A method for assessing heat waves based on remote sensing estimation and reconstructed temperature

Abstract

The present invention discloses a method for evaluating a high temperature heat wave based on remote sensing estimation and reconstruction of temperature, comprising: S1, estimating the daily maximum temperature of cloudless pixels based on remote sensing data, and obtaining the remote sensing climatic temperature by Fourier transform fitting; S2, obtaining the reanalysis climatic temperature by Fourier transform fitting based on the reanalysis temperature, and calculating the short-term fluctuation of the reanalysis temperature affected by the weather by combining the reanalysis temperature and the reanalysis climatic temperature; S3, combining the remote sensing climatic temperature in step S1 and the short-term fluctuation of the reanalysis temperature in step S2, interpolating the missing values ​​of cloud pixels caused by cloud coverage, and reconstructing and generating the all-weather daily maximum temperature with a resolution of 1 km; S4, quantitatively evaluating the intensity of the high temperature heat wave based on the remote sensing all-weather temperature. The present invention can accurately evaluate the intensity of the high temperature heat wave, and provide technical support for effectively responding to high temperature weather and protecting public health.

Description

Technical Field

[0001] The present invention belongs to the technical field of thermal infrared remote sensing, and in particular relates to a high temperature heat wave assessment method based on remote sensing estimation and reconstruction of air temperature. Background Art

[0002] Heat waves increase the risk of heat-related diseases with high mortality rates, such as heat stroke, aggravate the symptoms and incidence of underlying diseases, seriously threaten human life and health, and have a negative impact on people's quality of life and mental health. In addition, heat waves also lead to reduced crop yields, disruptions in transportation and energy networks, increased risks of forest fires, accelerated glacier melting, and threats to species diversity.

[0003] In order to effectively respond to the impact of heat waves, it is necessary to accurately assess the intensity of heat waves and understand their temporal and spatial changes. Meteorological stations are unevenly distributed in space, and the spatial representation range of observation data is limited, which cannot effectively reflect the spatial distribution characteristics of heat waves. Although meteorological reanalysis data has the advantage of spatial continuity, the spatial resolution is too low (usually less than 10km), making it difficult to reflect the spatial details of heat waves. Remote sensing surface temperature has a high spatial resolution, but compared with air temperature, the correlation between surface temperature and human health and comfort is relatively weak, making it difficult to accurately characterize the impact of heat waves on public health. Summary of the invention

[0004] In view of the technical problems that traditional meteorological station data can only reflect temperature characteristics at point scale, the spatial resolution of reanalysis data is too low to accurately reflect the spatial differences in temperature, and the remote sensing surface temperature has a weak correlation with human health, the present invention discloses a high temperature heat wave assessment method based on remote sensing estimation and reconstruction of temperature, which can accurately assess the intensity of high temperature heat waves and provide technical support for effectively responding to high temperature weather and protecting public health.

[0005] The present invention discloses a method for evaluating high temperature heat waves based on remote sensing estimation and reconstruction of air temperature, the method comprising the following steps:

[0006] S1, based on remote sensing data, the daily maximum temperature of cloud-free pixels is estimated, and the remote sensing climatological temperature is obtained by Fourier transform fitting;

[0007] S2, based on the reanalysis temperature, the reanalysis climatological temperature is obtained by Fourier transform fitting, and the short-term fluctuation of the reanalysis temperature affected by the weather is calculated by combining the reanalysis temperature and the reanalysis climatological temperature;

[0008] S3, combining the remote sensing climatological temperature in step S1 and the reanalysis temperature short-term fluctuation in step S2, interpolating the missing values ​​of cloud pixels caused by cloud cover, and reconstructing the all-weather daily maximum temperature with a resolution of 1 km;

[0009] S4, quantitative assessment of the intensity of heat waves based on remote sensing all-weather temperature.

[0010] As a preferred example, in step S1, the process of estimating the daily maximum temperature of cloud-free pixels based on remote sensing data includes the following steps:

[0011] The maximum temperature of the meteorological station observation day was taken as the dependent variable, and the surface temperature, vegetation index NDVI, water body index MNDWI, black sky albedo data, cumulative downlink shortwave radiation DSR, impervious surface coverage ISC, annual daily DOY and elevation value of the remote sensing pixels corresponding to the station were taken as spatial independent variables. Three machine learning methods, random forest, Cubist and XGBoost, were used to construct the random forest model, Cubist model and XGBoost model for temperature estimation respectively.

