short-term fluctuations. It is difficult to reproduce the real dynamic change process of near-surface air temperature and cannot meet the requirements of high-resolution and high-reliability refined applications. Summary of the Invention

[0005] To address the technical problem that existing hourly temperature data has low spatial resolution and is difficult to reflect local temperature change characteristics, this invention provides a method for generating hourly high-resolution temperature remote sensing data based on a diurnal temperature variation model. This method can achieve spatially continuous and temporally reliable generation of hourly high-resolution temperature data, thereby improving the ability to characterize the temporal variation features of local temperature.

[0006] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for generating hourly high-resolution air temperature remote sensing data based on a diurnal temperature variation model, the method comprising the following steps:

[0008] S1 generates high-resolution remote sensing data of daily maximum and minimum temperatures based on remote sensing features and meteorological station observation data, which are spatially continuously distributed.

[0009] S2, resample the coarse-resolution hourly temperature data of the reanalysis data to the same resolution as the remote sensing data, and fit the daily temperature variation model based on the resampled hourly temperature data of the reanalysis data to obtain the daily temperature variation trend curve;

[0010] S3, extract the short-time high-frequency fluctuation components of hourly air temperature relative to the diurnal temperature variation model from the reanalysis;

[0011] S4. The high-resolution remote sensing data of the highest and lowest daily temperatures, which are continuously distributed in space, are used as extreme value constraints to input the daily temperature variation model. The daily temperature variation trend curve of the daily temperature variation model is subjected to amplitude correction and shape reconstruction so that the curve is consistent with the local real thermal extreme value while maintaining the original intraday variation trend shape.

[0012] S5, the corrected and reconstructed daily temperature variation trend curve is superimposed and fused with the short-term high-frequency fluctuation component to obtain the final hourly high-resolution temperature data.

[0013] Step S1 further includes:

[0014] A model for estimating daily maximum temperature was constructed using the daily maximum temperature observed at meteorological stations as the dependent variable, and the daytime surface temperature, normalized difference vegetation index (NDVI), normalized difference water index (MNDWI), albedo, impervious surface cover, altitude, and annual DOY of the corresponding remote sensing pixels as independent variables. A model for estimating daily minimum temperature was constructed using the daily minimum temperature observed at meteorological stations as the dependent variable, and the nighttime surface temperature, NDVI, MNDWI, albedo, impervious surface cover, altitude, and annual DOY of the corresponding remote sensing pixels as independent variables.

[0015] During the model building process, the input features are standardized and outlier screening is performed. The model parameters are optimized through grid search, and K-fold cross-validation is used to complete the model training and validation, thereby obtaining the optimal estimation model for estimating the daily maximum and minimum temperatures.

[0016] The trained remote sensing estimation models for daily maximum and minimum temperatures were applied to the corresponding independent variables to obtain spatially continuous remote sensing data for daily maximum and minimum temperatures at a resolution of 1 km.

[0017] Furthermore, in step S2, the process of fitting a diurnal temperature variation model based on the resampled reanalysis hourly temperature data to obtain the diurnal temperature variation trend curve includes the following steps:

[0018] On a pixel-by-pixel, day-by-day basis, a diurnal temperature variation model is used to fit the hourly air temperature after resampling and reanalysis, constructing a diurnal variation curve to characterize the overall features of diurnal temperature variation; the expression of the diurnal temperature variation model is:

[0019] ;

[0020] In the formula, The fitted temperature at time t is the DTC. and These are the lowest and highest temperatures of the day; and For sunrise and sunset times; The length of a day is the result of subtracting the sunrise time from the sunset time. This is the time parameter by which the temperature peak lags behind the solar altitude angle peak. The nighttime cooling coefficient; Temperature at sunset; This is the adjusted time variable.

[0021] Step S3 further includes:

[0022] Pixel-by-pixel and day-by-day, short-time fluctuation components are calculated by fitting the hourly air temperature and its corresponding diurnal temperature variation model to the reanalysis data; the calculation formula for the short-time high-frequency fluctuation components is as follows:

[0023] ;

[0024] in, Let be the short-time high-frequency fluctuation component at time t. Let be the reanalysis temperature at time t; Let t be the DTC fitted temperature.

