CN111401644A - Rainfall downscaling space prediction method based on neural network - Google Patents
Rainfall downscaling space prediction method based on neural network Download PDFInfo
- Publication number
- CN111401644A CN111401644A CN202010197185.5A CN202010197185A CN111401644A CN 111401644 A CN111401644 A CN 111401644A CN 202010197185 A CN202010197185 A CN 202010197185A CN 111401644 A CN111401644 A CN 111401644A
- Authority
- CN
- China
- Prior art keywords
- data
- trmm
- measurement task
- neural network
- rainfall measurement
- 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.)
- Pending
Links
Images
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- 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
Abstract
The invention discloses a rainfall downscaling space prediction method based on a neural network, which is based on the neural network in machine learning and realizes the establishment of a nonlinear model between rainfall and related factors so as to realize the downscaling space prediction of rainfall, and the method comprises the following steps: acquiring TRMM rainfall data, vegetation index NDVI data and digital elevation model DEM data of a tropical rainfall measurement task, preprocessing the acquired data, unifying the data into the same coordinate system, normalizing the data, dividing the data according to a random principle, dividing the data into a training set and a testing set, respectively training and verifying the models, and finally performing downscaling prediction on spatial distribution of rainfall through the established models. In the application of the invention, the neural network method is applied to precipitation prediction, so that the prediction precision of precipitation spatial distribution is improved; the problem of insufficient coverage of the TRMM data space of the tropical rainfall measurement task is solved; the spatial resolution of the precipitation prediction is improved.
Description
Technical Field
The invention relates to the technical field of precipitation spatial distribution downscaling prediction, in particular to a precipitation downscaling spatial prediction method based on a neural network.
Background
Precipitation is one of the most important factors in meteorological related research, and has important significance on aspects of agricultural resource management, soil erosion, hydrological related research and the like. Uneven spatial distribution of precipitation is also an important cause of some natural disasters, such as drought and flood disasters, so that the spatial distribution of precipitation has great influence on daily life of people. Research on spatial distribution of precipitation has been carried out for many years, and initially people observed precipitation distribution on the spot through a ground observation station is a very reliable and high-precision method for measuring precipitation distribution, and the method is still used in many scientific researches. However, the observation data of the method can only observe the precipitation data of a specific site, and the method is time-consuming and labor-consuming if a large-scale precipitation distribution is measured. The development of satellite technology makes it possible to measure precipitation distribution in a wide range at low cost, so that precipitation data observed by satellites becomes an important data source in relevant research fields such as meteorology, environment and the like.
However, the current daily rainfall data is mostly tropical rainfall measurement task TRMM satellite data, which has limited coverage, cannot cover the world, and has too coarse resolution to meet the demand.at present, the research on rainfall distribution is limited by satellite data, and a prediction method for researching rainfall distribution is necessary.A plurality of relevant researchers solve the problem, and try to find some relevant indexes for predicting rainfall, and have made researches on the problem, and the researches search for factors influencing rainfall, such as elevation, vegetation index NDVI and the like, and also make a research on using a plurality of factors in combination.
Disclosure of Invention
In order to overcome the above problems or at least partially solve the above problems, embodiments of the present invention provide a precipitation downscaling spatial prediction method based on a neural network, which achieves higher-precision precipitation distribution prediction, solves the problem of incomplete coverage of TRMM data in a tropical rainfall measurement task, and improves precipitation prediction resolution.
The embodiment of the invention is realized by the following steps:
a rainfall downscaling space prediction method based on a neural network comprises the following steps:
s1: acquiring TRMM rainfall data, vegetation index NDVI data and digital elevation model DEM data of a tropical rainfall measurement task in an area to be predicted;
s2, preprocessing data, including:
s2-1: rotating and defining projection is carried out on TRMM data of the tropical rainfall measurement task;
s2-2: projecting TRMM rainfall data, vegetation index NDVI data and digital elevation model DEM data of a tropical rainfall measurement task into a proper coordinate system;
s2-3: resampling vegetation indexes NDVI and data of a digital elevation model DEM, and keeping the data consistent with data of a tropical rainfall measurement task TRMM;
s3: training and validation data acquisition, including:
s3-1: normalizing the data acquired in the step 2, and changing the value range of each pixel into [0,1 ];
s3-2: randomly classifying the normalized data into at least a training set and a test set, and distributing the training set and the test set according to a preset proportion;
s4: and training and parameter adjusting are carried out on the neural network model by utilizing the training set and the testing set, and rainfall distribution is predicted by utilizing the neural network model after training and parameter adjusting.