[0012] During the fitting process, the parameters of the random forest model, Cubist model and XGBoost model are debugged to optimize each model. The parameters to be optimized for the random forest model are n_estimators and max_features; the parameters to be optimized for the Cubist model are neighborhood, n_rules and n_committees; the parameters to be optimized for the XGBoost model are n_estimators, max_depth and learning_rate;

[0013] The accuracy of the random forest model, Cubist model, and XGBoost model was verified by the five-fold cross validation method, and the determination coefficient R was calculated. 2 , mean absolute error MAE and bias Bias to evaluate the model accuracy;

[0014] The model with the highest accuracy is applied to the spatial independent variable to estimate the daily maximum temperature in cloud-free pixels.

[0015] As a preferred example, in step S1, the remote sensing data acquisition process includes:

[0016] The daily daytime surface temperature with 1km resolution was extracted from the MODIS surface temperature product MOD11A1, and high-quality clear sky pixels were selected using the QC quality control band. The reflectance of the 8-day synthetic green band and shortwave infrared 1 band with 500m resolution was extracted from the MODIS reflectivity product MOD09A1, and the water body index MNDWI was obtained through band operation. The 16-day synthetic vegetation index NDVI with 250m resolution was extracted from the MODIS albedo product MCD43A3. The monthly synthetic black sky albedo data with 500m resolution was extracted from the MODIS albedo product MCD43A3. The impervious surface coverage ISC was calculated based on the CLCD_V01 land cover dataset. The 30m resolution elevation data was extracted from SRTMDEM. The daily maximum temperature and cumulative downward shortwave radiation DSR with 0.1° resolution were extracted from ERA5-Land. All spatial data were resampled to the daily scale and 1km spatial resolution, which was consistent with the MODIS surface temperature data. The calculation formula of the water body index MNDWI is:

[0017] ;

[0018] In the formula, is the green band reflectivity, It is the reflectivity of shortwave infrared 1 band.

[0019] As a preferred example, in step S1, the process of obtaining remote sensing climate temperature by Fourier transform fitting includes the following steps:

[0020] Based on the cloud-free pixel daily maximum temperature estimated by remote sensing, the remote sensing climatological temperature is obtained by fitting the Fourier transform pixel by pixel year by year:

[0021] ;

[0022] In the formula, The first Remote sensing climatological temperature of the day; is the remotely sensed annual mean temperature; is the number of harmonics of the fitted model; , and Harmonics The amplitude, frequency and phase of the cloud-free pixel are fitted by the least squares method using the remote sensing daily maximum temperature.

[0023] Step S2 further comprises:

[0024] The reanalysis climatological temperature is obtained by fitting the maximum temperature of the reanalysis day of cloud-free pixels extracted by ERA5-Land pixel by pixel and year by year using Fourier transform.

[0025] ;

[0026] In the formula, The first Reanalysis climatological temperature of the day; To reanalyze the annual mean temperature; is the number of harmonics of the fitted model; , and Harmonics The amplitude, frequency and phase of , whose values ​​are fitted by the reanalysis daily maximum temperature of cloud-free pixels using the least squares method;

[0027] Based on the difference between the ERA5-Land reanalysis daily maximum temperature and the fitted reanalysis climatological temperature, the short-term fluctuation of the reanalysis temperature affected by weather is obtained:

[0028] ;

[0029] In the formula, For the Reanalysis of short-term temperature fluctuations in the day, For the ERA5-Land reanalysis daily maximum temperature, The first Reanalysis climatological temperature for the day.

[0030] As a preferred example, in step S3, the following formula is used to calculate the maximum temperature of the cloud pixel day and interpolate the missing values ​​of the cloud pixel caused by cloud cover:

[0031] ;

[0032] In the formula, For the Remote sensing daily maximum temperature of cloud-covered pixels in the sky; The first Remote sensing climatological temperature of the day; For the Reanalysis of short-term temperature fluctuations.

