CN116050567B - Space-time degradation scale change analysis method for urban thermal environment - Google Patents
Space-time degradation scale change analysis method for urban thermal environment Download PDFInfo
- Publication number
- CN116050567B CN116050567B CN202211365577.3A CN202211365577A CN116050567B CN 116050567 B CN116050567 B CN 116050567B CN 202211365577 A CN202211365577 A CN 202211365577A CN 116050567 B CN116050567 B CN 116050567B
- Authority
- CN
- China
- Prior art keywords
- data
- surface temperature
- resolution
- model
- space
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000004458 analytical method Methods 0.000 title claims abstract description 24
- 230000008859 change Effects 0.000 title claims abstract description 10
- 230000015556 catabolic process Effects 0.000 title claims description 9
- 238000006731 degradation reaction Methods 0.000 title claims description 9
- 230000000694 effects Effects 0.000 claims abstract description 15
- 230000005855 radiation Effects 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims description 42
- 230000004913 activation Effects 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 21
- 238000000034 method Methods 0.000 claims description 20
- 210000002569 neuron Anatomy 0.000 claims description 18
- 230000002159 abnormal effect Effects 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 11
- 238000012952 Resampling Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 6
- 238000001556 precipitation Methods 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 5
- 238000006467 substitution reaction Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000005520 cutting process Methods 0.000 claims description 3
- 238000013079 data visualisation Methods 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000013075 data extraction Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000012800 visualization Methods 0.000 claims 1
- 238000011160 research Methods 0.000 abstract description 15
- 238000013527 convolutional neural network Methods 0.000 abstract description 4
- 230000036541 health Effects 0.000 abstract description 4
- 230000009467 reduction Effects 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 abstract description 3
- 230000015654 memory Effects 0.000 abstract description 3
- 238000007670 refining Methods 0.000 abstract description 3
- 238000003062 neural network model Methods 0.000 abstract description 2
- 241000282414 Homo sapiens Species 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000001932 seasonal effect Effects 0.000 description 2
- 235000001674 Agaricus brunnescens Nutrition 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000012368 scale-down model Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A30/00—Adapting or protecting infrastructure or their operation
- Y02A30/60—Planning or developing urban green infrastructure
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Operations Research (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Game Theory and Decision Science (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Primary Health Care (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention discloses a space-time downscaling change analysis method for urban thermal environment, which uses a convolutional neural network and a long-short-term memory neural network model to carry out the time scale of large-area earth surface temperature reduction, then uses the space-scale earth surface temperature reduction research of a multi-layer perceptron model on the basis of several times, determines the closely related influence factors of earth surface temperature, such as normalized vegetation index, solar radiation, relative humidity and the like, through consulting documents and related data, obtains high-molecular satellite and multi-load remote sensing image data of the earth surface temperature and related influence factors, and carries out the space-time downscaling aiming at MODIS earth surface temperature data products by the research main, and enhances the accuracy of the model through the abstract relation between the factors and the earth surface temperature. And (3) outputting high-resolution data, and finally refining the obtained downscaled data to obtain a final urban thermal environment downscaled analysis effect diagram, so as to realize the effect of urban thermal environment analysis and research and assist government decisions and urban resident life health.
Description
Technical Field
The invention relates to the fields of thermal infrared remote sensing and environmental monitoring, in particular to a space-time degradation scale change analysis method of urban thermal environment.
Background
Urban thermal environment refers to the integrated manifestation of urban spatial environment in a thermal force field. Specifically, the system takes the ground surface temperature and the air temperature of the urban under-pad surface as cores, and takes the transmission atmosphere condition, the under-pad surface condition and the solar radiation which are changed under the influence of human activities as a physical environment system which can influence human beings and activities thereof.