[0025] Further, in step S5, the corrected and reconstructed daily temperature variation trend curve is superimposed and fused with the short-term high-frequency fluctuation component using the following formula to generate hourly high-resolution temperature data:

[0026] ;

[0027] In the formula, To generate the final hourly high-resolution temperature at time t, To fit the temperature to the diurnal temperature variation model after correction by remote sensing daily extreme temperature values, Let t be the short-time high-frequency fluctuation component at time t.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0029] This invention presents a method for generating hourly high-resolution remote sensing air temperature based on a diurnal temperature variation model. By combining a diurnal temperature variation (DTC) model constructed from the same source with short-term fluctuation compensation, the temporal resolution is improved to the hourly level, based on the 1km resolution daily maximum and minimum temperatures estimated through remote sensing. This achieves refined generation of hourly high-resolution air temperature, effectively overcoming the problems of insufficient spatial resolution and low accuracy in traditional reanalysis hourly air temperature data. Compared to existing technologies that rely solely on statistical modeling or simply substitute daily extreme values, this invention overcomes the technical bottleneck of balancing spatial accuracy and temporal dynamics. It first fits the diurnal temperature variation pattern conforming to atmospheric physical laws based on reanalysis data, then introduces remote sensing-estimated extreme values ​​to constrain and correct the curve amplitude. Simultaneously, it superimposes short-term fluctuation components extracted from the same-source reanalysis data to recover hourly-level perturbation details, achieving an organic unity of high spatial resolution, reasonable diurnal variation pattern, and realistic temporal fluctuations. This method significantly improves the physical rationality and dynamic detail representation of hourly air temperature while ensuring spatial refinement, enhancing the accuracy and practicality of near-surface air temperature spatiotemporally continuous datasets. Attached Figure Description

[0030] Figure 1 This is a flowchart of the hourly high-resolution remote sensing method for generating air temperature based on a diurnal temperature variation model according to the present invention.

[0031] Figure 2 This is a reanalysis hourly temperature daily variation fitting result graph for a certain pixel from 6:00 on May 11, 2020 to 5:00 on May 12, 2020;

[0032] Figure 3 This is a short-term temperature fluctuation chart for a certain pixel from 6:00 on May 11, 2020 to 5:00 on May 12, 2020.

[0033] Figure 4 This is an hourly temperature change map of a certain pixel from 6:00 on May 11, 2020 to 5:00 on May 12, 2020, obtained by correcting for extreme remote temperature values ​​and superimposing short-term fluctuations.

[0034] Figure 5 A schematic diagram showing the comparison between the hourly temperature estimated by remote sensing at various stations and the hourly temperature observed at the stations in 2020;

[0035] Figure 6 This is a remote sensing temperature distribution map of the study area at 1km resolution on May 11, 2020, showing hourly temperature changes. Detailed Implementation

[0036] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0037] This invention discloses a method for generating hourly high-resolution air temperature remote sensing data based on a diurnal temperature variation model. The method includes the following steps:

[0038] S1 generates high-resolution remote sensing data of daily maximum and minimum temperatures based on remote sensing features and meteorological station observation data, which are spatially continuously distributed.

[0039] S2, resample the coarse-resolution hourly temperature data of the reanalysis data to the same resolution as the remote sensing data, and fit the daily temperature variation model based on the resampled hourly temperature data of the reanalysis data to obtain the daily temperature variation trend curve;

[0040] S3, extract the short-time high-frequency fluctuation components of hourly air temperature relative to the diurnal temperature variation model from the reanalysis;

[0041] S4. The high-resolution remote sensing data of the highest and lowest daily temperatures, which are continuously distributed in space, are used as extreme value constraints to input the daily temperature variation model. The daily temperature variation trend curve of the daily temperature variation model is subjected to amplitude correction and shape reconstruction so that the curve is consistent with the local real thermal extreme value while maintaining the original intraday variation trend shape.

[0042] S5, the corrected and reconstructed daily temperature variation trend curve is superimposed and fused with the short-term high-frequency fluctuation component to obtain the final hourly high-resolution temperature data.