In some embodiments of the present invention, a rainfall downscaling spatial prediction method based on a neural network includes, in step 1, acquiring tropical rainfall measurement task TRMM rainfall data, vegetation index NDVI data, and digital elevation model DEM data of an area to be predicted in a manner that:
the TRMM data of the tropical rainfall measurement task is downloaded and obtained from a NASA official website, and the TRMM data range of the tropical rainfall measurement task is 60 degrees S-60 degrees N;
the digital elevation model DEM data uses space shuttle radar terrain mapping mission SRTM data, the resolution ratio is 0.00083 degrees, and the coordinate system is WGS 1984;
the vegetation index NDVI data were obtained by a medium resolution imaging spectrometer MODIS downloaded from the L AADS website with a resolution of 0.005 ° and a coordinate system WGS 1984.
In some embodiments of the present invention, a precipitation downscaling spatial prediction method based on a neural network, the step of rotating and defining projection on the TRMM data of the tropical rainfall measurement task in step 2-1 includes:
s2-1-1: deviation of the TRMM data of the tropical rainfall measurement task from the normal remote sensing image is 90 degrees, and the TRMM data of the tropical rainfall measurement task is rotated by 90 degrees anticlockwise or 270 degrees clockwise;
s2-1-2: the TRMM data of the tropical rainfall measurement task adopts a WGS1984 coordinate system, the resolution is 0.25 degrees, geographic information and resolution information are added to the TRMM data of the tropical rainfall measurement task, and the unification of the TRMM data of the tropical rainfall measurement task and the residual data coordinate system is completed.
In some embodiments of the present invention, a method for predicting the downscaling space of precipitation based on a neural network, the method of projecting to a suitable coordinate system in step 2-2 comprises:
s2-2-1: selecting or designing a proper coordinate system according to the geographical position of the sample area and the application requirement;
s2-2-2: projecting the TRMM data of the tropical rainfall measurement task into the coordinate system, wherein the original resolution of the TRMM data of the tropical rainfall measurement task is 0.25 degrees, and the size of a projected pixel is set to be 25 km;
s2-2-3: projecting the vegetation index NDVI data into the coordinate system, wherein the resolution ratio of original data of the vegetation index NDVI is 0.005 degrees, and the size of a projected pixel is set to be 5 km;
s2-2-4: and projecting the data of the digital elevation model DEM into the coordinate system, wherein the resolution of the original data of the digital elevation model DEM is 0.00083 degrees, and the size of the projected pixel is set to be 90 m.
In some embodiments of the present invention, in step 2, a precipitation downscaling spatial prediction method based on a neural network further includes performing data pruning according to a prediction area range, where the pruning method includes:
and trimming TRMM data, vegetation index NDVI data and digital elevation model DEM data of the tropical rainfall measurement task by using the area range to be predicted.
In some embodiments of the present invention, a method for predicting a precipitation downscaling space based on a neural network, the step of resampling in step 2-3 comprises:
s2-3-1: resampling vegetation index NDVI data from original 5km resolution ratio to 25km, keeping the same with tropical rainfall measurement task TRMM data, and ensuring pixel coincidence;
s2-3-2: and resampling the DEM data of the digital elevation model from the original resolution of 90m to 25km, keeping the DEM data consistent with TRMM data of a tropical rainfall measurement task, and ensuring pixel coincidence.
In some embodiments of the present invention, a method for spatial prediction of precipitation downscaling based on a neural network, the step of normalizing data in step 3-1 includes:
s3-1-1: the input data requirement value field of the neural network training is [0,1], normalization processing needs to be carried out on the data obtained in the step 2 before training, normalization is carried out on the data by adopting a maximum value and minimum value normalization method, and the calculation mode is as follows:
in formula 1, x'iIs made ofA normalized value; x is the number ofiThe original value before normalization processing; x is the number ofmax，xminRespectively calculating the maximum value and the minimum value of the tropical rainfall measurement task TRMM data, and normalizing the tropical rainfall measurement task TRMM data according to formula 1;
s3-1-2: respectively counting the maximum value and the minimum value of the vegetation index NDVI data, and normalizing the TRMM data of the tropical rainfall measurement task according to the formula 1;
s3-1-3: and respectively counting the maximum value and the minimum value of the DEM data of the digital elevation model, and normalizing the TRMM data of the tropical rainfall measurement task according to the formula 1.
In some embodiments of the present invention, in a method for predicting a precipitation downscaling space based on a neural network, the step of classifying data into a training set and a test set in step 3-2 includes:
s3-2-1: firstly, counting the number C of pixels in TRMM data of a tropical rainfall measurement taskTatolCalculating the number C of samples in the training set according to the numberTrainAnd number of samples C in the test setTestThe calculation formula is as follows:
in the formula 2, CTatolThe total number of samples, namely the number of TRMM data pixels of the heat band rainfall measurement task; cTrainThe number of samples in the training set; cTestThe number of samples in the test set;
s3-2-2: randomly selecting C from TRMM data of tropical rainfall measurement taskTrainThe pixel values are used as samples of a training set, and the residual pixel values are used as samples of a testing set;
s3-2-3: acquiring pixel values of geographic positions corresponding to vegetation indexes NDVI and digital elevation model DEM data according to row and column numbers of sample pixels in a tropical rainfall measurement task TRMM training set, so as to obtain a training set of the vegetation indexes NDVI and the digital elevation model DEM data;
s3-2-4: and acquiring pixel values of geographic positions corresponding to the vegetation index NDVI and the digital elevation model DEM data according to the row and column numbers of the pixels of the samples in the tropical rainfall measurement task TRMM test set, so as to obtain a test set of the vegetation index NDVI and the digital elevation model DEM data.