The heat island effect is used as main content in the urban space heat environment research and management in the current global urban area scale range, the heat island research is limited by the time and space resolution of data, the urban surface temperature is closely related to the health and cold and warm feeling of human bodies, the heat island is an important parameter for describing the urban space heat environment condition and explaining the formation of the urban space heat environment, and is one of the core content of the current urban space heat environment research, and the common research method comprises a split window algorithm, a radiation transmission equation, a single window algorithm and the like. In 1989, roth et al found that the surface temperature distribution form and the temperature value obtained by inversion have smaller errors from the temperature value of the observation point, and proved the accuracy of thermal infrared remote sensing data; in 2001, the mushroom and the like first invert the surface temperature of the remote sensing data by using a single window algorithm, and calculate the related atmospheric parameters and inversion formulas required by the inversion of the surface temperature; in 2006, huang Miaofen et al used 3 methods to invert the Landsat (TM) image to invert the surface temperature, and found that the inversion value of the single window algorithm was closest to the measured value. The application research of the ground surface temperature by utilizing remote sensing inversion is correspondingly increased, in 2011, liu Chunguo and the like, inversion analysis is carried out on the thermal infrared band of Landsat EMT+, in the relation research of the ground surface temperature and the ecological space, the relation between the normalized vegetation index NDVI and the ground surface temperature is mostly concentrated, and the research of a learner based on remote sensing data and meteorological observation data shows that obvious correlation exists between the NDVI and the ground surface temperature, and the higher the vegetation proportion is, the lower the ground surface temperature is. The space downscaling algorithm for the surface temperature product of the moderate resolution imaging spectrometer MODIS is provided, wherein the space downscaling of the surface temperature is relatively less in supporting the application analysis of the space downscaling of the surface temperature, and the like, and multi-scale geographic weighted regression MGWR is introduced to analyze the scale difference of normalized vegetation index NDVI, digital elevation model DEM, gradient and longitude and latitude on the surface temperature space pattern image. The preprocessing of the data set and the selection of the influencing factors lack of standardization and comprehensiveness, cannot provide good basis for model training, and do not have detailed description of the applicable types of the models, so that the down-scale analysis with higher surface temperature precision is lacking, and meanwhile, the real-time large-scale research requirement of the thermal environment cannot be provided, which is a part of limitation existing in the current research.
The following problems exist in the prior art:
1. the ground surface temperature downscaling data cannot be obtained in a large range with higher precision, and urban thermal environment changes are researched and analyzed;
2. the remote sensing data are various in variety and different in specification, and a standard data set is required to be manufactured;
3. failing to construct the prediction effect of the downscaling model and the comparison model;
4. the data such as the surface temperature is seriously influenced by natural conditions such as weather and the like;
5. the influence of plaque aggregation of the downscaled image product is not weakened, and the space texture is refined.
Disclosure of Invention
Aiming at the problem to be solved, the invention provides a space-time degradation scale change analysis method of urban thermal environment, which is used for carrying out space-time degradation scale change analysis of urban thermal environment, using a roll neural network and a Long-short-term memory neural network (Convolutional Neural Network-Long-Short Term Memory, CNN-LSTM) model to carry out large-area earth surface temperature (Land surface temperature, LST) time degradation scale, then carrying out space-time degradation earth surface temperature research by using a multi-layer perceptron (MLP, multilayer Perceptron) model on several occasions, determining closely related influence factors of earth surface temperature, such as normalized interpolation vegetation index, solar radiation, relative humidity and the like, by consulting documents and related data, obtaining high-molecular satellites of the earth surface temperature and related influence factors and multi-load remote sensing image data, and carrying out space-time degradation on an MODIS earth surface temperature data product by a research master to strengthen the accuracy of the model through abstract relation between the factors and the earth surface temperature. The method comprises the steps of extracting in batches to obtain usable grid files, preprocessing data, adjusting space and time resolution, calling a fishing net to extract needed data, eliminating abnormal values contained in the data to obtain a high-resolution and low-resolution data set, constructing a CNN-LSTM model, and adding a convolutional neural network to accelerate the running speed of the model, wherein the long-term and short-term memory neural network model is suitable for time series data; adjusting each parameter of a model, training and outputting a low-resolution data set, waiting for earth time resolution data, constructing an MLP model to train and verify the earth time resolution data set obtained by adding a high-resolution data set, optimizing the model to obtain a high-space time resolution data set, training to obtain a specific surface temperature value and real remote sensing data to verify the real remote sensing data to meet requirements, continuing to perform the next step, calling the low-resolution data set by the adjusted downscale model to further train by a copying method, generating high-resolution data according to abstract relations existing in different influence factor values, finally refining the obtained downscale data by spatial textures to obtain a final urban thermal environment downscale analysis effect graph, realizing the effect of urban thermal environment analysis and research, and assisting government decisions and urban resident life health.