[0033] Step S4 further comprises:

[0034] Based on the remote sensing all-weather temperature, the thermal index including the number of high temperature days, high temperature effective accumulated temperature, cumulative number of heat wave days and the longest duration of heat waves is calculated to quantitatively assess the intensity of high temperature heat waves.

[0035] The beneficial effects of the present invention are:

[0036] The high temperature heat wave assessment method based on remote sensing estimation and reconstruction of temperature of the present invention uses remote sensing data to reconstruct the high-resolution all-weather daily maximum temperature through the "cloud-free pixel estimation-cloud pixel interpolation" framework, and then calculates the heat index based on the remote sensing all-weather temperature to quantitatively assess the intensity of the high temperature heat wave. The temperature generated by remote sensing estimation and reconstruction of the present invention has the advantages of spatiotemporal continuity and high spatial resolution, and can more carefully and accurately assess the intensity of high temperature heat waves, providing technical support for coping with high temperature impacts and protecting public health. Practical application results show that the method proposed by the present invention can be effectively used for quantitative assessment and monitoring of high temperature heat waves. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of a high temperature heat wave assessment method based on remote sensing estimation and reconstruction of air temperature of the present invention;

[0038] Figure 2 Scatter plot comparing estimated temperature of cloud-free pixels with observed temperature of the station;

[0039] Figure 3 Scatter plot comparing interpolated temperature of cloud pixels with observed temperature of the station;

[0040] Figure 4 Schematic diagram of the temperature comparison before and after interpolation on January 4, April 12, July 21 and October 23, 2023 in the study area (white area is the ocean, and light gray area is the cloud pixel); (a) and (e) are the temperature schematic diagrams before and after interpolation on January 4, respectively; (b) and (f) are the temperature schematic diagrams before and after interpolation on April 12, respectively; (c) and (g) are the temperature schematic diagrams before and after interpolation on July 21, respectively; (d) and (h) are the temperature schematic diagrams before and after interpolation on October 23, respectively;

[0041] Figure 5 This is a schematic diagram of the spatial distribution of the number of high temperature days, high temperature effective accumulated temperature, cumulative number of heat wave days and the longest duration of heat waves in the study area in 2023 (the white area is the ocean);

[0042] Figure 6 A scatter plot comparing the heat index calculated based on remote sensing temperature and station temperature. DETAILED DESCRIPTION

[0043] The following examples will enable those skilled in the art to more fully understand the present invention, but are not intended to limit the present invention in any way.

[0044] The present invention discloses a method for evaluating high temperature heat waves based on remote sensing estimation and reconstruction of air temperature, the method comprising the following steps:

[0045] S1, based on remote sensing data, the daily maximum temperature of cloud-free pixels is estimated, and the remote sensing climatological temperature is obtained by Fourier transform fitting;

[0046] S2, based on the reanalysis temperature, the reanalysis climatological temperature is obtained by Fourier transform fitting, and the short-term fluctuation of the reanalysis temperature affected by the weather is calculated by combining the reanalysis temperature and the reanalysis climatological temperature;

[0047] S3, combining the remote sensing climatological temperature in step S1 and the reanalysis temperature short-term fluctuation in step S2, interpolating the missing values ​​of cloud pixels caused by cloud cover, and reconstructing the all-weather daily maximum temperature with a resolution of 1 km;

[0048] S4, quantitative assessment of the intensity of heat waves based on remote sensing all-weather temperature.

[0049] The high temperature heat wave assessment method based on remote sensing estimation and reconstruction of temperature in this embodiment specifically includes the following steps:

[0050] 1) Data processing

[0051] The daily daytime surface temperature with a resolution of 1 km was extracted from the MODIS surface temperature product MOD11A1, and high-quality clear sky pixels were selected using the QC quality control band. The reflectance of the 8-day synthetic green band and shortwave infrared 1 band with a resolution of 500 m was extracted from the MODIS reflectivity product MOD09A1, and the water body index MNDWI was obtained through band operation (Formula (1)). The 16-day synthetic vegetation index NDVI with a resolution of 250 m was extracted from the MODIS albedo product MCD43A3. The monthly synthetic black sky albedo data with a resolution of 500 m was extracted. The impervious surface coverage ISC was calculated based on the CLCD_V01 land cover dataset. The 30-m resolution elevation data was extracted from the SRTM DEM. The daily maximum temperature and the accumulated downward shortwave radiation DSR with a resolution of 0.1° were extracted from ERA5-Land. All spatial data were resampled to the daily scale and 1 km spatial resolution to be consistent with the MODIS surface temperature data. The MNDWI calculation formula is:

[0052] (1);

[0053] In the formula, is the green band reflectivity, It is the reflectivity of shortwave infrared 1 band.