In some embodiments of the present invention, a rainfall downscaling space prediction method based on a neural network includes, in step 4, training and parameter-adjusting a neural network model by using a training set and a test set, as follows:
s4-1: training a neural network model by using samples in a training set, wherein vegetation indexes NDVI and digital elevation model DEM data are input data, and tropical rainfall measurement task TRMM data are true value data;
s4-2: using the trained neural network model and using vegetation index NDVI and digital elevation model DEM data in the test set as input data to obtain a prediction result of the test set;
s4-3: comparing the predicted value of the neural network model with TRMM data, namely a true value, of a concentrated tropical rainfall measurement task to be tested, and calculating an error RMSE, wherein the formula is as follows:
in formula 3, CTestNumber of samples in order to test, y'iIs the predicted value, y, of the i-th pixel of the modeliIs the true value of the ith pixel;
s4-3: and adjusting parameters in the neural network to reduce the RMSE, wherein the parameters comprise the learning rate of the neural network, the type of the neural network and the hidden layer of the type of the neural network.
In some embodiments of the present invention, a precipitation downscaling spatial prediction method based on a neural network further includes, in step 4:
resampling the high-resolution digital elevation model DEM data to 5km, enabling the high-resolution digital elevation model DEM data to be consistent with the resolution of vegetation index NDVI data, and ensuring pixel coincidence;
and (3) using the vegetation index NDVI with the resolution of 5km and the digital elevation model DEM data with the resolution of 5km as input data of the trained and parameter-adjusted model to obtain precipitation data with the resolution of 5 km.
The embodiment of the invention at least has the following advantages or beneficial effects:
the invention provides a rainfall downscaling space prediction method based on a neural network, which is based on the neural network in machine learning and realizes the establishment of a nonlinear model between rainfall and related factors, thereby realizing the downscaling space prediction of rainfall, and the method mainly comprises the following steps: acquiring TRMM rainfall data, vegetation index NDVI data and digital elevation model DEM data of a tropical rainfall measurement task, preprocessing the data, unifying the data into the same coordinate system, normalizing the data, dividing the data according to a random principle, dividing the data into a training set and a testing set, respectively training and verifying the models, and finally performing downscaling prediction on spatial distribution of rainfall through the established models. In the application of the invention, the neural network method is applied to precipitation prediction, so that the prediction precision of precipitation spatial distribution is improved; the problem of insufficient coverage of the TRMM data space of the tropical rainfall measurement task is solved; the spatial resolution of the precipitation prediction is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is a flowchart of an embodiment of a method for predicting a downscaling space of precipitation based on a neural network according to the present invention;
fig. 2 is a distribution diagram and a partial enlarged view of TRMM data of a tropical rainfall measurement task in a region to be predicted in an embodiment of a rainfall downscaling space prediction method based on a neural network;
fig. 3 is a distribution diagram of vegetation index NDVI data of a region to be predicted in an embodiment of a rainfall downscaling space prediction method based on a neural network;
FIG. 4 is a distribution diagram of digital elevation model DEM data of an area to be predicted in an embodiment of a rainfall downscaling space prediction method based on a neural network;
fig. 5 is a rainfall space distribution diagram of an area to be predicted, which is predicted by using a model in an embodiment of the rainfall downscaling space prediction method based on the neural network.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.
Examples
Referring to fig. 1, the present embodiment provides a method for predicting a precipitation downscaling space based on a neural network, including the following steps:
s1: acquiring TRMM rainfall data, vegetation index NDVI data and digital elevation model DEM data of a tropical rainfall measurement task in an area to be predicted;
s2, preprocessing data, including:
s2-1: rotating and defining projection is carried out on TRMM data of the tropical rainfall measurement task;
s2-2: projecting TRMM rainfall data, vegetation index NDVI data and digital elevation model DEM data of a tropical rainfall measurement task into a proper coordinate system;
s2-3: resampling vegetation indexes NDVI and data of a digital elevation model DEM to 25km, and keeping the data consistent with data of a tropical rainfall measurement task TRMM;
s3: training and validation data acquisition, including:
s3-1: normalizing the data acquired in the step 2, and changing the value range of each pixel into [0,1 ];
s3-2: randomly classifying the normalized data into at least a training set and a test set, and distributing the training set and the test set according to a preset proportion; the ratio of training set to test set is preferably 8: 2.