The high-precision large-range downscaling change analysis of the surface temperature can realize the change monitoring analysis of the space-time characteristics of the urban heat island, acquire the urban heat island intensity and distribution based on time sequences, and provide decision references for urban planning and development, health and life of urban residents and the like.
The implementation steps of the technical scheme of the method are as follows:
the first step: the method comprises the steps of creating a coarse resolution data set of a training and checking model (ground surface temperature can be inverted by using Landsat remote sensing image data), collecting remote sensing data of MODIS and ERA5 ground surface temperature (LST, land Surface Temperature) and related influence factors, including vegetation normalized index (NDVI, normalized differential vegetation index), relative humidity (RH, relative humidity), solar radiation, atmospheric pressure (atmospheric pressure), precipitation (precipitation) and air temperature (air temperature), and processing all collected remote sensing data to obtain two data sets with low resolution and high resolution respectively. The method comprises the following specific steps:
1. in the experiment, the LST uses MODIS11 surface temperature products, meanwhile ERA5 surface temperature analysis data are also included, the surface temperature products and the data sources thereof are different, the obtained remote sensing data are subjected to batch cutting, resampling, fishing net extraction, attribute table conversion and other pretreatment, and finally the remote sensing data are finished into EXCEL table data.
2. The time and the spatial resolution of each data are adjusted, LST data are adjusted according to actual conditions, MODIS ground surface temperature data are copied into 24 parts, manufacturing hours are unified into units of each hour and 1000M spatial resolution, NDVI data which are most relevant to the ground surface temperature are adjusted into two parts, the data comprise 1000M resolution and 100M resolution after linear resampling, ERA5 data are linearly resampled into 100M resolution, a data set is divided into two parts, the first part is a 1000M spatial resolution data set with the first step training model adjusting parameters and performances, and the second part is a 100M resolution data set for training a downscale model for improving the spatial resolution.
3. The influence of abnormal values in the data set is eliminated by using a latest value substitution method, and normal data values closest to the abnormal values are used for substitution.
And a second step of: constructing a CNN-LSTM downscaling model, wherein the CNN model comprises: a one-dimensional input convolution layer having 32 convolution kernels of size 7, an activation function of "relu", a pooling layer of size 7, and an lstm model comprising (1) an input layer and a hidden layer h1: there are 64 neurons, and the activation functions are all "relu"; (2) hidden layer h2: there are 32 neurons with an activation function of "relu" and a Dropout of 0.3; (3) output layer: there are 1 neuron, and the activation function is "sigmoid". The training period was adjusted to 50, the batch size was 30, and the monitor used the h5 file to monitor the training of the network and to attenuate the learning rate, and used the adam optimizer and set some parameters.
And a third step of: training and verifying a CNN-LSTM model, putting a 1000M resolution data set into a CNN-LSTM downscaling model, dividing data set training and verifying model parameters and accuracy, performing time downscaling on MODIS surface temperature product data to obtain surface temperature data in units of each hour, and performing linear resampling on the obtained data to obtain surface temperature data with 100M resolution.
Fourth step: constructing a multi-layer perceptron (MLP, multilayer Perceptron) downscaling model, (1) an input layer and a hidden layer h1: there are 64 neurons, and the activation functions are all "relu"; (2) hidden layer h2: there are 32 neurons, the activation function is "relu"; (3) output layer: there are 1 neuron, and the activation function is "sigmoid". The training period was adjusted to 50, the batch size was 30, and the monitor used the h5 file to monitor the training of the network and to attenuate the learning rate, and used the adam optimizer and set some parameters.
Fifth step: training and verifying an MLP model, and importing an NDVI and ERA5 ground surface temperature data set with 100M resolution and the time-scale-down ground surface temperature data just obtained into the model to form an 85% training set and a 15% verifying set, and training and verifying the MLP model to obtain ground surface temperature space scale-down precision and data.
