Rice high-temperature heat damage monitoring method based on remote sensing data

By fusing remote sensing data from multiple platforms and using a random forest model, a high-precision daily average temperature remote sensing dataset was constructed, which solved the problem of low data accuracy in monitoring high-temperature heat damage in rice and enabled rapid and accurate monitoring of high-temperature heat damage in rice.

CN119027801APending Publication Date: 2024-11-26CHINA THREE GORGES UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202410849159.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-27
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Existing technologies suffer from low accuracy in estimating daily average temperatures due to the reliance on single data sources and traditional linear models, making it impossible to effectively monitor the extent and severity of high-temperature heat damage to rice.

Method used

By employing multi-platform remote sensing data fusion and a random forest model, a comprehensive and high-precision daily average temperature remote sensing dataset was constructed. Combined with the distribution of rice planting areas and key growth stages, a remote sensing monitoring framework for high-temperature heat damage in rice was established.

Benefits of technology

It enables rapid, timely, and accurate acquisition of the disaster area and severity of high-temperature heat damage in rice over a wide area, providing a high-precision method for monitoring high-temperature heat damage in rice and overcoming the geographical limitations of traditional methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027801A_ABST
    Figure CN119027801A_ABST
Patent Text Reader

Abstract

A rice high-temperature heat damage monitoring method based on remote sensing data comprises the following steps: step 1, performing remote sensing inversion on daily average temperature to obtain a daily average temperature remote sensing data set of a target time period of a whole research area; 2, carrying out special processing on a daily average temperature remote sensing data set obtained by remote sensing inversion, and constructing a daily average temperature remote sensing data set with full coverage and high precision of a research area; and step 3, combining the full-coverage high-precision daily average temperature remote sensing data set to carry out full-flow rice high-temperature heat damage remote sensing monitoring.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of computer technology, and in particular to a rice high temperature heat damage monitoring method based on remote sensing data. Background Art

[0002] In the agricultural field, monitoring heat damage to rice is crucial to ensuring food production security. Affected by global climate change, continuous high temperatures have caused large-scale heat damage in crop-growing areas, reducing rice production and affecting my country's food production security. Therefore, monitoring the scope of heat damage to rice and assessing the degree of damage are of great significance for reducing the impact of disasters and ensuring the safety of rice food production.

[0003] The key prerequisite for monitoring the harm of high temperature and heat damage to rice is to estimate the daily average temperature of the whole coverage, extract the area of ​​rice planting area and determine the key growth period of rice. Some studies have shown that the land surface temperature (LST) product of Moderate Resolution Imaging Spectroradiometer (MODIS) can be used to estimate the daily average temperature. In the article "Estimation of daily mean air temperature from MODIS LST inAlpine areas", the authors demonstrated the feasibility of using MODIS-LST products to estimate the daily average temperature in the Italian Alps. In the article "An effective interpolation method for MODIS land surface temperature on the Qinghai–Tibet Plateau", the authors found that within a certain time and space area, the temperature changes with time are similar. The temperature values ​​of adjacent pixels can be calculated from the temperature of known pixels through a certain conversion method, so as to perform spatial interpolation of LST. The above methods all estimate the daily average temperature through LST, and the data source is single. Summary of the invention

[0004] The purpose of the present invention is to solve the technical problems of low LST estimation accuracy caused by a single data source and traditional linear model in the prior art, and to provide a method for integrating multi-platform data and using a machine learning model to construct a full-coverage, high-precision daily average temperature remote sensing dataset, and to build a rice high temperature heat damage remote sensing monitoring framework in combination with the distribution of rice-growing areas and the identification of the key growth period of rice, and to propose a rice high temperature heat damage monitoring method based on remote sensing data.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A method for monitoring high temperature heat damage to rice based on remote sensing data comprises the following steps:

[0007] Step 1: Perform remote sensing inversion of the daily average temperature to obtain the daily average temperature remote sensing data set for the target time period of the entire study area;

[0008] Step 2: Specially process the daily average temperature remote sensing dataset obtained by remote sensing inversion to construct a high-precision daily average temperature remote sensing dataset covering the entire study area;

[0009] Step 3: Combined with the full coverage high-precision daily average temperature remote sensing dataset, conduct full-process remote sensing monitoring of rice high temperature heat damage.