S4: and training and parameter adjusting are carried out on the neural network model by utilizing the training set and the testing set, and rainfall distribution is predicted by utilizing the neural network model after training and parameter adjusting.
In some embodiments of the present invention, a rainfall downscaling spatial prediction method based on a neural network includes, in step 1, acquiring tropical rainfall measurement task TRMM rainfall data, vegetation index NDVI data, and digital elevation model DEM data of an area to be predicted in a manner that:
the TRMM data of the tropical rainfall measurement task can be obtained from a website https:// disc.gsfc.nasa.gov./and numerous TRMM data products of the tropical rainfall measurement task are provided, wherein the TRMM3B43 data of the tropical rainfall measurement task is the most common data, the TRMM3B43 data of the tropical rainfall measurement task is monthly rainfall data, the resolution is 0.25 degrees, the coordinate system is WGS1984, data time is selected according to needs, corresponding data are downloaded, and the TRMM data range of the tropical rainfall measurement task is 60 degrees S-60 degrees N.
The digital elevation model DEM data uses space shuttle radar terrain mapping mission SRTM data, the resolution ratio is 0.00083 degrees, and the coordinate system is WGS 1984;
the vegetation index NDVI data were obtained by a medium resolution imaging spectrometer MODIS downloaded from the L AADS website with a resolution of 0.005 ° and a coordinate system WGS 1984.
In some embodiments of the present invention, a precipitation downscaling spatial prediction method based on a neural network, the step of rotating and defining projection on the TRMM data of the tropical rainfall measurement task in step 2-1 includes:
s2-1-1: deviation of the downloaded TRMM data of the tropical rainfall measurement task and a normal remote sensing image is 90 degrees, and the TRMM data of the tropical rainfall measurement task is rotated by 90 degrees anticlockwise or 270 degrees clockwise;
s2-1-2: the tropical rainfall measurement task TRMM data is lack of geographic coordinate information and resolution information, a WGS1984 coordinate system is adopted for the tropical rainfall measurement task TRMM data, and the resolution is 0.25 degrees, so that the geographic information and the resolution information are added to the tropical rainfall measurement task TRMM data, and the unification of the number of the tropical rainfall measurement tasks TRMM and other data coordinate systems is completed.
In some embodiments of the present invention, a method for predicting the downscaling space of precipitation based on a neural network, the method of projecting to a suitable coordinate system in step 2-2 comprises:
s2-2-1: selecting or designing a proper coordinate system according to the geographical position of the sample area and the application requirement;
s2-2-2: projecting the TRMM data of the tropical rainfall measurement task into the coordinate system, wherein the original resolution of the TRMM data of the tropical rainfall measurement task is 0.25 degrees, and the size of a projected pixel is set to be 25 km;
s2-2-3: projecting the vegetation index NDVI data into the coordinate system, wherein the resolution ratio of original data of the vegetation index NDVI is 0.005 degrees, and the size of a projected pixel is set to be 5 km;
s2-2-4: and projecting the data of the digital elevation model DEM into the coordinate system, wherein the resolution of the original data of the digital elevation model DEM is 0.00083 degrees, and the size of the projected pixel is set to be 90 m.
In some embodiments of the present invention, in step 2, a precipitation downscaling spatial prediction method based on a neural network further includes performing data pruning according to a prediction area range, where the pruning method includes:
trimming the TRMM data of the tropical rainfall measurement task by using the range of the area to be predicted, selecting Heilongjiang province in China as the area to be predicted, and referring to fig. 2 for the trimmed TRMM data of the tropical rainfall measurement task;
cutting vegetation index NDVI data by using the area range to be predicted, wherein the obtained result is shown in figure 3;
and (4) cutting the data of the digital elevation model DEM by using the area range to be predicted, wherein the obtained result is shown in FIG. 4.
In some embodiments of the present invention, a method for predicting a precipitation downscaling space based on a neural network, the step of resampling in step 2-3 comprises:
s2-3-1: resampling vegetation index NDVI data from original 5km resolution ratio to 25km, keeping the same with tropical rainfall measurement task TRMM data, and ensuring pixel coincidence;
s2-3-2: and resampling the DEM data of the digital elevation model from the original resolution of 90m to 25km, keeping the DEM data consistent with TRMM data of a tropical rainfall measurement task, and ensuring pixel coincidence.
In some embodiments of the present invention, a method for spatial prediction of precipitation downscaling based on a neural network, the step of normalizing data in step 3-1 includes:
s3-1-1: the input data requirement value field of the neural network training is [0,1], normalization processing needs to be carried out on the data obtained in the step 2 before training, normalization is carried out on the data by adopting a maximum value and minimum value normalization method, and the calculation mode is as follows:
in formula 1, x'iIs a normalized value; x is the number ofiThe original value before normalization processing; x is the number ofmax，xminRespectively calculating the maximum value and the minimum value of the tropical rainfall measurement task TRMM data, and normalizing the tropical rainfall measurement task TRMM data according to formula 1;
s3-1-2: respectively counting the maximum value and the minimum value of the vegetation index NDVI data, and normalizing the TRMM data of the tropical rainfall measurement task according to the formula 1;
s3-1-3: and respectively counting the maximum value and the minimum value of the DEM data of the digital elevation model, and normalizing the TRMM data of the tropical rainfall measurement task according to the formula 1.