Sixth step: determining the coefficient (R) by means of Mean Square Error (MSE), root Mean Square Error (RMSE), mean Absolute Error (MAE) 2 ) The multiple indexes are used as the basis, the model effect is displayed by data visualization, the MLP model prediction effect can be obtained by measuring the prediction precision, and the evaluation index calculation formula is shown as follows:
y in the above formulas (1) - (4) is an observed value;is the mean value of the observed values; />Is a predicted value; w (w) i The weight is corresponding to the ground surface temperature extraction points, the weight is reduced when the ground surface temperature extraction points are abnormal values, and n is the number of the ground surface temperature extraction points of the prediction area.
Seventh, changing the size of the fishing net by using the same data processing flow, extracting an image data set with higher spatial resolution, performing downscaling training on the model, performing space-time analysis on the urban thermal environment by using MLP model prediction data to obtain the hour ground surface temperature prediction data, and performing next precision index evaluation.
Eighth, the generated surface temperature prediction data is subjected to space texture refinement, and the formula is as follows:
w in the above formula (5) is the average difference value between two adjacent points in 1000M surface temperature data, t i Is the value of a point of the surface temperature in the data set, t i+1 At t i The value of the next surface temperature point of (C), T in the formula (6) i Is 100M after spatial downscalingAnd (3) measuring the difference value between two points in the predicted data by using the obtained average difference value W, and applying a formula (6) to respectively add and subtract the average error between the two points when the average difference value is larger than the average difference value, subtracting the average error with a large value, and adding the average error with a small value. And after all the steps are completed, a final surface temperature scale-down product is obtained, and information such as urban heat island strength, space-time distribution and the like with lower scale is generated, so that government decision making is further assisted, and healthy life of urban residents is guaranteed.
Preferably, the third step is finished, and space texture refinement can be performed to obtain the surface temperature time scale reduction data.
Preferably, the method can fully combine the medium spatial resolution characteristic in the satellite remote sensing image and the high time resolution characteristic of the MODIS ground surface temperature product to fuse the medium spatial resolution characteristic into the urban ground surface temperature data set with the optimal spatial scale of the month sequence, the seasonal sequence and the annual sequence.
The invention aims to solve the technical problems that: 1. obtaining ground surface temperature scale-down data with high precision in a large range, and researching and analyzing urban thermal environment changes; 2. the remote sensing data are various in variety and different in specification, and a standard data set is required to be manufactured; 3. constructing a scale-down model and comparing the prediction effect of the model; 4. the data such as the surface temperature is seriously influenced by natural conditions such as weather and the like; 5. weakening the plaque aggregation influence of the downscaled image product and refining the space texture.
The beneficial effects of the invention are as follows: 1. the accuracy of the model downscaling result is higher; 2. the data source which is easy to obtain and free to open is used, so that the data loss and other problems are strong in inclusion; 3. the ground surface temperature can be reduced for a large area.
Drawings
FIG. 1 is a process structure diagram of a space-time downscaling variation analysis method of an urban thermal environment according to the invention;
FIG. 2 is a CNN-LSTM model structure of the present invention;
FIG. 3 is a view showing the structure of an MLP model according to the present invention;
Detailed Description
The invention is further described below with reference to examples. The following examples are presented only to aid in the understanding of the invention. It should be noted that it will be apparent to those skilled in the art that modifications can be made to the present invention without departing from the principles of the invention, and such modifications and adaptations are intended to be within the scope of the invention as defined in the following claims.
Other features and advantages of the present invention will become apparent to those skilled in the art from the following description, taken in conjunction with the accompanying drawings, illustrating embodiments of the invention by way of specific examples. The invention may be practiced or carried out in other embodiments and details within the scope and range of equivalents of the various features and advantages of the invention.
FIG. 1 is a process structure diagram of a space-time downscaling variation analysis method of an urban thermal environment, the invention comprises the steps of:
the first step: the method comprises the steps of preparing a coarse resolution data set of a training and checking model, inverting the surface temperature by utilizing Landsat remote sensing image data, collecting remote sensing data of MODIS and ERA5 surface temperature (LST, land Surface Temperature) and related influence factors, including vegetation normalized index (NDVI, normalized differential vegetation index), relative humidity (RH, relative humidity), solar radiation, atmospheric pressure (atmospheric pressure), precipitation (precipitation) and air temperature (air temperature), and processing all collected remote sensing data to obtain two data sets with low resolution and high resolution respectively. The method comprises the following specific steps:
1. in the experiment, the LST uses MODIS11 surface temperature products, meanwhile ERA5 surface temperature analysis data are also included, the surface temperature products and the data sources thereof are different, the obtained remote sensing data are subjected to batch cutting, resampling, fishing net extraction, attribute table conversion and other pretreatment, and finally the remote sensing data are finished into EXCEL table data.