[0054] 2) Temperature estimation of cloud-free pixels

[0055] The maximum temperature of the observation day at the meteorological station was used as the dependent variable, and the surface temperature, NDVI, MNDWI, albedo, DSR, ISC, annualized daily DOY and elevation value of the remote sensing pixels corresponding to the station were used as independent variables. Three machine learning methods, random forest (RF), Cubist and XGBoost, were used to build temperature estimation models. During the fitting process, two key parameters of the RF model (n_estimators and max_features), three key parameters of the Cubist model (neighborhood, n_rules and n_committees), and three key parameters of the XGBoost model (n_estimators, max_depth and learning_rate) were debugged to optimize the model. The model accuracy was verified by the five-fold cross validation method, and the determination coefficient R was calculated. 2 The model accuracy was evaluated by using the mean absolute error (MAE) and bias. The model with the highest accuracy was applied to the spatial independent variable to estimate the daily maximum temperature of the clear sky pixels.

[0056] 3) Temperature interpolation for cloud pixels

[0057] There are a large number of missing values ​​in remote sensing images due to cloud cover, which leads to a large number of missing values ​​in remote sensing estimated temperatures. In order to improve the spatiotemporal integrity of remote sensing temperature data, it is necessary to interpolate cloud pixels to obtain a reconstructed all-weather seamless temperature for evaluating high temperature heat waves.

[0058] Based on the daily maximum temperature of cloud-free pixels estimated by remote sensing, the time series Fourier analysis method is used to interpolate the daily maximum temperature of cloud-covered pixels. The time series Fourier analysis method decomposes the temperature time series within the year into climatic temperature and short-term temperature fluctuations affected by weather. The climatic temperature is fitted by Fourier transform for each pixel year by year based on the remote sensing estimated temperature and the reanalysis temperature in the clear sky phase:

[0059] (2);

[0060] In the formula, The first The climatic temperature of the day, is the annual average temperature, is the number of harmonics in the fitting model, which is 3 in this example. , and Harmonics The amplitude, frequency and phase of the clear sky phase are fitted by the least squares method using the actual air temperature (remote sensing temperature or reanalysis temperature).

[0061] Then, based on the reanalysis temperature and the fitted reanalysis climatological temperature, the short-term change of the reanalysis temperature (the difference between the actual temperature and the model-fitted climatological temperature) is calculated:

[0062] (3);

[0063] In the formula, To reanalyze the short-term changes in temperature, To reanalyze the temperature, is the reanalysis climatological temperature obtained by fitting.

[0064] The short-term fluctuations of remote sensing temperature and reanalysis temperature affected by weather such as clouds and rain are consistent. The real remote sensing temperature of cloud pixels is calculated based on the fitted remote sensing climatological temperature and reanalysis temperature short-term changes:

[0065] (4);

[0066] In the formula, is the remote sensing real temperature of the cloud pixel; is the remote sensing climatological temperature obtained by fitting; To reanalyze the short-term changes in temperature.

[0067] 4) Heat index calculation

[0068] The cloud-free pixel temperature estimated by remote sensing and the cloud pixel temperature interpolated were combined to obtain the reconstructed 1km resolution all-weather daily maximum temperature. Based on this data, four thermal indices were calculated, including the number of high temperature days, high temperature accumulated temperature, cumulative number of heat wave days, and the longest duration of heat waves, to quantitatively characterize the intensity of high temperature heat waves.

[0069] According to the definition of the China Meteorological Administration, a day with a maximum temperature above 35°C is a high temperature day, and three consecutive days of high temperature are considered a heat wave.