In some embodiments of the present invention, in a method for predicting a precipitation downscaling space based on a neural network, the step of classifying data into a training set and a test set in step 3-2 includes:
s3-2-1: firstly, counting the number C of pixels in TRMM data of a tropical rainfall measurement taskTatolCalculating the number C of samples in the training set according to the numberTrainAnd number of samples C in the test setTestThe calculation formula is as follows:
in the formula 2, CTatolThe total number of samples, namely the number of TRMM data pixels of the heat band rainfall measurement task; cTrainThe number of samples in the training set; cTestThe number of samples in the test set; total number of pixels C of area to be predictedTatolTo 1676, calculate the number of training set samples CTrain1341, test set sample CTestIs 335.
S3-2-2: randomly selecting C from TRMM data of tropical rainfall measurement taskTrainThe pixel values are used as samples of a training set, and the residual pixel values are used as samples of a testing set;
s3-2-3: acquiring pixel values of geographic positions corresponding to vegetation indexes NDVI and digital elevation model DEM data according to row and column numbers of sample pixels in a tropical rainfall measurement task TRMM training set, so as to obtain a training set of the vegetation indexes NDVI and the digital elevation model DEM data;
s3-2-4: and acquiring pixel values of geographic positions corresponding to the vegetation index NDVI and the digital elevation model DEM data according to the row and column numbers of the pixels of the samples in the tropical rainfall measurement task TRMM test set, so as to obtain a test set of the vegetation index NDVI and the digital elevation model DEM data.
In some embodiments of the present invention, a rainfall downscaling space prediction method based on a neural network includes, in step 4, training and parameter-adjusting a neural network model by using a training set and a test set, as follows:
s4-1: training a neural network model by using samples in a training set, wherein vegetation indexes NDVI and digital elevation model DEM data are input data, and tropical rainfall measurement task TRMM data are true value data;
s4-2: using the trained neural network model and using vegetation index NDVI and digital elevation model DEM data in the test set as input data to obtain a prediction result of the test set;
s4-3: comparing the predicted value of the neural network model with TRMM data, namely a true value, of a concentrated tropical rainfall measurement task to be tested, and calculating an error RMSE, wherein the formula is as follows:
in formula 3, CTestNumber of samples in order to test, y'iIs the predicted value, y, of the i-th pixel of the modeliIs the true value of the ith pixel;
s4-3: and adjusting parameters in the neural network to reduce the RMSE, wherein the parameters comprise the learning rate of the neural network, the type of the neural network and the hidden layer of the type of the neural network.
In some embodiments of the present invention, a precipitation downscaling spatial prediction method based on a neural network further includes, in step 4:
resampling the high-resolution digital elevation model DEM data to 5km, enabling the high-resolution digital elevation model DEM data to be consistent with the resolution of vegetation index NDVI data, and ensuring pixel coincidence;
and (3) using the vegetation index NDVI with the resolution of 5km and the digital elevation model DEM data with the resolution of 5km as input data of the trained and parameter-adjusted model to obtain precipitation data with the resolution of 5km, and referring to fig. 5 as a final result.
In summary, an embodiment of the present invention provides a method for predicting a precipitation downscaling space based on a neural network, which is based on the neural network in machine learning to implement establishment of a nonlinear model between precipitation and a correlation factor, thereby implementing downscaling space prediction of precipitation, and the method mainly includes: acquiring TRMM rainfall data, vegetation index NDVI data and digital elevation model DEM data of a tropical rainfall measurement task, preprocessing the data, unifying the data into the same coordinate system, normalizing the data, dividing the data according to a random principle, dividing the data into a training set and a testing set, respectively training and verifying the models, and finally performing downscaling prediction on spatial distribution of rainfall through the established models. In the application of the invention, the neural network method is applied to precipitation prediction, so that the prediction precision of precipitation spatial distribution is improved; the problem of insufficient coverage of the TRMM data space of the tropical rainfall measurement task is solved; the spatial resolution of the precipitation prediction is improved.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes will occur to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A rainfall downscaling space prediction method based on a neural network is characterized by comprising the following steps:
s1: acquiring TRMM rainfall data, vegetation index NDVI data and digital elevation model DEM data of a tropical rainfall measurement task in an area to be predicted;
s2, preprocessing data, including:
s2-1: rotating and defining projection is carried out on TRMM data of the tropical rainfall measurement task;
s2-2: projecting TRMM rainfall data, vegetation index NDVI data and digital elevation model DEM data of a tropical rainfall measurement task into a proper coordinate system;
s2-3: resampling vegetation indexes NDVI and data of a digital elevation model DEM, and keeping the data consistent with data of a tropical rainfall measurement task TRMM;
s3: training and validation data acquisition, including:
s3-1: normalizing the data acquired in the step 2, and changing the value range of each pixel into [0,1 ];
s3-2: randomly classifying the normalized data into at least a training set and a test set, wherein the training set and the test set are distributed according to a preset proportion;
s4: and training and adjusting parameters of the neural network model by using the training set and the testing set, and predicting precipitation distribution by using the trained and adjusted neural network model.
