33 results about "Satellite precipitation" patented technology

The invention discloses a method for simulating and forecasting flood in a cold highland area based on hydrological and hydrodynamic coupling, and belongs to the technical field of disaster forecasting. The method specifically comprises the following steps: S1, multi-source basic data collection and preprocessing: collecting multi-type and multi-scale basic data for a target cold and cold mountainous area drainage basin; and S2, deep learning correction and fusion of the satellite rainfall data: aiming at the local overestimation and underestimation problems of the satellite rainfall data, a deep learning algorithm is adopted to carry out hour scale correction and fusion. Four types of core data of satellite remote sensing, reanalysis, ground observation and geographic space are collected, total factors of'rainfall-runoff-terrain-underlying surface 'required by flood simulation in the cold and cold mountainous area are covered, simulation one-sidedness caused by lack of data types in traditional modeling is avoided, rainfall and runoff abnormal values are eliminated by adopting a 3-sigma criterion, data formats and spatial-temporal scales are unified, and the modeling efficiency is improved. A standardized data set is formed, and interference of abnormal values, format incompatibility and space-time mismatching on subsequent model input is avoided.

The invention belongs to the technical field of hydrology and water resources, and discloses a cold and dry mountainous area drainage basin runoff simulation method based on an MIKE SHE / MIKE 11 coupling model in combination with GIS / RS data. According to the multi-source data fusion technology, a cold and dry mountainous area distributed input data set is constructed through TRMM satellite rainfall data, an MODIS LAI product and an ASTER DEM, and the rainfall space representation precision is improved by 40%. According to the model coupling mechanism, an MIKE SHE / MIKE 11 collaborative simulation framework is provided, earth surface-groundwater flow data interaction is achieved through river channel linkage, and the complex river network flood routing simulation error is reduced to 10% or below. According to the parameter partition calibration method, a drainage basin is divided into five hydrogeological regions, sensitive parameters such as a Manning coefficient and a hydraulic conduction coefficient are optimized in combination with a genetic algorithm, and the basic flow simulation precision is improved by 25%.

The invention discloses a short temporary rainfall intensity prediction and identification method based on space-time composite attention, and the method comprises the following steps: obtaining an IMERG satellite rainfall data set, and carrying out the data construction and marking of a target region; establishing a space-time composite attention network, accessing an MFGA module and a TSMA module layer by layer, and performing non-negative activation on output to ensure physical interpretability; proposing a C-DWABP loss function, performing direction asymmetrical penalty aiming at a pixel-level prediction error, and introducing a grading weight mapped according to a real rainfall intensity grade at the same time; based on the output, intensity classification is carried out through a microthresholding ordered layer, and learnable threshold stability training is set; generating a forecast product, and obtaining an hour-by-hour precipitation grid in future 1-6 hours and a corresponding intensity identification result; according to the method, the precision of short temporary rainfall intensity prediction is improved, the false alarm rate is reduced, and a heavy rainfall event is effectively identified.

The invention discloses a satellite precipitation product improvement method, device and system based on a dual-time scale minimum error, and the method comprises the steps: carrying out the improvement of a first satellite precipitation product, a second satellite precipitation product and an interpolation precipitation product according to the obtained first satellite precipitation product, second satellite precipitation product and interpolation precipitation product; sequentially adopting a dual tool variable algorithm, a least square fusion algorithm and an inverse error weighting algorithm to respectively calculate first fusion rainfall products under a daily scale and a monthly scale; and fusing the first fused precipitation products under the daily scale and the monthly scale to generate a second fused precipitation product. Ground observation data are not needed, daily scale and monthly scale rainfall errors can be minimized at the same time, the thought of data fusion is incorporated, the performance of satellite rainfall products is enhanced, and detailed data support can be provided for natural disaster monitoring and early warning and regional water resource management.

The invention discloses a multi-source rainfall measurement product data fusion method, system and device and a medium, and belongs to the technical field of meteorological data processing, and the method comprises the steps: obtaining meteorological satellite rainfall measurement data, weather radar rainfall measurement data and ground rain gauge rainfall measurement data, and constructing a multi-source rainfall data set; performing preprocessing through the multi-source rainfall data set to obtain satellite rainfall data with a unified coordinate system and resolution, radar rainfall data after noise reduction and rainfall gauge rainfall data after interpolation completion; and in combination with the preprocessed multi-source rainfall data, constructing a time-space consistency data set. According to the method, meteorological satellite, radar and rain gauge data are integrated, a multi-source rainfall data set is constructed, a coordinate system and resolution are unified, noise reduction interpolation ensures time-space consistency, a Kriging interpolation method is adopted to fuse data, resolution and accuracy are improved, an evaluation model is subjected to cross validation, and accurate support is provided for weather forecast and disaster prevention and reduction.

The invention discloses a satellite precipitation two-stage error correction method based on an actual measurement site, and the method comprises the steps: extracting drainage basin DEM data, obtaining a mask file of a drainage basin, and extracting the satellite precipitation of the drainage basin; performing inverse distance weighted interpolation on the rainfall data of the actual measurement site to a corresponding resolution to obtain grid data of the actual measurement site; error correction data are obtained through a deep learning model or dynamic quantile mapping; calculating a residual error between the rainfall data after error correction and grid rainfall of the actual measurement site, and learning a nonlinear relationship between the rainfall data and the residual error by using a gradient lifting regression tree model; and adding the precipitation residual error of the nonlinear residual error correction model to the error-corrected satellite precipitation error to obtain a residual error-corrected precipitation product. According to the method, the satellite precipitation product is corrected based on the actual measurement site data, the high-precision precipitation product is obtained, point-to-surface conversion of the precipitation data can be realized, and data support is provided for input of a refined hydrological model.

The invention provides a runoff prediction method based on multi-source satellite rainfall data fusion and a deep learning optimization model. The runoff prediction method comprises the following steps: collecting rainfall data and hydro meteorological data of a multi-source satellite product in a research basin; carrying out noise reduction processing on the precipitation data of the multi-source satellite product by adopting Kalman filtering; and fusing the multi-source satellite product rainfall data after noise reduction processing by adopting a Bayesian model average algorithm. Constructing a deep learning runoff prediction model with adaptive periodic position coding, taking runoff data in hydro meteorological data as supervision, training and testing the runoff prediction model by using the fused multi-source satellite product rainfall data, and performing hyper-parameter optimization by adopting a Newton-Raphson optimization algorithm in the training process; and performing runoff prediction by using the trained runoff prediction model. According to the method, the problems that in existing runoff forecasting model fusion, static weighting cannot describe the change of model performance along with time and time-varying errors are ignored are solved, and the accuracy of input data is improved.

The invention provides a satellite-ground cross-scale multi-source precipitation three-stage fusion method. The method comprises the following steps: extracting DEM data and a drainage basin surface file of a drainage basin; local and global feature factors such as topography and meteorology of the drainage basin are obtained; extracting original satellite rainfall data of the watershed; a Kriging interpolation method is used for unifying the spatial resolution of the rainfall data of the multiple original satellites; constructing a hybrid geographically weighted regression model based on a kernel function, and establishing a mapping relationship between the multiple pieces of satellite rainfall data and global and local feature factors; calculating to obtain a plurality of primary fused precipitation products; and integrating a plurality of precipitation products fused once by using a Bayesian triangular cap method to finally obtain a high-precision and high-resolution precipitation fusion product obtained through three-stage and two-time fusion. Multi-source precipitation fusion is optimized through a refined data processing technology, the precipitation data precision and applicability are improved, high-quality data support is provided, and the method is suitable for different hydrological and meteorological research scenes.

The invention discloses a radar satellite precipitation rapid fusion algorithm based on a radar quality index, and the algorithm comprises the steps: designing a radar shielding area index RSA as a piecewise function changing with a climate proportion correction factor RFC through employing the climate proportion correction factor RFC, optimizing the weight of a radar quality index analysis field through employing a radar terrain shielding index analysis field, and carrying out the calculation of the radar quality index. And rapidly fusing radar and satellite rainfall data by using the radar quality index fusion weight to generate a radar-satellite fusion background field. According to the method, the radar quality index and the terrain shielding coefficient are utilized to dynamically quantify the spatial credibility of the radar data, and the fusion weight of the radar and the satellite is optimized, so that the fusion background field can reflect the actual rainfall condition more accurately, especially in a radar coverage blind area and a site sparse area, the real rainfall condition can be reflected more accurately. And the problems of local distortion, rainfall intensity underestimation or structure fuzziness and the like are effectively reduced.

The invention provides a complex watershed short-time flood response intelligent forecasting method based on satellite rainfall perception, and relates to the technical field of hydro-meteorological monitoring and disaster prevention and reduction. The method comprises the following steps: acquiring satellite rainfall data of a target drainage basin observed by a multi-channel satellite, and preprocessing the satellite rainfall data; inputting the preprocessed satellite precipitation data into a short-time precipitation prediction model, and outputting a short-time satellite precipitation prediction result; and inputting the short-time satellite precipitation prediction result into a semi-distributed hydrological model, simulating and generating a flood response process of a watershed scale, and outputting predicted flood process data. According to the method, the precision problem of short-time rainfall prediction of the complex drainage basin is solved, and the timeliness and stability of flood prediction are improved.

This invention discloses a method for downscaling satellite precipitation based on measured stations. It extracts watershed DEM data, including satellite precipitation, wind speed, and temperature. Weighted interpolation of precipitation data from measured stations yields grid data. A deep learning model learns the nonlinear, spatiotemporal coupling relationship between multiple factors and precipitation distribution, and predicts low-resolution satellite precipitation using dynamic weights. High-resolution satellite precipitation is predicted by upsampling satellite precipitation, temperature, wind speed, and station precipitation using a deep learning model and dynamic weights. The residual between the low-resolution precipitation and the measured station precipitation is calculated and upsampled. The high-resolution satellite precipitation is then added to the sampled residual to obtain the downscaled precipitation product. This invention downscales satellite precipitation products based on measured station data to obtain high-resolution precipitation products, achieving refined precipitation data and providing data support for inputting refined hydrological models.

The invention discloses a near-real-time satellite precipitation error correction method and system. The method comprises the following steps: performing spatial-temporal resolution unified processing on an obtained daily satellite precipitation product, a daily interpolation precipitation product and a monthly interpolation precipitation product; in each grid unit, estimating a relative error of a daily interpolation precipitation product and a daily satellite precipitation product at each moment; based on the relative error, the daily satellite precipitation product and the daily interpolation precipitation product are split into corresponding overestimation error data pairs, zero error data pairs and underestimation error data pairs; with the minimum error as the target, the overestimation error data pair, the zero error data pair and the underestimation error data pair are fused, and a daily scale corrected precipitation product is obtained; and based on a total water quantity control mode, the distribution characteristics of the rainfall products corrected by the daily scale are corrected by using the monthly interpolation rainfall products, and corrected satellite rainfall products are obtained. According to the method, the error and the distribution mode of near-real-time satellite precipitation are calibrated through error decomposition and total water volume control.

A method for back-calculating satellite precipitation data in mountainous areas includes: reading APHRODITE and IMERG precipitation data and calculating daily precipitation statistics at observation stations; building a model using a random forest algorithm at a 0.25° scale to simulate the long-term trend of the two types of precipitation data; generating spatial disaggregation weights for precipitation data at a 0.1° scale using a 3×3 sliding window; and combining the temporal and spatial scales to obtain the 0.1° IMERG precipitation data before 2001, which is the NIMERG data. This method utilizes the temporal and spatial relationships between APHRODITE reanalysis precipitation products and IMERG satellite precipitation products over the same timeframe to reversely back-calculate and extend the IMERG satellite precipitation data. This method can provide a set of high-quality, long-term precipitation products for research related to mountain hydrology, ecology, or geological hazards.

This application provides a satellite precipitation error correction and fusion method based on elevation gradient under complex terrain, belonging to the field of meteorological data processing technology. The method includes: acquiring multi-source precipitation observation data of complex terrain areas, obtaining gridded precipitation data and inputting it into an elevation-error correlation model, obtaining an error intensity distribution map and identifying grid areas with error intensity values ​​greater than or equal to a preset error threshold as uncertain areas and performing feature extraction, obtaining regional feature vectors and inputting them into an error correction strategy selection model, determining the correction strategy type and applicability of the uncertain areas, obtaining the correction areas that need to be deeply fused from the uncertain areas and obtaining ground verification point data, calculating the local precipitation true field and error spatial structure parameters; using a correction algorithm to correct the satellite precipitation data of the correction areas, fusing the satellite precipitation data of the correction areas with those of the uncorrected areas, and generating precipitation products and uncertainty assessment reports.

This invention discloses a real-time precipitation retrieval method for meteorological satellites based on a deep learning geographic attention mechanism. The method includes: constructing a training dataset of FY-4B AGRI infrared cloud images and IMERG Final Run labels, preprocessing it, and inputting it into a U-shaped encoder-decoder network; achieving deep fusion of multi-scale features by embedding a GeoAB geographic attention module; using a dual-objective loss function to complete rain area classification and precipitation intensity regression training; and outputting precipitation data with 0.05° spatial resolution and 15-minute temporal resolution through T-transform decomposition of the cloud image, independent inference, T-inverse transform fusion, and average pooling to eliminate grid effects. Compared with existing technologies, this invention solves the technical problems of balancing real-time performance and accuracy in satellite precipitation retrieval, and the blurred distinction between rain areas and clear skies. This invention achieves a CSI of 0.693, a 3.3% improvement over IMERG Final Run, with a total retrieval time of less than 20 seconds. It balances high accuracy and real-time performance, providing data support for disaster early warning, agricultural management, and other applications, and has promising application prospects.

The invention discloses a possible maximum flood calculation method suitable for a data-free area, and belongs to the field of drainage basin water circulation simulation and hydrological design. The method comprises the following steps: calculating a daily scale possible maximum rainfall result of a drainage basin of the data-free area by adopting an improved Hirschfield method; classifying and processing possible maximum rainfall calculation problems of areas without data; for small and medium-sized drainage basins with the drainage basin area smaller than or equal to 1000 km < 2 >, the 24-hour hour-by-hour possible maximum rainfall process is deduced in the mode that frequency calculation is coupled with a Chicago rainfall pattern on the basis of satellite rainfall inversion data; for a large watershed with the watershed area larger than 1000 km < 2 >, the possible maximum rainfall process of the next day scale of several days is deduced based on satellite rainfall inversion data in a same-frequency amplification mode. According to the method disclosed by the invention, a technical path with complete calculation is constructed, a more reasonable and reliable possible maximum flood calculation result of the watershed in the non-data region is obtained, and key and powerful theoretical guidance and technical support are provided for scale argumentation of a flood control system of water conservancy and hydropower engineering in the non-data region.

A method to train a diffusion regression model for satellite-based precipitation data bias correction may include obtaining satellite observation precipitation data for a training geographic region. The method may include obtaining corresponding ground observation precipitation data for the training geographic region, the corresponding ground observation precipitation data having a same resolution as the satellite observation precipitation data. The method may include generating training residuals from the satellite observation precipitation data and the corresponding ground observation precipitation data. The method may include adding noise to the training residuals. The method may include de-noising the noisy training residuals using a diffusion regression model to generate predicted residuals. The method may include updating the diffusion regression model using a loss function that depends on the training residuals and the predicted residuals.

This invention relates to the field of precipitation fusion technology and discloses a multi-source precipitation data fusion method based on metaheuristics and machine learning, comprising the following steps: collecting ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data, and performing data preprocessing; dividing the preprocessed ground precipitation observation data, multi-source satellite precipitation product data, and auxiliary variable data into training and validation sets according to a certain ratio; setting the initial value range of hyperparameters of a multilayer perceptron model, and inputting them along with the training set into the multilayer perceptron model for training; optimizing the hyperparameters of the multilayer perceptron model using a sparrow search method to obtain a trained multilayer perceptron model; inputting the validation set into the trained multilayer perceptron model to identify the precipitation observation values ​​of the validation set, thereby obtaining the predicted ground precipitation observation data, i.e., the fused precipitation data; this method improves the accuracy of fused precipitation data while avoiding reliance on assumptions that are not valid in reality.

The application discloses a hydrological and hydrodynamic coupling alpine mountain flood simulation and prediction method, and belongs to the technical field of disaster prediction. The method specifically comprises the following steps: S1, multi-source basic data collection and preprocessing: for the target alpine mountain basin, collect multi-type and multi-scale basic data; S2, deep learning correction and fusion of satellite precipitation data: for the local overestimation and underestimation problems of satellite precipitation data, a deep learning algorithm is used for hour-scale correction and fusion. By collecting satellite remote sensing, reanalysis, ground observation and geographic space four kinds of core data, covering all factors of 'precipitation-runoff-terrain-underlying surface' required for alpine mountain flood simulation, avoiding the one-sidedness of traditional modeling caused by the lack of data types, using 3 sigma criterion to remove precipitation and runoff outliers, unifying data format and space-time scale, forming a standardized data set, avoiding the interference of abnormal values, incompatible format and space-time mismatch on subsequent model input.

The invention provides a method and device for determining the irrigation water consumption of an arid region, and relates to the technical field of agricultural irrigation, and the method comprises the steps: firstly obtaining various evapotranspiration data, ground observation rainfall data, satellite observation rainfall data and soil environment characteristic data of the arid region in an irrigation period; carrying out weighted fusion on the plurality of evapotranspiration data to obtain first data; correcting the satellite observation rainfall data pixel by pixel based on the ground observation data to obtain second data; inputting the soil environment characteristic data into a first deep learning model to obtain first non-evaporation loss data of the arid region in the irrigation period; then correcting the first non-evaporation loss data to obtain second non-evaporation loss data; and finally, based on the first data, the second data and the second non-evaporation loss data, determining the irrigation water consumption of the arid region in the irrigation period. According to the method for determining the irrigation water consumption of the arid region, accurate detection of the irrigation water consumption of the arid region is achieved.

The application provides a complex basin short-time flood response intelligent prediction method based on satellite precipitation perception, and relates to the technical field of hydrological meteorological monitoring and disaster prevention and mitigation. The method comprises the following steps: acquiring satellite precipitation data of a target basin observed by a multi-channel satellite, and performing pretreatment; inputting the pretreated satellite precipitation data into a short-time precipitation prediction model to output a short-time satellite precipitation prediction result; and inputting the short-time satellite precipitation prediction result into a semi-distributed hydrological model to simulate a flood response process at a basin scale and output predicted flood process data. The application solves the precision problem of short-time precipitation prediction of a complex basin and improves the timeliness and stability of flood prediction.

The present invention discloses a method for simulating the installation of virtual rain gauges based on satellite precipitation data. The method comprises the following steps: obtaining potential installation points based on existing initial observation sites, calculating the probability of continuous precipitation variables taking values ​​in different intervals between the initial observation sites and the potential installation points; calculating the weighted cosine similarity of the distribution function of the value probability; calculating the information entropy between the initial observation points and the potential installation points; and selecting potential installation points that meet the requirements as virtual rain gauges based on the distribution characteristics and weighted cosine similarity of the satellite precipitation information entropy at different site locations. The present invention addresses the problems of the inconsistency between accuracy and coverage, as well as insufficient data processing efficiency, in existing precipitation information acquisition technologies, thereby providing more accurate, comprehensive, and timely precipitation information support.

This invention provides a precipitation product correction method based on climate adjustment, combining dry and wet season division and weight allocation. The method calculates adjustment amounts and scaling factors based on historical precipitation data of the region, generates a deviation adjustment value using the adjustment amount AA and scaling factor SF, and applies the deviation adjustment value to the corresponding IMERG precipitation estimate to obtain an adjusted and corrected precipitation estimate. This improves the accuracy of precipitation estimate correction, effectively corrects errors in satellite precipitation data, and enhances the reliability of satellite precipitation data.

The invention provides a ground snowfall identification method based on an FY3G satellite rainfall measurement radar, and the method comprises the steps: carrying out the equiangular cylindrical projection, bilinear interpolation and Kriging filling operation of L1-level preprocessing data and L2-level product data of the radar, and obtaining equal-latitude and equal-longitude projection data; filtering the radar reflectivity factor vertical profile of the equal latitude and longitude projection data based on a three-factor threshold, and then carrying out spatial registration and cubic spline interpolation dual-frequency profile data calculation to obtain a dual-frequency ratio profile; and according to a double-frequency-ratio profile slope calculated by a regression algorithm, and according to a maximum reflectivity value and a storm top height on the profile, a snowfall index is obtained, so that snowfall identification is carried out on the screened to-be-calibrated area. According to the method, through L1-level pre-processing data and L2-level product data of the radar, in combination with technologies of data threshold filtering, factor discrimination and the like, effective identification of large-range ground snowfall is realized, and the spatial coverage and precision of snowfall identification are improved.

The application discloses a Gaussian transformation and mixed four-dimensional variation assimilation method for instantaneous precipitation rate of a stationary satellite, constructs a FY-4 stationary meteorological satellite precipitation rate observation operator and a mixed four-dimensional variation assimilation scheme, establishes a Gaussian transformation and error optimization statistical method meeting different precipitation rate data requirements, introduces high-temporal and high-spatial resolution stationary satellite precipitation rate data into a regional model mixed four-dimensional variation assimilation system, and improves analysis and prediction effects of the model on a typhoon. Meanwhile, two selectable Gaussian transformation schemes meeting different requirements and a precipitation rate observation error scheme which changes adaptively with rain intensity are established in the assimilation system, and the problems that the assimilation effect of precipitation physics in the model is weakened due to the strong non-Gaussian distribution characteristics of the precipitation physics are effectively improved.

The embodiment of the invention provides a consistency evaluation method, device and equipment based on satellite-ground rainfall detection products, and is applied to the technical field of atmospheric detection. The method comprises the following steps: acquiring foundation rainfall data and a satellite rainfall detection product; based on a preset identification conversion relation, obtaining a ground precipitation identification product according to the foundation precipitation data; performing time matching and space matching on the satellite rainfall detection product and the ground rainfall identification product to obtain a satellite-ground rainfall detection product; the satellite-ground rainfall detection products are fused according to stations and time; and based on a preset evaluation index and a preset evaluation dimension, carrying out consistency evaluation on the fused satellite-ground rainfall detection product. In this way, data can be observed by using the ground rain gauge, authenticity inspection can be carried out through matching and analysis of satellite-ground data, consistency analysis and evaluation work of satellite-ground rainfall detection products can be carried out, and support is provided for manufacturing and application of satellite-ground rainfall fusion products.

The application provides a ground snow identification method based on FY3G satellite precipitation measurement radar, comprising: performing equiangular cylindrical projection, bilinear interpolation and Kriging filling operation on L1 level pretreatment data and L2 level product data of the radar to obtain equi- latitude and longitude projection data; filtering the vertical profile of the radar reflectivity factor of the equi- latitude and longitude projection data based on three-factor threshold value, then performing spatial registration and three spline interpolation double-frequency profile data calculation to obtain a double-frequency ratio profile; then calculating the snowfall index according to the slope of the double-frequency ratio profile calculated by a regression algorithm and according to the maximum reflectivity on the profile and the storm top height, so as to identify the snowfall of the screened calibration area. The method realizes effective identification of large-scale ground snowfall by combining the L1 level pretreatment data and the L2 level product data of the radar with data threshold filtering and factor discrimination technology, and improves the spatial coverage and accuracy of snowfall identification.

The invention relates to the technical field of meteorology and climatology, in particular to a watershed hydro meteorology prediction method, which comprises the following steps of: performing multi-dimensional visual angle evaluation on a plurality of candidate satellite rainfall data of a watershed, and performing weight parameter construction on visual angle evaluation indexes of the obtained candidate satellite rainfall data; in combination with view angle evaluation indexes and weight parameters, a comprehensive evaluation system covering multiple dimension view angles such as spatial-temporal variability, extreme rainfall events, extreme rainfall intensity, rainfall probability density distribution and uncertain analysis characteristics is constructed, and the performance of multiple candidate satellite rainfall data of a drainage basin is comprehensively analyzed; a more targeted decision basis is provided for hydrometeorological prediction of the watershed, and the accuracy of hydrometeorological prediction is improved.

The application discloses a satellite precipitation data correction method with a self-adjusting window, comprising the following steps: S1, using satellite grid data and ground observation station measured precipitation data, calculating the ground observation station density in each grid of the satellite grid data; S2, combining the ground observation station density in each grid, determining the optimal correction window width corresponding to the distribution state of each ground observation station; S3, setting an initial window based on the optimal correction window width, moving the initial window to successively correct the satellite grid data in each window, obtaining the function relationship between the satellite grid data and the ground observation station measured precipitation data, and inputting the satellite grid data into the function relationship to obtain the corrected satellite grid data. The method can consider the space-time distribution characteristics of the satellite precipitation data, take the ground observation station measured precipitation data as a reference, locally correct the grid data, reduce the correction error, and make the correction result have relatively high precision.

The application discloses a short-term precipitation intensity prediction and identification method based on space-time composite attention, comprising the following steps: acquiring an IMERG satellite precipitation dataset, and constructing and labeling data of a target region; a space-time composite attention network is built, MFGA modules and TSMA modules are sequentially connected, and non-negative activation is outputted to ensure physical interpretability; a C-DWABP loss function is proposed, a direction asymmetric penalty is carried out on pixel-level prediction error, and a hierarchical weight mapped according to real rain intensity is introduced; intensity classification is carried out based on output through a differentiable thresholding ordered layer, and a learnable threshold is set to stabilize training; a prediction product is generated, and future 1-6 hour precipitation grids and corresponding intensity identification results are obtained; the application improves the precision of short-term precipitation intensity prediction, reduces the false alarm rate, and effectively identifies strong precipitation events.