[0070] The number of high temperature days is the number of days with the highest daily temperature above 35°C each year (in d);

[0071] The effective high temperature accumulated temperature is the cumulative temperature (in °C) when the maximum daily temperature exceeds 35 °C each year. The calculation formula is as follows:

[0072] (5);

[0073] In the formula, is the high temperature effective accumulated temperature, For the The maximum daily temperature of the day.

[0074] The cumulative number of heat wave days is the cumulative sum of the number of days of all heat wave events each year (in d);

[0075] The maximum duration of a heat wave is the longest duration of a single heat wave among all heat wave events each year (in days).

[0076] Examples

[0077] The present invention uses high-resolution all-weather daily maximum temperature generated by remote sensing estimation and reconstruction to evaluate high temperature heat waves. The following are the specific implementation steps of the example. Figure 1 .

[0078] 1) Taking the Yangtze River Delta region of China as the study area, the 1km resolution daytime surface temperature was extracted from the MODIS surface temperature product MOD11A1, and high-quality clear sky pixels were extracted using the QC quality control band; the green band and shortwave infrared 1 band were extracted from the MODIS reflectance product MOD09A1, and the 500m resolution 8-day composite water index MNDWI was obtained through band operation (Formula (1)); the 250m resolution 16-day composite vegetation index NDVI was extracted from the MODIS albedo product MCD43A3; the 500m resolution monthly composite albedo data Albedo was extracted from the MODIS albedo product MCD43A3; the 1km resolution annual impervious surface coverage ISC was calculated based on the land cover product CLCD_V01; the daily 0.1° resolution solar downwelling radiation data DSR and daily maximum temperature were obtained from ERA5-Land; and the 30m resolution elevation data were extracted from SRTM / DEM. All spatial data were uniformly resampled to daily scale and 1km resolution.

[0079] 2) According to longitude, latitude and time, extract the surface temperature, NDVI, MNDWI, ISC, Albedo, DSR and elevation values ​​of the remote sensing pixels corresponding to the maximum temperature measured at the meteorological station. Taking the maximum temperature measured at the station as the dependent variable and the surface temperature, NDVI, MNDWI, ISC, Albedo, DSR, elevation value and DOY of the corresponding remote sensing pixels as independent variables, use RF, XGBoost and Cubists algorithms to build temperature estimation models respectively, and debug the parameter values ​​of each model during the model fitting process to optimize the model. The accuracy of each model is verified by the five-fold cross validation method, and the determination coefficient R is calculated. 2 , mean absolute error MAE and bias characterize the model accuracy. The results show that the accuracy of the Cubist model is better than the other two machine learning models. 2 , MAE and Bias are 0.99, 0.75℃ and 0.003℃ respectively (see Figure 2 ). Therefore, the Cubist model is selected to estimate the daily maximum temperature of clear sky pixels.

[0080] 3) The constructed Cubist model was applied to the surface temperature, NDVI, MNDWI, ISC, Albedo, DSR, elevation and DOY spatial variables to obtain the daily maximum temperature of clear sky pixels with a resolution of 1 km.

[0081] 4) Based on the daily maximum temperature in the clear sky phase estimated by remote sensing, the annual variation of the remote sensing estimated temperature is fitted pixel by pixel using formula (2) to calculate the remote sensing climatological temperature.

[0082] 5) Extract the ERA5-Land daily maximum temperature in the clear sky phase, and use formula (2) to fit and calculate the reanalysis climatological temperature for each pixel year by year.

[0083] 6) The short-term fluctuation of reanalysis temperature is calculated pixel by pixel year by year according to formula (3), and then the daily maximum temperature of cloud pixels is calculated by formula (4) in combination with remote sensing climatological temperature. Figure 3 The scatter plot of interpolated temperature and measured temperature at the station is given. The sample points are evenly distributed near the 1:1 line without obvious deviation. The determination coefficient R 2 , mean absolute error MAE and bias are 0.97, 1.38℃ and 0.71℃ respectively.