2. The rainfall downscaling spatial prediction method based on the neural network according to claim 1, wherein in the step 1, the acquiring modes of TRMM rainfall data, vegetation index NDVI data and digital elevation model DEM data of the tropical rainfall measurement task in the area to be predicted comprise:
the TRMM data of the tropical rainfall measurement task is downloaded and obtained from a NASA official website, and the TRMM data range of the tropical rainfall measurement task is 60 degrees S-60 degrees N;
the digital elevation model DEM data uses space shuttle radar terrain mapping mission SRTM data, the resolution ratio is 0.00083 degrees, and the coordinate system is WGS 1984;
the vegetation index NDVI data is obtained by a medium resolution imaging spectrometer MODIS downloaded from the L AADS website, with a resolution of 0.005 ° and a coordinate system WGS 1984.
3. The method for spatial prediction of rainfall downscaling based on neural network of claim 1, wherein the step of rotating and defining the projection of the TRMM data of the tropical rainfall measurement task in step 2-1 comprises:
s2-1-1: deviation of the TRMM data of the tropical rainfall measurement task from the normal remote sensing image is 90 degrees, and the TRMM data of the tropical rainfall measurement task is rotated by 90 degrees anticlockwise or 270 degrees clockwise;
s2-1-2: the TRMM data of the tropical rainfall measurement task adopts a WGS1984 coordinate system, the resolution is 0.25 degrees, geographic information and resolution information are added to the TRMM data of the tropical rainfall measurement task, and the unification of the TRMM data of the tropical rainfall measurement task and the residual data coordinate system is completed.
4. The method for spatial prediction of precipitation downscaling based on neural network as claimed in claim 1, wherein the method of projecting to a suitable coordinate system in step 2-2 comprises:
s2-2-1: selecting or designing a proper coordinate system according to the geographical position of the sample area and the application requirement;
s2-2-2: projecting the TRMM data of the tropical rainfall measurement task into the coordinate system, wherein the original resolution of the TRMM data of the tropical rainfall measurement task is 0.25 degrees, and the size of a projected pixel is set to be 25 km;
s2-2-3: projecting the vegetation index NDVI data into the coordinate system, wherein the resolution ratio of original data of the vegetation index NDVI is 0.005 degrees, and the size of a projected pixel is set to be 5 km;
s2-2-4: and projecting the data of the digital elevation model DEM into the coordinate system, wherein the resolution of the original data of the digital elevation model DEM is 0.00083 degrees, and the size of the projected pixel is set to be 90 m.
5. The method for spatial prediction of precipitation downscaling based on the neural network as claimed in claim 1, wherein the step 2 further comprises performing data pruning according to the prediction area range, and the pruning method comprises:
and trimming TRMM data, vegetation index NDVI data and digital elevation model DEM data of the tropical rainfall measurement task by using the area range to be predicted.
6. The method for spatial prediction of precipitation downscaling based on neural network as claimed in claim 1, wherein the step of resampling in step 2-3 comprises:
s2-3-1: resampling vegetation index NDVI data from original 5km resolution ratio to 25km, keeping the same with tropical rainfall measurement task TRMM data, and ensuring pixel coincidence;
s2-3-2: and resampling the DEM data of the digital elevation model from the original resolution of 90m to 25km, keeping the DEM data consistent with TRMM data of a tropical rainfall measurement task, and ensuring pixel coincidence.
7. The method for spatial prediction of precipitation downscaling based on the neural network as claimed in claim 1, wherein the step of normalizing data in step 3-1 comprises:
s3-1-1: the input data requirement value field of the neural network training is [0,1], normalization processing needs to be carried out on the data obtained in the step 2 before training, normalization is carried out on the data by adopting a maximum value and minimum value normalization method, and the calculation mode is as follows:
in formula 1, x'iIs a normalized value; x is the number ofiThe original value before normalization processing; x is the number ofmax，xminRespectively calculating the maximum value and the minimum value of the tropical rainfall measurement task TRMM data, and normalizing the tropical rainfall measurement task TRMM data according to formula 1;
s3-1-2: respectively counting the maximum value and the minimum value of the vegetation index NDVI data, and normalizing the TRMM data of the tropical rainfall measurement task according to the formula 1;
s3-1-3: and respectively counting the maximum value and the minimum value of the DEM data of the digital elevation model, and normalizing the TRMM data of the tropical rainfall measurement task according to the formula 1.