2. The time and the spatial resolution of each data are adjusted, LST data are adjusted according to actual conditions, MODIS ground surface temperature data are copied into 24 parts, manufacturing hours are unified into units of each hour and 1000M spatial resolution, NDVI data which are most relevant to the ground surface temperature are adjusted into two parts, the data comprise 1000M resolution and 100M resolution after linear resampling, ERA5 data are linearly resampled into 100M resolution, a data set is divided into two parts, the first part is a 1000M spatial resolution data set with the first step training model adjusting parameters and performances, and the second part is a 100M resolution data set for training a downscale model for improving the spatial resolution.
3. The influence of abnormal values in the data set is eliminated by using a latest value substitution method, and normal data values closest to the abnormal values are used for substitution.
And a second step of: constructing a CNN-LSTM downscaling model, as shown in FIG. 2, wherein the CNN model comprises: a one-dimensional input convolution layer having 32 convolution kernels of size 7, an activation function of "relu", a pooling layer of size 7, and an lstm model comprising (1) an input layer and a hidden layer h1: there are 64 neurons, and the activation functions are all "relu"; (2) hidden layer h2: there are 32 neurons with an activation function of "relu" and a Dropout of 0.3; (3) output layer: there are 1 neuron, and the activation function is "sigmoid". The training period was adjusted to 50, the batch size was 30, and the monitor used the h5 file to monitor the training of the network and to attenuate the learning rate, and used the adam optimizer and set some parameters.
And a third step of: training and verifying a CNN-LSTM model, putting a 1000M resolution data set into a CNN-LSTM downscaling model, dividing data set training and verifying model parameters and accuracy, performing time downscaling on MODIS surface temperature product data to obtain surface temperature data in units of each hour, and performing linear resampling on the obtained data to obtain surface temperature data with 100M resolution.
Fourth step: constructing a multi-layer perceptron (MLP, multilayer Perceptron) downscaling model, as shown in fig. 3, (1) an input layer and a hidden layer h1: there are 64 neurons, and the activation functions are all "relu"; (2) hidden layer h2: there are 32 neurons, the activation function is "relu"; (3) output layer: there are 1 neuron, and the activation function is "sigmoid". The training period was adjusted to 50, the batch size was 30, and the monitor used the h5 file to monitor the training of the network and to attenuate the learning rate, and used the adam optimizer and set some parameters.
Fifth step: training and verifying an MLP model, and importing an NDVI and ERA5 ground surface temperature data set with 100M resolution and the time-scale-down ground surface temperature data just obtained into the model to form an 85% training set and a 15% verifying set, and training and verifying the MLP model to obtain ground surface temperature space scale-down precision and data.
Sixth step: determining the coefficient (R) by means of Mean Square Error (MSE), root Mean Square Error (RMSE), mean Absolute Error (MAE) 2 ) The multiple indexes are used as the basis, the model effect is displayed by data visualization, the MLP model prediction effect can be obtained by measuring the prediction precision, and the evaluation index calculation formula is shown as follows:
y in the above formulas (1) - (4) is an observed value;is the mean value of the observed values; />Is a predicted value; w (w) i The weight is corresponding to the ground surface temperature extraction points, the weight is reduced when the ground surface temperature extraction points are abnormal values, and n is the number of the ground surface temperature extraction points of the prediction area.
Seventh, changing the size of the fishing net by using the same data processing flow, extracting an image data set with higher spatial resolution, performing downscaling training on the model, performing space-time analysis on the urban thermal environment by using MLP model prediction data to obtain the hour ground surface temperature prediction data, and performing next precision index evaluation.