[0084] 7) Combining the daily maximum temperature of cloud-free pixels estimated by remote sensing and the daily maximum temperature of cloud-covered pixels obtained by interpolation, the reconstructed all-weather daily maximum temperature with a resolution of 1 km is obtained. Figure 4 The spatial distribution of the cloud-free pixel temperatures before interpolation and the reconstructed all-weather temperature on January 4, April 12, July 21, and October 23, 2023 are given. Figure 4 It can be seen that the reconstructed temperature can more accurately describe the spatial heterogeneity of temperature due to the influence of latitude, land and sea location, etc., as well as the changes caused by changes in factors such as seasons and underlying surface.

[0085] 8) Based on the 1km resolution remote sensing all-weather daily maximum temperature, the number of high temperature days, high temperature effective accumulated temperature, cumulative number of heat wave days and the longest duration of heat waves are calculated pixel by pixel to quantitatively characterize the intensity of high temperature heat waves. Figure 5 The spatial distribution of four thermal indices in the study area in 2023 is given. Figure 5 It can be seen that the heat index calculated from remote sensing temperature has obvious spatial heterogeneity. The heat wave is the strongest in the central region, with frequent high temperature weather and many extreme high temperatures. The heat wave in the urban area is stronger than that in the suburbs. Large cities such as Shanghai, Hangzhou, Nanjing and valley cities such as Quzhou are obvious heat wave centers. The heat index is calculated based on the highest temperature observed on the day of the station observation and compared with the heat index of the corresponding pixel of remote sensing to verify the reliability of the heat index based on remote sensing temperature. Figure 6A scatter plot comparing the thermal index based on remote sensing temperature and station temperature is given. The thermal index based on remote sensing temperature is highly correlated with the thermal index based on station temperature. The determination coefficient R 2 All of them are greater than 0.8, among which, the high temperature effective accumulated temperature has the highest correlation, and the determination coefficient R 2 The average absolute error (MAE) is 0.97, the average absolute error (MAE) is 4.48℃, and the bias is -1.10℃. The MAE of the number of high temperature days, the cumulative number of heat wave days, and the duration of heat waves is around 1~3d, and the bias is greater than 0. The sample points of each index are all concentrated near the 1:1 line, which shows the reliability of assessing high temperature heat waves based on remote sensing estimation and reconstruction of temperature.

[0086] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A method for assessing high temperature heat waves based on remote sensing estimation and reconstruction of air temperature, characterized in that: The method comprises the following steps: S1, based on remote sensing data, the daily maximum temperature of cloud-free pixels is estimated, and the remote sensing climatological temperature is obtained by Fourier transform fitting; S2, based on the reanalysis temperature, the reanalysis climatological temperature is obtained by Fourier transform fitting, and the short-term fluctuation of the reanalysis temperature affected by the weather is calculated by combining the reanalysis temperature and the reanalysis climatological temperature; S3, combining the remote sensing climatological temperature in step S1 and the reanalysis temperature short-term fluctuation in step S2, interpolating the missing values ​​of cloud pixels caused by cloud cover, and reconstructing the all-weather daily maximum temperature with a resolution of 1 km; S4, quantitative assessment of heat wave intensity based on remote sensing all-weather temperature; Step S2 further comprises: The reanalysis climatological temperature is obtained by fitting the maximum temperature of the reanalysis day of cloud-free pixels extracted by ERA5-Land pixel by pixel and year by year using Fourier transform. Where, T r ′ a (t) is the reanalysis climatological temperature on day t obtained by fitting; is the annual mean temperature of the reanalysis; N is the number of harmonics in the fitting model; a i 、f i and are the amplitude, frequency and phase of harmonic i, respectively, and their values ​​are fitted by the reanalysis daily maximum temperature of cloud-free pixels using the least squares method; Based on the difference between the ERA5-Land reanalysis daily maximum temperature and the fitted reanalysis climatological temperature, the short-term fluctuation of the reanalysis temperature affected by weather is obtained: ΔT ra (t)=T ra (t)-T r ′ a (t) In the formula, ΔT ra (t) is the short-term fluctuation of reanalysis temperature on day t, T ra (t) is the maximum daily temperature of the ERA5-Land reanalysis on day t, T r ′ a (t) is the reanalysis climatological temperature on day t obtained by fitting; In step S3, the following formula is used to calculate the maximum temperature of the cloud pixel on the day and interpolate the missing values ​​of the cloud pixel caused by cloud cover: T rs (t)=T r ′ s (t)+ΔT ra (t) Where, T rs (t) is the remote sensing maximum temperature of the cloud pixel on the tth day; T r ′ s (t) is the remote sensing climatological temperature of the tth day obtained by fitting; ΔT ra (t) is the short-term fluctuation of reanalysis temperature on day t; Step S4 further comprises: Based on the remote sensing all-weather temperature, the thermal index including the number of high temperature days, high temperature effective accumulated temperature, cumulative number of heat wave days and the longest duration of heat waves is calculated to quantitatively assess the intensity of high temperature heat waves.