[0010] In step 1, remote sensing inversion is a method that combines remote sensing data from multiple sources and a random forest (RF) model to obtain the daily average temperature in the target time period of the study area;

[0011] In step 2, the special processing includes three steps: data fusion, spatial interpolation, and conditional time series interpolation;

[0012] In step 3, the rice heat damage level index is selected, and the high-precision daily average temperature remote sensing data set covering the entire study area obtained in step 2 is used to monitor the occurrence of rice heat damage in the study area.

[0013] Step 1 specifically includes the following sub-steps:

[0014] Step 1-1: Obtain some remote sensing data required for the study;

[0015] Step 1-2: Divide the land surface temperature (LST) remote sensing dataset collected in step 1-1 in a specific designed way to improve the accuracy of remote sensing inversion;

[0016] Step 1-3: Screen the effective values ​​of the LST remote sensing data sets from different sources obtained after the division;

[0017] Step 1-4: Combine the screened LST remote sensing dataset and the meteorological station data (T station ), digital elevation data (DEM), two-band enhanced vegetation index (EVI2), longitude (Lon) and latitude (Lat) combined with RF model, several regression models were trained;

[0018] Step 1-5: According to the trained regression model, the daily average temperature remote sensing dataset T of the target time period in the entire study area is inverted. mean :T mean=f(LST,DEM,EVI2,Lon,Lat), and the root mean square error RMSE index is used to verify the inversion accuracy R 2 .

[0019] In step 2, the following steps are used to process the daily average temperature remote sensing dataset obtained by remote sensing inversion:

[0020] Step 2-1: Perform multi-platform data fusion on the daily average temperature remote sensing datasets from different sources (different satellite sources and different acquisition times);

[0021] Step 2-2: Perform spatial interpolation on the fused daily average temperature remote sensing image;

[0022] Step 2-3: For the few pixels that are still empty after multi-platform fusion and spatial interpolation, the Conditional Temporal Interpolation Filtering (CTIF) (GROTEN, 1993) algorithm is used to interpolate all daily average temperature remote sensing images to achieve the goal of increasing the effective pixel coverage of all daily average temperature remote sensing images;

[0023] Step 2-4: Use the temperature data observed by the meteorological stations collected in step 1-1 to verify the validity of the daily average temperature remote sensing image results interpolated in step 2-3. The verification evaluation indicators are correlation coefficient R, root mean square error RMSE and mean absolute error MAE;

[0024] Step 2-5: Estimate and obtain a full coverage, high-precision daily average temperature remote sensing dataset.

[0025] In step 3, the steps for monitoring rice heat damage are as follows:

[0026] Step 3-1: According to the provisions of GB / T 21985-2008, the national standard of the People's Republic of China, "Temperature Indicators of High Temperature Hazards of Major Crops", issued on June 3, 2008, select the index of occurrence level of high temperature heat damage to rice;

[0027] Step 3-2: Combined with the full coverage high-precision daily average temperature remote sensing dataset obtained in step 2 and the previous research results of our team: rice planting area distribution data and rice key growth period data (Liu et al., 2020), estimate the area proportion of rice high temperature heat damage and conduct disaster level assessment to produce a disaster distribution map;

[0028] Step 3-3: Obtain the occurrence ratio table of high temperature heat damage to rice over many years;

[0029] Step 3-4: Obtain the distribution map of rice heat damage levels over multiple years.