[0043] This invention provides a method for generating hourly high-resolution remote sensing air temperature based on a diurnal temperature variation model, enabling the refined generation of hourly air temperature data with a high spatial resolution of 1 km. This method comprehensively utilizes remote sensing data and reanalysis air temperature data, constructing air temperature data with both high spatial and temporal resolution through a path of "diurnal variation trend modeling - extreme value constraint correction - high-frequency fluctuation compensation." The specific steps are as follows:

[0044] 1) Preprocessing of multi-source remote sensing and reanalysis data

[0045] Data for temperature estimation was obtained from MODIS remote sensing products, including extracting daily daytime LST and nighttime LST from MOD11A1 products, calculating the Normalized Difference Water Index (MNDWI) from surface reflectance from MOD09A1 products, extracting the Normalized Difference Vegetation Index (NDVI) from MOD13Q1 products, extracting monthly synthetic albedo from MCD43A3 products, and extracting elevation information from SRTM / DEM products. All data were standardized to a 1km resolution, and NDVI and albedo were interpolated to a diurnal scale. Additionally, 10m resolution impervious surface information was extracted from DynamicWorld land cover products, and then the impervious surface cover (ISC) at 1km resolution was calculated.

[0046] The coarse-resolution hourly temperature was extracted from the reanalysis data, and then resampled to a spatial resolution of 1 km to match the spatial resolution of the remote sensing data. The daily extreme temperatures, namely the daily maximum and minimum temperatures, were calculated based on the hourly temperatures.

[0047] 2) Construction of a remote sensing estimation model for daily extreme temperature

[0048] Using the daily maximum temperature observed at meteorological stations as the dependent variable, and the daytime surface temperature, NDVI, MNDWI, albedo, impermeable surface coverage, altitude, and annual DOY of the corresponding remote sensing pixels as independent variables, a daily maximum temperature estimation model was constructed based on the XGBoost machine learning algorithm.

[0049] Using the daily minimum temperature observed at meteorological stations as the dependent variable, and the nighttime surface temperature, NDVI, MNDWI, albedo, impermeable surface coverage, altitude, and annual DOY of the corresponding remote sensing pixels as independent variables, a daily minimum temperature estimation model was constructed based on the XGBoost machine learning algorithm.

[0050] During model construction, input features are standardized and outlier filtering is performed to improve model stability. Subsequently, a grid search method is used to optimize key parameters of the XGBoost model, including the maximum tree depth, learning rate, subsampling ratio, column sampling ratio, and number of weak learners, to obtain the optimal model. To avoid overfitting, K-fold cross-validation is used for training and validation. The optimal parameter combination is selected based on the cross-validation results to complete the final model training.

[0051] 3) Remote sensing estimation of daily extreme temperature

[0052] The trained remote sensing estimation models for daily maximum and minimum temperatures were applied to the corresponding independent variables: daytime surface temperature (or nighttime surface temperature if estimating daily minimum temperature), NDVI, MNDWI, albedo, impermeable surface cover, altitude, and annual DOY, to obtain spatially continuous 1km resolution remote sensing data for daily maximum and minimum temperatures.

[0053] 4) DTC model fitting based on hourly temperature reanalysis

[0054] On a pixel-by-pixel and day-by-day basis, the daily temperature variation (DTC) model is used to fit the hourly temperature after resampling and reanalysis to construct a daily variation curve to characterize the overall features of the daily temperature variation.

[0055] (1);

[0056] In the formula, The fitted temperature at time t is the DTC. and These are the lowest and highest temperatures of the day; and For sunrise and sunset times; The length of a day is the result of subtracting the sunrise time from the sunset time. The time parameter (in hours) in which the temperature peak lags behind the solar altitude angle peak. This is the nighttime cooling coefficient (exponential decay rate). Temperature at sunset; The adjusted time was to account for the temperature drop after midnight.

[0057] 5) Extraction of hourly temperature short-term fluctuation components within the day

[0058] In addition to the overall trend dominated by daytime solar radiation loss and nighttime longwave radiation loss, hourly temperature data also superimposes high-frequency short-term fluctuations caused by factors such as weather system changes, cloud cover disturbances, and local surface differences. These fluctuations reflect rapid temperature changes on an hourly scale and are an important component of intraday temperature dynamics. The DTC model can only fit the overall trend of temperature change and cannot effectively capture the short-term characteristics of temperature loss. The short-term high-frequency fluctuation components are calculated pixel-by-pixel and day-by-day based on the hourly temperature from the reanalysis data and its corresponding DTC-fitted temperature:

[0059] (2);

[0060] in, Let be the short-time high-frequency fluctuation component at time t. Let be the reanalysis temperature at time t; Let t be the DTC fitted temperature.

[0061] The calculated short-term fluctuation components are used to characterize the short-term deviation of the actual temperature from the DTC diurnal variation curve, providing high-frequency correction information for subsequent hourly temperature calculations.