2. The high temperature heat wave assessment method based on remote sensing estimation and reconstruction of air temperature according to claim 1 is characterized in that: In step S1, the process of estimating the daily maximum temperature of cloud-free pixels based on remote sensing data includes the following steps: The maximum temperature of the meteorological station observation day was taken as the dependent variable, and the surface temperature, vegetation index NDVI, water body index MNDWI, black sky albedo data, cumulative downlink shortwave radiation DSR, impervious surface coverage ISC, annual daily DOY and elevation value of the remote sensing pixels corresponding to the station were taken as spatial independent variables. Three machine learning methods, random forest, Cubist and XGBoost, were used to construct the random forest model, Cubist model and XGBoost model for temperature estimation respectively. During the fitting process, the parameters of the random forest model, Cubist model and XGBoost model are debugged to optimize each model. The parameters to be optimized for the random forest model are n_estimators and max_features; the parameters to be optimized for the Cubist model are neighborhood, n_rules and n_committees; the parameters to be optimized for the XGBoost model are n_estimators, max_depth and learning_rate; The accuracy of the random forest model, Cubist model, and XGBoost model was verified by the five-fold cross validation method, and the determination coefficient R was calculated. 2 , mean absolute error MAE and bias Bias to evaluate the model accuracy; The model with the highest accuracy is applied to the spatial independent variable to estimate the daily maximum temperature in cloud-free pixels.

3. The high temperature heat wave assessment method based on remote sensing estimation and reconstruction of air temperature according to claim 2 is characterized in that: In step S1, the remote sensing data acquisition process includes: The daily daytime surface temperature with 1km resolution was extracted from the MODIS surface temperature product MOD11A1, and high-quality clear sky pixels were selected using the QC quality control band. The reflectance of the 8-day synthetic green band and shortwave infrared 1 band with 500m resolution was extracted from the MODIS reflectivity product MOD09A1, and the water body index MNDWI was obtained through band operation. The 16-day synthetic vegetation index NDVI with 250m resolution was extracted from the MODIS albedo product MCD43A3. The monthly synthetic black sky albedo data with 500m resolution was extracted from the MODIS albedo product MCD43A3. The impervious surface coverage ISC was calculated based on the CLCD_V01 land cover dataset. The 30m resolution elevation data was extracted from the SRTM DEM. The daily maximum temperature and the accumulated downward shortwave radiation DSR with 0.1° resolution were extracted from ERA5-Land. All spatial data were resampled to the daily scale and 1km spatial resolution, which was consistent with the MODIS surface temperature data. The calculation formula of the water body index MNDWI is: In the formula, ρ Green is the green band reflectivity, ρ SWIR1 It is the reflectivity of shortwave infrared 1 band.

4. The high temperature heat wave assessment method based on remote sensing estimation and reconstruction of air temperature according to claim 1 is characterized in that: In step S1, the process of obtaining remote sensing climatic temperature by Fourier transform fitting includes the following steps: Based on the cloud-free pixel daily maximum temperature estimated by remote sensing, the remote sensing climatological temperature is obtained by fitting the Fourier transform pixel by pixel year by year: Where, T r ′ s (t) is the remote sensing climatological temperature on the tth day obtained by fitting; is the remote sensing annual average temperature; N is the number of harmonics in the fitting model; a i 、f i and are the amplitude, frequency and phase of harmonic i, respectively, and their values ​​are obtained by fitting the remote sensing daily maximum temperature of cloud-free pixels using the least squares method.

Images