8. The method for spatial prediction of precipitation downscaling based on neural network as claimed in claim 1, wherein the step of classifying the data into training set and test set in step 3-2 comprises:
s3-2-1: firstly, counting the number C of pixels in TRMM data of a tropical rainfall measurement taskTatolCalculating the number C of samples in the training set according to the numberTrainAnd number of samples C in the test setTestThe calculation formula is as follows:
in the formula 2, CTatolThe total number of samples, namely the number of TRMM data pixels of the heat band rainfall measurement task; cTrainThe number of samples in the training set; cTestThe number of samples in the test set;
s3-2-2: randomly selecting C from TRMM data of tropical rainfall measurement taskTrainThe pixel values are used as samples of a training set, and the residual pixel values are used as samples of a testing set;
s3-2-3: acquiring pixel values of geographic positions corresponding to vegetation indexes NDVI and digital elevation model DEM data according to row and column numbers of sample pixels in a tropical rainfall measurement task TRMM training set, so as to obtain a training set of the vegetation indexes NDVI and the digital elevation model DEM data;
s3-2-4: and acquiring pixel values of geographic positions corresponding to the vegetation index NDVI and the digital elevation model DEM data according to the row and column numbers of the pixels of the samples in the tropical rainfall measurement task TRMM test set, so as to obtain a test set of the vegetation index NDVI and the digital elevation model DEM data.
9. The method for predicting the precipitation downscaling space based on the neural network according to claim 1, wherein the step 4 of training and parameter-adjusting the neural network model by using the training set and the test set comprises the following steps:
s4-1: training a neural network model by using samples in a training set, wherein vegetation indexes NDVI and digital elevation model DEM data are input data, and tropical rainfall measurement task TRMM data are true value data;
s4-2: using the trained neural network model and using vegetation index NDVI and digital elevation model DEM data in the test set as input data to obtain a prediction result of the test set;
s4-3: comparing the predicted value of the neural network model with TRMM data, namely a true value, of a concentrated tropical rainfall measurement task to be tested, and calculating an error RMSE, wherein the formula is as follows:
in formula 3, CTestNumber of samples in order to test, y'iIs the predicted value, y, of the i-th pixel of the modeliIs the true value of the ith pixel;
s4-3: and adjusting parameters in the neural network to reduce the RMSE, wherein the parameters comprise the learning rate of the neural network, the type of the neural network and the hidden layer of the type of the neural network.
10. The method for spatial prediction of precipitation downscaling based on a neural network according to claim 1, wherein the step 4 further comprises:
resampling the high-resolution digital elevation model DEM data to 5km, enabling the high-resolution digital elevation model DEM data to be consistent with the resolution of vegetation index NDVI data, and ensuring pixel coincidence;
and (3) using the vegetation index NDVI with the resolution of 5km and the digital elevation model DEM data with the resolution of 5km as input data of the trained and parameter-adjusted model to obtain precipitation data with the resolution of 5 km.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010197185.5A CN111401644A (en) | 2020-03-19 | 2020-03-19 | Rainfall downscaling space prediction method based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010197185.5A CN111401644A (en) | 2020-03-19 | 2020-03-19 | Rainfall downscaling space prediction method based on neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111401644A true CN111401644A (en) | 2020-07-10 |
Family
ID=71431034
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010197185.5A Pending CN111401644A (en) | 2020-03-19 | 2020-03-19 | Rainfall downscaling space prediction method based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111401644A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112017135A (en) * | 2020-07-13 | 2020-12-01 | 香港理工大学深圳研究院 | Method, system and equipment for spatial-temporal fusion of remote sensing image data |
| CN113923594A (en) * | 2021-10-12 | 2022-01-11 | 中国人民解放军国防科技大学 | Weather distinguishing method based on time division long term evolution network |
| CN114139819A (en) * | 2021-12-07 | 2022-03-04 | 成都理工大学 | Geochemical variable space prediction method based on geostatistical weighted random forest |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017209548A1 (en) * | 2016-06-01 | 2017-12-07 | 서울대학교산학협력단 | Device and method for generating artificial neural network-based prediction model |
| CN109472393A (en) * | 2018-09-30 | 2019-03-15 | 广州地理研究所 | Space NO emissions reduction precipitation data detection method, device and electronic equipment |
| CN110334446A (en) * | 2019-07-05 | 2019-10-15 | 中国水利水电科学研究院 | The mountain torrents Critical Rainfall calculation method of NO emissions reduction processing based on satellite precipitation data |
| CN110738252A (en) * | 2019-10-14 | 2020-01-31 | 广州地理研究所 | Space autocorrelation machine learning satellite precipitation data downscaling method and system |
| CN110737875A (en) * | 2019-09-24 | 2020-01-31 | 浙江省水利河口研究院 | medium-and-long-term torrential flood disaster early warning and prejudging method |
-
2020
- 2020-03-19 CN CN202010197185.5A patent/CN111401644A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017209548A1 (en) * | 2016-06-01 | 2017-12-07 | 서울대학교산학협력단 | Device and method for generating artificial neural network-based prediction model |
| CN109472393A (en) * | 2018-09-30 | 2019-03-15 | 广州地理研究所 | Space NO emissions reduction precipitation data detection method, device and electronic equipment |
| CN110334446A (en) * | 2019-07-05 | 2019-10-15 | 中国水利水电科学研究院 | The mountain torrents Critical Rainfall calculation method of NO emissions reduction processing based on satellite precipitation data |
| CN110737875A (en) * | 2019-09-24 | 2020-01-31 | 浙江省水利河口研究院 | medium-and-long-term torrential flood disaster early warning and prejudging method |
| CN110738252A (en) * | 2019-10-14 | 2020-01-31 | 广州地理研究所 | Space autocorrelation machine learning satellite precipitation data downscaling method and system |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112017135A (en) * | 2020-07-13 | 2020-12-01 | 香港理工大学深圳研究院 | Method, system and equipment for spatial-temporal fusion of remote sensing image data |
| CN112017135B (en) * | 2020-07-13 | 2021-09-21 | 香港理工大学深圳研究院 | Method, system and equipment for spatial-temporal fusion of remote sensing image data |
| CN113923594A (en) * | 2021-10-12 | 2022-01-11 | 中国人民解放军国防科技大学 | Weather distinguishing method based on time division long term evolution network |
| CN113923594B (en) * | 2021-10-12 | 2023-10-03 | 中国人民解放军国防科技大学 | Weather distinguishing method based on time division long term evolution network |
| CN114139819A (en) * | 2021-12-07 | 2022-03-04 | 成都理工大学 | Geochemical variable space prediction method based on geostatistical weighted random forest |
| CN114139819B (en) * | 2021-12-07 | 2023-05-23 | 成都理工大学 | Geochemical variable space prediction method based on geostatistically weighted random forest |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110738252B (en) | Space autocorrelation machine learning satellite precipitation data downscaling method and system | |
| Cooper et al. | Object-based correction of LiDAR DEMs using RTK-GPS data and machine learning modeling in the coastal Everglades | |
| Tariq et al. | Monitoring land use and land cover changes using geospatial techniques, a case study of Fateh Jang, Attock, Pakistan | |
| Jin et al. | Downscaling AMSR-2 soil moisture data with geographically weighted area-to-area regression kriging | |
| CN114037911B (en) | Large-scale forest height remote sensing inversion method considering ecological zoning | |
| CN105912836B (en) | A kind of watershed hydrologic cycle analogy method of pure remotely-sensed data driving | |
| CN111723524B (en) | PM2.5 satellite remote sensing inversion method based on daily variation constraint | |
| CN111401644A (en) | Rainfall downscaling space prediction method based on neural network | |
| CN106776481A (en) | A kind of NO emissions reduction bearing calibration for acting on satellite precipitation data | |
| Arowolo et al. | Comparison of spatial interpolation techniques to generate high‐resolution climate surfaces for Nigeria | |
| CN113297528A (en) | NO based on multi-source big data2High-resolution space-time distribution calculation method | |
| EP3794386B1 (en) | Determination of location-specific weather information for agronomic decision support | |
| CN114091764A (en) | Weather forecast element correction method, weather forecast element correction device, computer equipment and storage medium | |
| CN110059745A (en) | A kind of Basin Rainfall product correction method based on star merged and system | |
| KR101521576B1 (en) | Method for producing meteorological data with high-resolution using K-PRISM and high-resolution meteorological data production system using thereof | |
| CN112800827A (en) | Hyperspectral image classification experimental method | |
| CN111079835B (en) | Himapari-8 atmospheric aerosol inversion method based on deep full-connection network | |
| CN115544889A (en) | Numerical mode precipitation deviation correction method based on deep learning | |
| CN114091756B (en) | Township tsunami risk assessment method based on Thiessen polygon | |
| CN112285808B (en) | Method for reducing scale of APHRODITE precipitation data | |
| Khesali et al. | A method in near-surface estimation of air temperature (NEAT) in times following the satellite passing time using MODIS images | |
| CN110310370B (en) | Method for point-plane fusion of GPS (Global positioning System) and SRTM (short Range TM) | |
| CN114819737B (en) | Method, system and storage medium for estimating carbon reserves of highway road vegetation | |
| CN113205014B (en) | Time sequence data farmland extraction method based on image sharpening | |
| CN115830464A (en) | Plateau mountain agricultural greenhouse automatic extraction method based on multi-source data |
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 |