Eighth, the generated surface temperature prediction data is subjected to space texture refinement, and the formula is as follows:
w in the above formula (5) is the average difference value between two adjacent points in 1000M surface temperature data, t i Is the value of a point of the surface temperature in the data set, t i+1 At t i The value of the next surface temperature point of (C), T in the formula (6) i For a certain point ground surface temperature value of 100M resolution after spatial downscaling, n is the number of the ground surface temperature data extraction points of 1000M resolution in the downscaling area, the obtained average difference W is used for measuring the difference value between two points in predicted data, and when the difference value is larger than the average difference value, the average error is respectively added and subtracted to the two points, subtraction is carried out with large value, and addition is carried out with small value. And after all the steps are completed, a final surface temperature scale-down product is obtained, and information such as urban heat island strength, space-time distribution and the like with lower scale is generated, so that government decision making is further assisted, and healthy life of urban residents is guaranteed.
Preferably, the third step is finished, and space texture refinement can be performed to obtain the surface temperature time scale reduction data.
Preferably, the method can fully combine the medium spatial resolution characteristic in the satellite remote sensing image and the high time resolution characteristic of the MODIS ground surface temperature product to fuse the medium spatial resolution characteristic into the urban ground surface temperature data set with the optimal spatial scale of the month sequence, the seasonal sequence and the annual sequence.
The technical key points of the invention are as follows: 1. adding the data of the influence factors such as vegetation normalization index (NDVI), relative Humidity (RH), solar radiation (SISF), atmospheric pressure, precipitation, air temperature and the like except the data of MODIS and ERA5 surface temperature (LST) to prepare a data set; 2. providing data preprocessing for processing remote sensing images, abnormal values and the like by using a batch extraction program of the fishing nets with resolution of 1000M and 100M respectively; 3. constructing a CNN-LSTM model to realize time downscaling, and then constructing an MLP model to realize space downscaling to obtain the target time and space downscaling surface temperature data;
is only a preferred embodiment of the present invention; the scope of the invention is not limited in this respect. Any person skilled in the art, within the technical scope of the present disclosure, may apply to the present invention, and the technical solution and the improvement thereof are all covered by the protection scope of the present invention.
And the invention is applicable to the prior art where it is not described.
Claims (3)
1. The space-time degradation scale change analysis method for the urban thermal environment is characterized by comprising the following steps of:
step 1: making a coarse resolution data set of a training and checking model, collecting remote sensing data of MODIS and ERA5 surface temperature and related influence factors, including vegetation normalization index, relative humidity, solar radiation, atmospheric pressure, precipitation and air temperature, and processing all collected remote sensing data to obtain two data sets respectively with low resolution and high resolution; the method comprises the following specific steps:
in the step 1.1, the LST uses MODIS11 surface temperature products, meanwhile ERA5 surface temperature analysis data are also used, the obtained remote sensing data are subjected to batch cutting, resampling, fishing net extraction and attribute table conversion, pretreatment is carried out, and finally the remote sensing data are arranged into EXCEL form data;
step 1.2, adjusting time and spatial resolution of each data, wherein LST data are adjusted according to actual conditions, MODIS ground surface temperature data are copied into 24 parts, manufacturing hours are unified into units of each hour and 1000M spatial resolution, NDVI data which are most relevant to ground surface temperature are adjusted into two parts, the data comprise 1000M resolution and 100M resolution after linear resampling, ERA5 data are linearly resampled into 100M resolution, a data set is divided into two parts, the first part is a 1000M spatial resolution data set for adjusting parameters and performance of a first training model, and the second part is a 100M resolution data set for improving a spatial resolution training downscale model;
step 1.3, eliminating the influence of abnormal values in the data set by using a nearest value substitution method, and substituting the normal data values nearest to the abnormal values;
step 2: constructing a CNN-LSTM downscaling model, wherein the CNN model comprises: a one-dimensional input convolution layer with 32 convolution kernels of size 7, an activation function "relu", a pooling layer of size 7; wherein the LSTM model comprises an input layer and a hidden layer h1: there are 64 neurons, and the activation functions are all "relu"; hidden layer h2: there are 32 neurons with an activation function of "relu" and a Dropout of 0.3; output layer: there are 1 neuron, and the activation function is "sigmoid"; adjusting the training period to be 50, the batch size to be 30, using an h5 file to monitor the training of the network by a monitor, performing attenuation processing on the learning rate, and using an adam optimizer to perform parameter setting;