[0030] In step 1-2, the steps of data set division are as follows:

[0031] Step 1-2-1: First, divide the two different satellite sensor sources of Terra and Aqua;

[0032] Step 1-2-2: Then divide the data collection time into two different time periods: day and night;

[0033] Step 1-2-3: Press four different seasons [Spring (S 1 ) is March, April, and May; Summer (S 2 ) is June, July, and August; Autumn (A) is September, October, and November; Winter (W) is December, January, and February]:

[0034]

[0035] Access remote sensing data from multiple sources.

[0036] In step 2-1, the steps of multi-platform data fusion are as follows:

[0037] Step 1) For each data source (different satellite sources and different acquisition times) in steps 1-5, the inversion accuracy of the four seasons is averaged;

[0038] Step 2) sort the accuracy results;

[0039] Step 3) In order of accuracy from high to low, the temperature inversion results of data sources with higher accuracy are retained first.

[0040] The step 3) comprises the following sub-steps:

[0041] (1) For the daily average temperature inversion image of any day in the target time period, for any pixel X in the image, the daily average temperature value inverted by the data with higher accuracy is given priority;

[0042] (2) If the pixel value on the data image with higher accuracy is null (caused by the lack of pixel value in the original image), then the daily average temperature value inverted from the data with lower accuracy will be considered;

[0043] (3) Repeat step (2) to fill empty pixels;

[0044] (4) For pixels that are still empty after step (3), assign pixel X a null value.

[0045] In step 2-2, the steps of spatial interpolation of daily average temperature image are as follows:

[0046] Step 2-2-1: Set the daily average temperature images with all effective pixel coverage greater than the specified value B in the study area as the reference image T ref , the image smaller than the specified value B is set as the image to be interpolated T interp , where B is the specified percentage;

[0047] Step 2-2-2: Create pixels with values ​​in the image to be interpolated The reference image pixel at the same location Regression relationship between DEM pixels and EVI2 pixels:

[0048] , where α 1 , α 2 , α 3 , α 4 represents the regression coefficient;

[0049] Step 2-2-3: Use the regression coefficient, reference image pixel array, DEM pixel array and EVI2 pixel array to infer the missing pixels on the image to be interpolated at the same location Values:

[0050]

[0051] Compared with the prior art, the present invention has the following technical effects:

[0052] 1) This paper innovatively combines multi-platform remote sensing data and random forest models, and proposes a method for multi-platform data fusion and spatial interpolation of daily average temperature images; a high-precision 1 km daily temperature remote sensing dataset covering the entire study area is constructed, providing a reliable dataset for rice high temperature heat damage monitoring;

[0053] 2) Based on the high-precision temperature remote sensing data set, the present invention combines rice high temperature heat damage indicators, rice planting area distribution and identified key rice growth period to construct a full-process rice high temperature heat damage remote sensing monitoring process framework. Remote sensing technology can be used to quickly and timely obtain ground feature data in a large area. With the help of this method, the disaster area and degree of rice high temperature heat damage can be quickly, efficiently and accurately obtained.

[0054] 3) Compared with other research methods on high temperature heat damage to rice, traditional research methods based on agronomy have regional limitations, and high temperature heat damage to rice in the field of agronomy focuses more on the mechanism research and variety breeding of high temperature heat damage to rice. However, the high temperature heat damage to rice based on meteorology in the present invention focuses more on the research on the occurrence law of high temperature heat damage to rice, risk assessment, loss assessment, etc., builds a remote sensing monitoring framework for high temperature heat damage to rice, provides a high temperature heat damage monitoring method for rice based on remote sensing data, and realizes more accurate monitoring of high temperature heat damage to rice. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0056] Figure 1 It is the research technology roadmap of the present invention;

[0057] Figure 2 This is a verification diagram of the remote sensing estimation accuracy of the daily average temperature dataset of the present invention. DETAILED DESCRIPTION

[0058] A method for monitoring high temperature heat damage to rice based on remote sensing data comprises the following steps:

[0059] Step 1: After preprocessing, the data is input into the random forest model to perform remote sensing inversion of the daily average temperature;

[0060] Step 2: Process the daily average temperature dataset obtained by remote sensing inversion to obtain a full coverage and high-precision daily average temperature remote sensing dataset;

[0061] Step 3: Select rice heat damage occurrence level indicators to monitor rice heat damage, estimate the area proportion of rice heat damage and conduct disaster level assessment based on the full coverage and high-precision daily average temperature remote sensing data set, and produce a disaster distribution map.