[0062] 6) Constraint-based correction of DTC curves based on remotely sensed extreme air temperatures

[0063] The DTC model is based on reanalysis data, which has low spatial resolution and a smoothing effect, leading to an underestimation of the daily temperature variation amplitude. To address this bias, remotely sensed estimated daily maximum and minimum temperatures are introduced to constrain and correct the DTC fitting curve.

[0064] Specifically, the daily maximum and minimum temperatures of the remote sensing day are substituted into the DTC model pixel by pixel to replace the original maximum and minimum temperatures, resulting in the corrected hourly temperatures. Through this substitution process, the amplitude of the DTC fitting curve is constrained and reconstructed, so that it is consistent with the extreme values ​​of remote sensing temperature with high spatial resolution while maintaining the diurnal variation trend, thereby improving the ability of the DTC curve to characterize local temperature differences.

[0065] 7) Hourly high-resolution temperature generation

[0066] Based on the constrained DTC fitting curve and short-term fluctuation components, the two are superimposed and fused to generate hourly high-resolution temperature data.

[0067] (3);

[0068] In the formula, To generate the final hourly high-resolution temperature, The fitted temperature of DTC is the temperature after correction by remote sensing daily extreme temperature constraints. This refers to the short-time high-frequency fluctuation component calculated in step 4.

[0069] Short-term temperature fluctuations This primarily reflects high-frequency temperature disturbances, whose formation mechanisms are closely related to weather system changes, cloud cover evolution, and regional-scale energy exchange processes, exhibiting strong temporal continuity and regional consistency. Therefore, superimposing the short-term fluctuation components extracted from the reanalysis temperature onto the DTC fitting curve constrained by remote sensing extrema can effectively recover the details of hourly-scale dynamic changes in temperature while maintaining the overall trend and spatial distribution characteristics of daily temperature variations. This ensures that the final generated hourly high-resolution temperature data accurately reflects the intraday temperature variation process over time.

[0070] Example

[0071] This example demonstrates a method for generating hourly high-resolution air temperature remote sensing data based on a diurnal temperature variation model, using a specific region as the study area. The technical process is as follows: Figure 1 As shown, the specific implementation steps are as follows:

[0072] 1) Core data were obtained from MODIS series products and related data sources and standardized. Daily inter-day surface temperature (LST) was extracted from MOD11A1, Normalized Difference Water Index (MNDWI) was calculated using MOD09A1, Normalized Difference Vegetation Index (NDVI) was extracted using MOD13Q1, monthly synthetic albedo was obtained using MCD43A3, impermeable surface cover (ISC) at 1km resolution was calculated by distinguishing land cover types using DynamicWorld, and elevation data was extracted using SRTM1. The ERA5_land reanalysis data was selected, and the initial values ​​of daily maximum and minimum temperatures were extracted. All data were uniformly resampled to a daily time scale and 1km spatial resolution, and bilinear interpolation was used to ensure data continuity. Simultaneously, the hourly temperature data from the ERA5_land reanalysis were resampled in the same way to ensure subsequent spatial matching.

[0073] 2) Construction of the XGBoost Temperature Estimation Model and Generation of Remote Sensing Extreme Temperatures: The daily maximum and minimum temperatures observed at meteorological stations were used as dependent variables (true values), and relevant parameters of the corresponding remote sensing pixels at each station were used as independent variables, specifically including nighttime surface temperature, NDVI, MNDWI, albedo, impermeable surface cover (ISC), altitude, and annual day of day (DOY). DOY reflects the impact of seasonal variations on temperature, while ISC helps characterize local thermal differences. The XGBoost model from machine learning was used to construct the temperature estimation model. This model has strong feature fitting ability and anti-overfitting ability, effectively uncovering the complex nonlinear relationship between independent variables and temperature. During model training, the training and validation sets were divided according to a reasonable ratio. Parameter optimization (such as learning rate, tree depth, and number of iterations) improved the model's estimation accuracy. Finally, the trained model outputs remote sensing estimates of the daily maximum and minimum temperatures at a spatial resolution of 1 km for the study area, providing accurate spatialized extreme value data for subsequent constraint correction of the DTC model.