step 3: training and verifying a CNN-LSTM model, putting a 1000M resolution data set into a CNN-LSTM downscaling model, dividing data set training and verifying model parameters and accuracy, performing time downscaling on MODIS surface temperature product data to obtain surface temperature data in units of each hour, and performing linear resampling on the obtained data to obtain surface temperature data with 100M resolution;
step 4: constructing a multi-layer perceptron downscaling MLP model, wherein the MLP model comprises an input layer and a hidden layer h1: there are 64 neurons, and the activation functions are all "relu"; hidden layer h2: there are 32 neurons, the activation function is "relu"; output layer: there are 1 neuron, and the activation function is "sigmoid"; adjusting the training period to be 50, the batch size to be 30, and using an h5 file to monitor the training of the network, performing attenuation processing on the learning rate, using an adam optimizer and performing some parameter settings by a monitor;
step 5: training and verifying an MLP model, and importing an NDVI and ERA5 ground surface temperature data set with 100M resolution and the time-scale ground surface temperature data obtained just to the MLP model to form an 85% training set and a 15% verifying set, and training and verifying the MLP model to obtain ground surface temperature space scale-down precision and data;
step 6: determining the coefficient (R) by means of Mean Square Error (MSE), root Mean Square Error (RMSE), mean Absolute Error (MAE) 2 ) The multiple indexes are used as the basis, the model effect is displayed by data visualization, the MLP model prediction effect can be obtained by measuring the prediction precision, and the evaluation index calculation formula is shown as follows:
y in the above formulas (1) - (4) is an observed value;is the mean value of the observed values; />Is a predicted value; w (W) i The weight is reduced when the ground surface temperature extraction points are abnormal values, and n is the number of the ground surface temperature extraction points of the prediction area;
step 7, changing the size of the fishing net by using the same data processing flow, extracting an image data set with higher spatial resolution, performing downscaling training on the model, performing space-time analysis on the urban thermal environment by using MLP model prediction data to obtain hour ground surface temperature prediction data, and performing next precision index evaluation;
and 8, performing space texture refinement on the generated surface temperature prediction data, wherein the formula is as follows:
w in the above formula (5) is the average difference value between two adjacent points in 1000M surface temperature data, t i Is the value of a point of the surface temperature in the data set, t i+1 At t i The value of the next surface temperature point of (C), T in the formula (6) i For a certain point ground surface temperature value with 100M resolution after spatial downscaling, n is the number of ground surface temperature data extraction points with 1000M resolution in a downscaling area, and the average difference W is used for measuring the difference between two points in predicted data; applying a formula (6), when the average difference value is larger than the average difference value, respectively adding and subtracting the average difference value from two points, subtracting the difference value with large value, and adding the difference value with small value; finishing and space visualization are carried out after all the steps are completed, so that a final ground surface temperature dimension-reducing product is obtained, and urban heat island strength and space-time distribution information with lower dimensions are generated.
2. The method for analyzing the space-time downscaling variation of the urban thermal environment according to claim 1, wherein the space texture refinement is carried out after the step 1.3 is finished to obtain the surface temperature time downscaling data.
3. The method for analyzing the space-time downscaling change of the urban thermal environment according to claim 2, wherein the method can fully combine the characteristics of the medium space resolution in the satellite remote sensing image and the characteristics of the high time resolution of the MODIS surface temperature product to fuse the characteristics into the urban surface temperature data set with the optimal space scale, namely a month sequence, a season sequence and a year sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211365577.3A CN116050567B (en) | 2022-10-31 | 2022-10-31 | Space-time degradation scale change analysis method for urban thermal environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211365577.3A CN116050567B (en) | 2022-10-31 | 2022-10-31 | Space-time degradation scale change analysis method for urban thermal environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116050567A CN116050567A (en) | 2023-05-02 |