[0062] In step 1, the specific steps are as follows:

[0063] Step 1-1: Download various data required for the study;

[0064] Step 1-2: Divide the land surface temperature (LST) dataset collected in step 1-1 in a certain way to improve the inversion accuracy;

[0065] Step 1-3: Screen the divided data set for valid values;

[0066] Step 1-4: Combine the filtered data and the weather station data collected in step 1-1 (T station ), digital elevation data (DEM), two-band enhanced vegetation index (EVI2), longitude (Lon) and latitude (Lat) are input into the random forest model for training, and a total of 16 regression models can be obtained, and the models are verified using validation samples;

[0067] Step 1-5: Use the 16 trained random forest models to invert the daily average temperature T of the entire study area from 2003 to 2019. mean :T mean =f(LST,DEM,EVI2,Lon,Lat).

[0068] In step 1-1, the steps to download various data are as follows:

[0069] (1) Download the 1km spatial resolution daily land surface temperature (LST) dataset and the 500m spatial resolution eight-day synthetic reflectance dataset (Enhanced Vegetation Index (EVI) dataset) from 2003 to 2019 from the Distributed Archive Center of the Earth Science Data and Information Services Division of the Goddard Space Flight Center of NASA;

[0070] (2) Download meteorological data from the China Meteorological Science Data Sharing Network, including daily average temperature and latitude and longitude data sets of different regional stations

[0071] (3) Download the 90m spatial resolution digital elevation model from the Shuttle Radar Topography Mission Platform (Digital

[0072] Elevation Model, DEM) dataset;

[0073] (4) Download administrative division data from the National Basic Geographic Information System Center and produce a land use type distribution map of the study area.

[0074] In step 1-2, the LST dataset is divided as follows:

[0075] (1) Classify the two different satellite sensor sources, Terra and Aqua;

[0076] (2) Divide the data collection time into two different time periods: daytime and nighttime;

[0077] (3) Divided into four different seasons (spring is March, April, and May; summer is June, July, and August; autumn is September, October, and November; winter is December, January, and February):

[0078]

[0079] In steps 1-4: the steps for data entry are as follows:

[0080] (1) The 1 km daily LST data, 1 km 8-day enhanced two-band vegetation index (EVI2) data (using the EVI2 with the closest time distance to match LST), DEM data, longitude and latitude data of the entire study area from 2005 to 2016 were used as the independent variable input of the model, and the daily average temperature monitored by the meteorological station was used as the dependent variable input of the model;

[0081] (2) 7 / 10 of the data is used as model training samples, and the remaining 3 / 10 of the data is used as validation samples.

[0082] In step 2, the following steps are used to process the daily average temperature dataset obtained by remote sensing inversion:

[0083] Step 2-1: Perform multi-platform data fusion. For temperature inversion results from different sources (different satellite sources and different acquisition times), the temperature inversion results with higher accuracy are retained first;

[0084] Step 2-2: Perform spatial interpolation on the daily average temperature image;

[0085] Step 2-3: Use the Conditional Temporal Interpolation Filtering (CTIF) (GROTEN, 1993) algorithm to interpolate all temperature images to increase the effective pixel coverage of all images;

[0086] Step 2-4: Verify the interpolated results using weather station data.

[0087] In step 2-1, the steps of multi-platform data fusion are as follows:

[0088] (1) The inversion accuracy of the four seasons in each data source (different satellite sources and different acquisition times) is averaged;

[0089] (2) Sort the accuracy results;

[0090] (3) In order of accuracy from high to low, the temperature inversion results from data sources with higher accuracy are retained first.

[0091] The operation steps of step (3) of step 2-1 are as follows:

[0092] 1) For the daily average temperature inversion image of any day from 2003 to 2019, for any pixel X in the image, the temperature value inverted by the data with higher accuracy is given priority;

[0093] 2) If the pixel value on the data image with higher accuracy is null (caused by the lack of pixel value in the original image), the temperature value inverted from the data with lower accuracy will be considered;

[0094] 3) Repeat step 2) to fill empty pixels;

[0095] 4) For pixels that are still empty after fusion, assign pixel X a null value.

[0096] In step 2-2, the steps of spatial interpolation of daily average temperature image are as follows:

[0097] (1) The daily average temperature image with all effective pixel coverage greater than 80% in the study area is set as the reference image T ref , the image with less than 80% is set as the image to be interpolated T interp ;

[0098] (2) Establishing the valued pixels of the image to be interpolated The reference image pixel at the same location Regression relationship between DEM pixels and EVI2 pixels:

[0099]

[0100] In the formula, α 1 , α 2 , α 3 , α 4 represents the regression coefficient.

[0101] (3) Through the regression coefficient, reference image pixel array, DEM pixel array and EVI2 pixel array, the missing pixels on the image to be interpolated at the same location are deduced Values:

[0102]

[0103] In steps 2-4, the steps for verification using weather station data are as follows:

[0104] (1) Use the correlation coefficient R to test the meteorological station data and interpolation results:

[0105]

[0106] In the formula: yi and represent the actual reference value and estimated value of pixel i respectively, and represents the mean of the measured sample and the mean of the estimated sample, and n represents the sample size.

[0107] (2) Use the root mean square error (RMSE) to test the meteorological station data and interpolation results:

[0108]

[0109] Where: y i and represent the actual reference value and estimated value of pixel i respectively, and represents the mean of the measured sample and the mean of the estimated sample, and n represents the sample size.

[0110] Step 3: The steps for monitoring rice heat damage are as follows:

[0111] Step 3-1: Select the rice high temperature heat damage occurrence index to monitor the high temperature heat damage occurrence in the study area;

[0112] Step 3-2: Estimate the area proportion of rice heat damage and conduct disaster level assessment, and make a disaster distribution map.

[0113] In step 3-1, the specific indicators of rice heat damage level are:

[0114] According to the provisions of GB / T 21985-2008, the national standard of the People's Republic of China, "Temperature Indicators of High Temperature Hazards for Major Crops", issued on June 3, 2008: During the heading and flowering period of rice, if the meteorological process with a daily average temperature T_mean ≥ 30°C occurs for three consecutive days, it is considered that rice heat damage has occurred. If the rice heat damage occurs for 3-4 consecutive days, it is considered to be a mild disaster; if it lasts for 5-7 days, it is considered to be a moderate disaster; if the disaster lasts for more than 8 days (including 8 days), the rice heat damage level is a severe disaster (a total of 15 days).

[0115] Example

[0116] A rice heat damage monitoring method based on remote sensing data, such as Figure 1 As shown, the following steps are included:

[0117] Step 1: Input the data into the random forest model to perform remote sensing inversion of the daily average temperature.

[0118] Table 1 shows the dataset information.

[0119] Table 1 Main research data sources and related information

[0120]

[0121] The data input and daily mean temperature remote sensing inversion are:

[0122] Step 1-1: Use the 1 km daily LST data, 1 km 8-day Enhanced Vegetation Index (EVI2) data (matching LST with the EVI2 with the closest time distance) of the entire study area (Hubei, Hunan, Jiangxi, Jiangsu and Anhui provinces) from 2005 to 2016 as the independent variable input of the model, and the daily average temperature monitored by the meteorological station as the dependent variable input of the model.

[0123] 7 / 10 of the data are used as model training samples, and the remaining 3 / 10 of the data are used as validation samples.

[0124] Step 1-2: Data from two different satellite sensors (Terra / Aqua), collected in two different time periods (daytime and nighttime), and four different seasons (spring is March, April, and May; summer is June, July, and August; autumn is September, October, and November; and winter is December, January, and February) are input into the model for training. A total of 16 random forest regression models are obtained, and the models are validated using validation samples.

[0125] The 16 trained random forest models were used to invert the temperature values ​​of the entire study area from 2003 to 2019.

[0126] Table 2 shows the accuracy verification results of each random forest model. Except for the daytime data in summer, the R 2 The accuracy of the models varies greatly from 0.75 to 0.95, and the RMSE ranges from 1.68 to 3.07℃. The accuracy of different models varies greatly. In general, the accuracy of the night model in the study area is higher than that of the day model; in addition, the accuracy of the autumn model is the highest, and the accuracy of the summer model is the lowest.

[0127] Table 2 Accuracy verification results of random forest models trained based on 16 different sources and seasonal data

[0128]

[0129]

[0130] Step 2: Screen, fuse and interpolate the 16 inverted daily average temperature values ​​from 2003 to 2019. The multi-platform data fusion and daily average temperature image spatial interpolation and interpolation are as follows:

[0131] Step 2-1: Take the average of the model verification accuracy of the four seasons in each data source, and then sort the accuracy results. Assign different priorities to data from different sources in order of accuracy from high to low. That is, for the daily average temperature inversion image of any day from 2003 to 2019, for any pixel X in the image, the temperature value inverted from the data with the highest accuracy is given priority; if the pixel value on the data image with the highest accuracy is null (caused by the lack of pixel value in the original image), then consider the temperature value inverted from the data with the second highest accuracy, and so on; if the pixel values ​​of the above four are all null, pixel X is finally assigned a null value.

[0132] Step 2-2: Set the daily average temperature images with all effective pixel coverage greater than 80% in the study area as the reference image T ref , the image with less than 80% is set as the image to be interpolated T interp .

[0133] (1) Establish the valued pixels of the image to be interpolated (T interp_valid ) and the reference image pixel at the same location Regression relationship between DEM pixels and EVI2 pixels:

[0134]

[0135] In the formula, α 1 , α 2 , α 3 , α 4 represents the regression coefficient.

[0136] (2) After the regression equation is established, the missing pixels on the image to be interpolated at the same position can be deduced through the regression coefficient, reference image pixel array, DEM pixel array and EVI2 pixel array. Values:

[0137]

[0138] Step 2-3: For the few pixels that are still empty after the above platform fusion and interpolation operations, the Conditional Temporal Interpolation Filtering (CTIF) (GROTEN, 1993) algorithm is used to interpolate all temperature images, and finally the goal of increasing the effective pixel coverage of all images is achieved. The correlation coefficient R and root mean square error RMSE are used to verify the results:

[0139]

[0140] In the formula: yi and represent the actual reference value and estimated value of pixel i respectively, and represents the mean of the measured sample and the mean of the estimated sample, and n represents the sample size.

[0141] The meteorological station data are used to verify the results after temperature inversion, fusion and interpolation, such as Figure 2 As shown. The correlation coefficient R between the daily average temperature observed by the meteorological station in all years and the daily average temperature inverted according to the random forest model is ≥0.93, and the correlation between the two is good. The root mean square error RMSE of the estimated daily average temperature in all years is ≤3.43℃, and the mean absolute error MAE is ≤2.48℃. The root mean square error and mean absolute error of the estimated daily average temperature in all years are small, indicating that the temperature data estimated by remote sensing based on MO[Y]D11A1 land surface temperature data can be used for remote sensing monitoring of high temperature heat damage to rice.

[0142] Step 3: According to the provisions of GB / T 21985-2008, the National Standard of the People's Republic of China, "Temperature Indicators of High Temperature Hazards for Major Crops", issued on June 3, 2008, the rice high temperature heat damage level index was selected to monitor the occurrence of high temperature heat damage in the first-season rice in the study area. During the heading and flowering period of rice, if the meteorological process with a daily average temperature T_mean ≥ 30°C occurs for three consecutive days, it is considered that rice high temperature heat damage has occurred. If the rice high temperature heat damage occurs for 3-4 consecutive days, it is a mild disaster; if it lasts for 5-7 days, it is a moderate disaster; if the disaster lasts for more than 8 days (including 8 days), the rice high temperature heat damage level is a severe disaster (a total of 15 days). The rice distribution dataset and the rice key growth period dataset used are from the datasets obtained based on the previous research of our team.

[0143] The occurrence level of rice heat damage was monitored according to the duration of rice heat damage. Table 3 shows the proportion of rice heat damage in the study area. A large area of ​​rice heat damage was monitored in 2013, and the degree of damage was higher than that in other years. The proportion of slightly affected areas in the study area was 38.78%, the proportion of moderately affected areas was 19.8%, and the proportion of severely affected areas was 6.27%, and the overall damage was serious. The scope of rice heat damage in 2006 was smaller than that in 2013, and the overall damage in the study area was relatively serious.

[0144] Table 3 Proportion of rice affected by high temperature heat damage in the study area

[0145]

[0146] In general, the method proposed in the present invention can use remote sensing and artificial intelligence technology to quickly and timely obtain ground feature data in a large area, and with the help of this method, the disaster area and degree of rice high temperature heat damage can be quickly, efficiently and accurately obtained.

Claims

1. A method for monitoring high temperature heat damage to rice based on remote sensing data, characterized in that: The following steps are involved: Step 1: Perform remote sensing inversion of the daily average temperature to obtain the daily average temperature remote sensing data set for the target time period of the entire study area; Step 2: Specially process the daily average temperature remote sensing dataset obtained by remote sensing inversion to construct a high-precision daily average temperature remote sensing dataset covering the entire study area; Step 3: Combined with the full coverage high-precision daily average temperature remote sensing dataset, conduct full-process remote sensing monitoring of rice high temperature heat damage.

2. The method according to claim 1, characterized in that In step 1, remote sensing inversion is a method that combines remote sensing data from multiple sources and the random forest RF model to obtain the daily average temperature in the target time period of the study area; In step 2, the special processing includes three steps: data fusion, spatial interpolation, and conditional time series interpolation; In step 3, the rice heat damage level index is selected, and the high-precision daily average temperature remote sensing data set covering the entire study area obtained in step 2 is used to monitor the occurrence of rice heat damage in the study area.

3. The method according to claim 2, characterized in that Step 1 specifically includes the following sub-steps: Step 1-1: Obtain some remote sensing data required for the study; Step 1-2: Divide the land surface temperature (LST) remote sensing dataset collected in step 1-1 in a specific designed way to improve the accuracy of remote sensing inversion; Step 1-3: Screen the effective values ​​of the LST remote sensing data sets from different sources obtained after the division; Step 1-4: Combine the screened LST remote sensing dataset and the meteorological station data collected in step 1-1 station , digital elevation data DEM, two-band enhanced vegetation index EVI2, longitude Lon and latitude Lat combined with RF model, trained to obtain several regression models; Step 1-5: According to the trained regression model, the daily average temperature remote sensing dataset T of the target time period in the entire study area is inverted. mean :Tmean=f(LST,DEM,EVI2,Lon,Lat), and use the root mean square error RMSE indicator to verify the inversion accuracy R 2 .

4. The method according to claim 2, characterized in that: In step 2, the following steps are used to process the daily average temperature remote sensing dataset obtained by remote sensing inversion: Step 2-1: Perform multi-platform data fusion on daily average temperature remote sensing datasets from different sources; Step 2-2: Perform spatial interpolation on the fused daily average temperature remote sensing image; Step 2-3: For the few pixels that are still null after multi-platform fusion and spatial interpolation, use the conditional time series interpolation algorithm CTIF to interpolate all daily average temperature remote sensing images to achieve the goal of increasing the effective pixel coverage of all daily average temperature remote sensing images; Step 2-4: Use the temperature data observed by the meteorological stations collected in step 1-1 to verify the validity of the daily average temperature remote sensing image results interpolated in step 2-3. The verification evaluation indicators are correlation coefficient R, root mean square error RMSE and mean absolute error MAE; Step 2-5: Estimate and obtain a full coverage, high-precision daily average temperature remote sensing dataset.

5. The method according to any one of claims 1 to 4, characterized in that: In step 3, the steps for monitoring rice heat damage are as follows: Step 3-1: Select the level index of high temperature heat damage to rice; Step 3-2: Combine the full coverage high-precision daily average temperature remote sensing data set obtained in step 2 with the rice planting area distribution data and rice key growth period data to estimate the area proportion of rice high temperature heat damage and conduct disaster level assessment to produce a disaster distribution map; Step 3-3: Obtain the occurrence ratio table of high temperature heat damage to rice over many years; Step 3-4: Obtain the distribution map of rice heat damage levels over multiple years.

6. The method according to claim 3, characterized in that In step 1-2, the steps of data set division are as follows: Step 1-2-1: First, divide the two different satellite sensor sources of Terra and Aqua; Step 1-2-2: Then divide the data collection time periods into two different time periods: day and night; Step 1-2-3: Divide into four different seasons [Spring S1 is March, April, and May; Summer S2 is June, July, and August; Autumn A is September, October, and November; Winter W is December, January, and February]: i∈(Terra_day,Terra_night,Aqua_day,Aqua_night),j∈(S1,S2,A,W), Access remote sensing data from multiple sources.

7. The method according to claim 2, 3 or 4, characterized in that: In step 2-1, the steps of multi-platform data fusion are as follows: Step 1) Take the average of the inversion accuracy of the four seasons for each data source in steps 1-5; Step 2) sort the accuracy results; Step 3) In order of accuracy from high to low, the temperature inversion results of data sources with higher accuracy are retained first.

8. The method according to claim 7, characterized in that The step 3) comprises the following sub-steps: (1) For the daily average temperature inversion image of any day in the target time period, for any pixel X in the image, the daily average temperature value inverted by the data with higher accuracy is given priority; (2) If the pixel value on the data image with higher accuracy is null, then the daily average temperature value inverted from the data with lower accuracy will be considered; (3) Repeat step (2) to fill empty pixels; (4) For pixels that are still empty after step (3), assign pixel X a null value.

9. The method according to claim 4, characterized in that In step 2-2, the steps of spatial interpolation of daily average temperature image are as follows: Step 2-2-1: Set the daily average temperature images with all effective pixel coverage greater than the specified value B in the study area as the reference image Tref, and the images with less than the specified value B as the interpolation image Tinterp, where B is the specified percentage; Step 2-2-2: Create pixels with values ​​in the image to be interpolated The reference image pixel at the same location Regression relationship between DEM pixels and EVI2 pixels: Where α1, α2, α3, and α4 represent regression coefficients; Step 2-2-3: Use the regression coefficient, reference image pixel array, DEM pixel array and EVI2 pixel array to infer the missing pixels on the image to be interpolated at the same location Values:

Citation Information

Patent Citations

  • Assessment method for high-temperature heat damage risk of corn

    CN107392503A

  • Rice high temperature and heat damage remote sensing monitoring method of middle and lower reaches of Yangtze River based on satellite-ground multi-source data

    CN108961089A