[0074] 3) DTC Model Fitting and Reanalysis of Hourly Temperatures: The DTC (Diurnal Temperature Variation) model was used to fit the resampled 1km hourly reanalysis temperature data to construct a theoretical curve for diurnal temperature variation. While the resampled ERA5_land hourly reanalysis temperature data possesses high temporal resolution and reflects hourly temperature changes, it still exhibits spatial smoothing effects due to limitations in the reanalysis algorithm itself, making it difficult to accurately reflect temperature differences in small-scale regions (such as local temperature differences between mountainous and plain areas, or between urban and suburban areas). Therefore, a diurnal temperature variation model was used for fitting. This model captures the overall trend of temperature "warming after sunrise, peaking in the afternoon, and cooling after sunset" based on the intraday temperature variation pattern, constructing a continuous and smooth theoretical curve for diurnal temperature variation. This provides a stable benchmark for subsequent short-term fluctuation extraction. The fitting results are shown in [link to data]. Figure 2 .

[0075] 4) Extraction of short-term temperature fluctuation components: In addition to containing the overall trend of intraday changes, hourly temperature data also contains high-frequency short-term fluctuation characteristics caused by weather fluctuations (such as short-term cloud cover changes and local wind disturbances) and local differences in underlying surfaces (such as the thermal regulation of small-scale water bodies and forests). These characteristics are an important part of the actual temperature changes and directly affect the authenticity and detail of the temperature data. If only the DTC model is used for fitting, this high-frequency fluctuation information will be lost, resulting in overly smooth temperature data that cannot reflect the dynamic details of the actual temperature. Therefore, short-term fluctuation components are extracted from the hourly temperature data of the reanalysis and calculated using formula (2) (i.e., short-term fluctuation component = reanalysis hourly temperature - DTC model fitted temperature). The extracted short-term fluctuation components can accurately reflect the high-frequency change details of the temperature. The results are shown in […]. Figure 3 .

[0076] 5) DTC Model Extreme Value Constraints and Corrections: Using the remote sensing-estimated daily maximum and minimum temperatures obtained in step 2), the initial reanalysis daily maximum and minimum temperature values ​​used in the DTC model are replaced to precisely constrain and correct the amplitude of the DTC fitting curve. Due to the spatial smoothing effect of the reanalysis data, the extracted daily extreme temperatures are difficult to reflect local small-scale differences. In contrast, the remote sensing-estimated daily extreme temperatures have a high spatial resolution of 1 km, accurately depicting the temperature extreme differences across different underlying surfaces and topography within the region. This constraint and correction step allows the amplitude of the theoretically fitted DTC curve to accurately match the actual surface temperature extremes, preserving the overall trend of intraday temperature variation while incorporating local extreme value differences, achieving a precise fit between the theoretical curve and actual temperature. The correction results are shown in [link to relevant documentation]. Figure 4 .

[0077] 6) Generation and accuracy verification of hourly high-resolution air temperature: Based on the above steps, hourly high-resolution air temperature for the study area was generated. The estimated air temperature was then compared with the observed air temperature at meteorological stations within the study area for verification. The verification results are shown in […]. Figure 5 The validation statistics show that the effective sample size N=35136, the coefficient of determination R²=0.95 (the closer R² is to 1, the stronger the correlation between the estimated and observed values), the mean absolute error (MAE) MAE=1.62℃, and the root mean square error (RMSE) RMSE=2.11℃ (the smaller the MAE and RMSE, the smaller the estimation error). These validation results fully demonstrate that this method can accurately generate hourly temperatures at multiple stations and with high resolution, possessing good reliability and practicality, and can meet the needs of subsequent related research and applications.

[0078] 7) Generation of hourly temperature spatial distribution in the study area: Based on the single-pixel temperature generation method, batch calculations were performed on all raster pixels in the study area on May 11, 2020, to generate a complete 1km resolution hourly temperature dataset for the study area. The results are shown in [link to dataset]. Figure 6 . Figure 6 The data clearly shows the spatial distribution of temperature in the study area at 24 times from 00:00 to 23:00, fully presenting the diurnal variation pattern of regional temperature: the temperature is generally low at night (00:00-06:00), and the spatial difference within the region is relatively small; the temperature gradually increases with the increase of solar radiation during the day (07:00-18:00), reaching the peak of the day in the afternoon (13:00-15:00); at the same time, the results accurately capture the spatial heterogeneity of local temperature caused by factors such as topography and underlying surface type, and the temperature differences between mountainous areas and plains, and between water bodies and land are clearly distinguishable (e.g., the temperature in mountainous areas is lower than that in plains, and the temperature in water areas is lower than that in surrounding land). This fully demonstrates the core advantage of the dataset generated by this method, which has a high spatial resolution of 1km, and can provide accurate data support for regional-scale temperature-related studies.

[0079] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0080] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for generating hourly high-resolution air temperature remotely based on temperature diurnal model, characterized in that, The method includes the following steps: S1 generates high-resolution remote sensing data of daily maximum and minimum temperatures based on remote sensing features and meteorological station observation data, which are spatially continuously distributed. S2, resample the coarse-resolution hourly temperature data of the reanalysis data to the same resolution as the remote sensing data, and fit the daily temperature variation model based on the resampled hourly temperature data of the reanalysis data to obtain the daily temperature variation trend curve; S3, extract the short-time high-frequency fluctuation components of hourly air temperature relative to the diurnal temperature variation model from the reanalysis; S4. The high-resolution remote sensing data of the highest and lowest daily temperatures, which are continuously distributed in space, are used as extreme value constraints to input the daily temperature variation model. The daily temperature variation trend curve of the daily temperature variation model is subjected to amplitude correction and shape reconstruction so that the curve is consistent with the local real thermal extreme value while maintaining the original intraday variation trend shape. S5, the corrected and reconstructed daily temperature variation trend curve is superimposed and fused with the short-term high-frequency fluctuation component to obtain the final hourly high-resolution temperature data; Step S2, which involves fitting a diurnal temperature variation model based on the resampled hourly temperature data to obtain the diurnal temperature variation trend curve, includes the following steps: On a pixel-by-pixel, day-by-day basis, a diurnal temperature variation model is used to fit the hourly air temperature after resampling and reanalysis, constructing a diurnal variation curve to characterize the overall features of diurnal temperature variation; the expression of the diurnal temperature variation model is: ； In the formula, The fitted temperature for DTC at time t; and These are the lowest and highest temperatures of the day; and For sunrise and sunset times; The length of a day is the result of subtracting the sunrise time from the sunset time. This is the time parameter by which the temperature peak lags behind the solar altitude angle peak. The nighttime cooling coefficient; Temperature at sunset; This is the adjusted time variable.

2. The method for generating hourly high-resolution air temperature remote sensing based on a diurnal temperature variation model according to claim 1, characterized in that, Step S1 further includes: A model for estimating daily maximum temperature was constructed using the daily maximum temperature observed at meteorological stations as the dependent variable, and the daytime surface temperature, normalized difference vegetation index (NDVI), normalized difference water index (MNDWI), albedo, impervious surface cover, altitude, and annual DOY of the corresponding remote sensing pixels as independent variables. A model for estimating daily minimum temperature was constructed using the daily minimum temperature observed at meteorological stations as the dependent variable, and the nighttime surface temperature, NDVI, MNDWI, albedo, impervious surface cover, altitude, and annual DOY of the corresponding remote sensing pixels as independent variables. During the model building process, the input features are standardized and outlier screening is performed. The model parameters are optimized through grid search, and K-fold cross-validation is used to complete the model training and validation, thereby obtaining the optimal estimation model for estimating the daily maximum and minimum temperatures. The trained remote sensing estimation models for daily maximum and minimum temperatures were applied to the corresponding independent variables to obtain spatially continuous remote sensing data for daily maximum and minimum temperatures at a resolution of 1 km.

3. The method for generating hourly high-resolution air temperature remote sensing based on a diurnal temperature variation model according to claim 1, characterized in that, Step S3 further includes: Pixel-by-pixel and day-by-day, short-time fluctuation components are calculated by fitting the hourly air temperature and its corresponding diurnal temperature variation model to the reanalysis data; the calculation formula for the short-time high-frequency fluctuation components is as follows: ； in, Let be the short-time high-frequency fluctuation component at time t. Let be the reanalysis temperature at time t; Let t be the DTC fitted temperature.

4. The method for generating hourly high-resolution air temperature remote sensing based on a diurnal temperature variation model according to claim 1, characterized in that, In step S5, the corrected and reconstructed daily temperature variation trend curve is superimposed and fused with the short-term high-frequency fluctuation component using the following formula to generate hourly high-resolution temperature data: ； In the formula, To generate the final hourly high-resolution temperature at time t, To fit the temperature to the diurnal temperature variation model after correction by remote sensing daily extreme temperature values, Let t be the short-time high-frequency fluctuation component at time t.

Images