| CN116050567B true CN116050567B (en) | 2023-12-15 |
Family
ID=86122448
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211365577.3A Active CN116050567B (en) | 2022-10-31 | 2022-10-31 | Space-time degradation scale change analysis method for urban thermal environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116050567B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016132161A1 (en) * | 2015-02-16 | 2016-08-25 | Kontoes Charalampos | Method that detects areas of active fire hotspots in real-time, calculates the most probable ignition point and assesses fire probability indicators, using satellite images and fuel data. |
| CN106483147A (en) * | 2016-10-14 | 2017-03-08 | 中国科学院遥感与数字地球研究所 | The long-term sequence passive microwave soil moisture accuracy improvements research worked in coordination with based on MODIS and measured data |
| CN108896185A (en) * | 2018-05-14 | 2018-11-27 | 河海大学 | Remote Sensing temperature space NO emissions reduction method based on normalization desert index |
| CN112560570A (en) * | 2020-09-29 | 2021-03-26 | 中国科学院大学 | High-resolution earth surface temperature estimation method based on cooperative downscaling and data fusion |
| CN113486000A (en) * | 2021-09-08 | 2021-10-08 | 中国测绘科学研究院 | Surface evapotranspiration data downscaling method based on multi-source data and deep learning |
-
2022
- 2022-10-31 CN CN202211365577.3A patent/CN116050567B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016132161A1 (en) * | 2015-02-16 | 2016-08-25 | Kontoes Charalampos | Method that detects areas of active fire hotspots in real-time, calculates the most probable ignition point and assesses fire probability indicators, using satellite images and fuel data. |
| CN106483147A (en) * | 2016-10-14 | 2017-03-08 | 中国科学院遥感与数字地球研究所 | The long-term sequence passive microwave soil moisture accuracy improvements research worked in coordination with based on MODIS and measured data |
| CN108896185A (en) * | 2018-05-14 | 2018-11-27 | 河海大学 | Remote Sensing temperature space NO emissions reduction method based on normalization desert index |
| CN112560570A (en) * | 2020-09-29 | 2021-03-26 | 中国科学院大学 | High-resolution earth surface temperature estimation method based on cooperative downscaling and data fusion |
| CN113486000A (en) * | 2021-09-08 | 2021-10-08 | 中国测绘科学研究院 | Surface evapotranspiration data downscaling method based on multi-source data and deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 多尺度地理加权回归的地表温度降尺度研究;祝新明等;遥感学报(第8期);第1749-1766页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116050567A (en) | 2023-05-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110751094B (en) | Crop yield estimation method based on GEE comprehensive remote sensing image and deep learning method | |
| Ramirez et al. | Artificial neural network technique for rainfall forecasting applied to the Sao Paulo region | |
| CN110610054B (en) | Method and system for constructing cuboid inversion model of soil humidity | |
| CN104239706B (en) | A kind of preparation method of ground observation temperature space-time data collection | |
| Dong et al. | Automatic system for crop pest and disease dynamic monitoring and early forecasting | |
| CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
| CN108984803B (en) | Method and system for spatializing crop yield | |
| CN112836610A (en) | Land use change and carbon reserve quantitative estimation method based on remote sensing data | |
| Long et al. | Reconstruction of historical arable land use patterns using constrained cellular automata: A case study of Jiangsu, China | |
| CN105608293A (en) | Forest aboveground biomass inversion method and system fused with spectrum and texture features | |
| CN106951979A (en) | The crop maturity phase Forecasting Methodology that remote sensing, crop modeling are merged with weather forecast | |
| CN109711102A (en) | A kind of crop casualty loss fast evaluation method | |
| CN109800921B (en) | Regional winter wheat yield estimation method based on remote sensing phenological assimilation and particle swarm optimization | |
| CN113837499A (en) | Ultra-short-term wind power prediction method and system | |
| CN113505923A (en) | Regional power grid short-term load prediction method and system | |
| CN116403048B (en) | Crop growth estimation model construction method based on multi-mode data fusion | |
| CN114529097A (en) | Multi-scale crop phenological period remote sensing dimensionality reduction prediction method | |
| CN115544889A (en) | Numerical mode precipitation deviation correction method based on deep learning | |
| CN112651108A (en) | Method for decoupling influences of meteorological elements and vegetation dynamics on hydrological elements | |
| CN116029419A (en) | Deep learning-based long-term new energy daily average generation power prediction method and system | |
| CN112001121A (en) | Large-area gentle region predictive soil mapping method based on solar radiation | |
| Lou et al. | An effective method for canopy chlorophyll content estimation of marsh vegetation based on multiscale remote sensing data | |
| Xin et al. | Satellite mapping of maize cropland in one-season planting areas of China | |
| CN116050567B (en) | Space-time degradation scale change analysis method for urban thermal environment | |
| Yang et al. | On grass yield remote sensing estimation models of China’s northern farming-pastoral ecotone |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |