1309results about "Rainfall/precipitation gauges" patented technology

The invention belongs to the technical field of rainfall prediction, and discloses a short and temporary rainfall prediction method based on multi-model random scheduling integration, which develops a robust training and pushing framework based on a continuous rolling prediction strategy, and decomposes long-sequence prediction into manageable stages. According to the method, training is carried out through teacher forcing and planned sampling, error propagation is relieved, and the training process is stabilized. The invention further designs asymmetric encoder-decoders (DSE and AFD) that achieve lower FLOPs than competitive baselines under standardized assessment, where DSE selectively compresses significant features and AFD stepwise reconstructs details to mitigate excessive smoothing problems. Finally, an intensity weighted Gaussian KL divergence loss function is designed, and the key problem of data balance is solved by modeling and predicting on a distribution level and endowing a large weight to a meteorological important heavy rainfall event.

The invention provides an urban road ponding monitoring and drainage system. The urban road ponding monitoring and drainage system comprises a monitoring module which collects ponding depth, water velocity, rainfall intensity and field image data in real time; the initialization configuration module sets parameters such as an accumulated water depth threshold value; the data transmission module encrypts and transmits data to the central control module; the model algorithm optimization module constructs a three-dimensional ponding diffusion model and outputs a predicted ponding depth and an optimized drainage strategy; the central control module generates a drainage pump rotating speed instruction and the like; the drainage execution module dynamically adjusts drainage pump power and a drainage path; the self-checking and fault-tolerant module calculates a health index of drainage equipment and switches redundant equipment; the early warning module sends risk level information; and the interaction module displays information such as a real-time accumulated water thermodynamic diagram. The urban road drainage efficiency can be improved, the water accumulation risk is reduced, and smooth urban traffic and public travel safety are guaranteed.

The invention relates to the technical field of building construction, in particular to a foundation pit dewatering prediction method based on a digital twinborn technology, which comprises the following steps of: 1, establishing a multi-source data fusion layer; 2, establishing a digital twin modeling layer; 3, establishing an intelligent prediction algorithm layer; 4, establishing a dynamic optimization decision layer, and generating a precipitation well optimization scheme based on a prediction result; according to the method, engineering geological data of a geological survey report before excavation of the building foundation pit, historical meteorology of a building location and hydrogeological data are utilized, digital twinning is carried out on precipitation during the whole construction period of the proposed building foundation pit on the basis of a digital twinning technology, the precipitation period, daily precipitation and influence of precipitation on foundation pit supporting are predicted, and the foundation pit supporting effect is improved. According to the method, the influence of settlement deformation of surrounding buildings is reduced, digital simulation is provided for smooth construction of a whole project, and measurement and prediction are provided for the project and the surrounding environment, so that the construction cost is greatly reduced.

A computer-implemented method includes executing, via a flood simulation model, a computer simulation that simulates flooding within a target spatial area based on an input of a geospatial dataset and the rainfall intensity data of one or more flooding events associated with the target spatial area; determining normalized inundation scores for a set of geographical cells and one or more clusters of interconnected geographical cells based on executing the computer simulation; determining, via an optimization algorithm, a set of optimal sensor locations in the target spatial area based on the normalized inundation scores and the one or more clusters of interconnected geographical cells; and generating a sensor location map that includes the set of geographical cells and a set of geospatial markers identifying the set of sensor locations; and generating a visual output that displays the sensor location map.

The invention provides a GNSS-based mountain area highway loess slope stability evaluation method and system, and relates to the technical field of loess slope stability evaluation.The method comprises the steps that firstly, a slope area is divided into grids, geologic feature data of each grid are obtained, super nodes are divided according to the geologic feature data, and the super nodes are divided into super nodes; monitoring points are determined, and the soil moisture content is monitored, so that the moisture content of the soil at different depths is analyzed; secondly, the original displacement is monitored through a GNSS, displacement correction is carried out based on the Fredld unsaturated soil shear strength theory, and the actual displacement is obtained; the influence of the accumulated rainfall on the soil body strength parameters is analyzed, and the soil body strength parameters are corrected in real time; coupling the corrected soil body strength parameter with the actual displacement, calculating a safety coefficient, and judging an early warning state according to the rainfall sensitivity coefficient; and finally, comprehensively judging the stability of the slope according to the early warning state. According to the method, through multi-parameter comprehensive analysis, the accuracy of slope stability evaluation is improved.

The invention provides a runoff forecasting method and device based on soil humidity calibration, and relates to the technical field of forecasting. The method comprises the steps of generating weather forced data of a WRF mode based on an atmosphere reanalysis data set under the condition that a target area has no runoff observation data; acquiring an observed value of surface soil humidity of a target area in the external meteorological data set; inputting the meteorological forced data into a distributed hydrological model WRF-Hydro model coupled with the WRF mode, and performing simulation to obtain a simulation value of the surface soil humidity of the target area; calibrating a key parameter set of the hydrological model according to the difference between the observed value and the simulated value of the soil humidity and a historical flood event of the target area to obtain a general parameter set of the target area; and using the general parameter set to drive a WRF-Hydro model to carry out runoff forecasting on the target area. Based on the scheme, runoff forecasting can be accurately carried out on the area without runoff observation data.

This invention discloses a method for optimizing the layout of a rain gauge network in mountainous areas based on radar rainfall data, which includes Step 1: collecting rainfall data and dividing the target area into meshes, Step 2: identifying abnormal rain gauge station locations, Step 3: determining candidate locations for additional rain gauge stations, Step 4: evaluating the accuracy and effectiveness of different rain gauge network optimization methods, and Step 5: determining the optimal rain gauge network layout method. [Effects] The method of the present invention, combined with radar rainfall data, realizes effective planning and optimization of rainfall observation networks in complex mountainous terrain conditions, further improves the accuracy and reliability of precipitation observation in mountainous watersheds, and provides more accurate data support for the management and prevention of mountainous flood disasters.

The invention discloses a high-resolution peak accumulated snow depth inversion device and method based on a geographically weighted random forest, and the device comprises a data obtaining module which is used for obtaining the multi-source data of a research region, and the multi-source data comprise MODIS remote sensing accumulated snow data, Sentinel-1VV / VH backscattering data, topographic data, and meteorological station snow depth observation data; the high-resolution snow coverage data set construction module is used for processing the MODIS remote sensing snow data to obtain a high-resolution cloudless binary snow coverage data set; the feature construction module is used for constructing a comprehensive feature set containing a snow space-time process feature index, a back scattering feature index and a topographic feature index; and the model construction and analysis module is used for training a geographically weighted random forest model through the comprehensive feature set, and inverting the peak snow depth of the research area according to the optimal model. According to the invention, estimation of high-resolution peak snow depth can be realized.

The disclosure relates generally to methods and systems for ground water prediction using integration of satellite and auxiliary observations. Conventional techniques in the art for the ground water level prediction are not accurate and efficient for predicting the ground water level. The present disclosure combines time-series archived remote sensing data with a wide array of past auxiliary datasets specific to a given region for precise prediction of future ground water condition, to predict the ground water situation i.e. rising or declining for a given region. These encompass diverse information such as weather conditions, soil properties, agricultural data, population statistics, water resources information, industrial details, drought records, seismic data, and surface deformation patterns. The present disclosure also considers different future scenarios which can be built on future cropping patterns, forecasted weather conditions, use of irrigation system in future, and future population growth of the given region.

The invention relates to the technical field of disaster early warning, and discloses a rainstorm early warning method based on multi-source forecast product dynamic fusion, which comprises the steps of collecting multi-mode rainfall forecast data, constructing a rainfall cumulative distribution function, performing dynamic updating based on Kalman filtering, and obtaining frequency correction rainfall forecast products of each mode; historical forecast and live monitoring data samples are obtained, comprehensive weight coefficients corresponding to all levels of rainfall are calculated, and a fusion initial rainfall distribution field is obtained; performing matching reconstruction on the corrected rainfall forecast field to generate a reconstructed rainfall distribution field; collecting latest real-time monitoring data, and carrying out space-time dynamic correction on the reconstructed rainfall distribution field to generate a real-time rainstorm potential field; and comparing the real-time rainstorm potential field with the multistage rainstorm early warning threshold, generating rainstorm early warning information, and issuing the rainstorm early warning information. According to the invention, the timeliness and accuracy of rainstorm early warning response are improved, the public personal safety is guaranteed, and the city emergency disposal capability is improved.

The invention discloses a microwave link-weather radar combined rainfall inversion method based on Z-R relation adaptive adjustment, and the method comprises the steps: defining a related parameter set of microwave link rainfall estimation and weather radar reflectivity measurement in observation time, and carrying out the matching and space-time averaging processing, and obtaining rainfall intensity-radar reflectivity characteristics; fitting a radar reflectivity-rainfall intensity relation within observation time by using a least square method based on the rainfall intensity-radar reflectivity characteristics; according to the radar reflectivity-rainfall intensity relation sample of the known position, calculating the spatial distribution of the radar reflectivity-rainfall intensity relation in the research area by using an inverse distance weighted interpolation method; and according to the spatial distribution of the radar reflectivity-rainfall intensity relationship within the observation time and the average reflectivity factor measured by the radar, calculating to obtain a spatial distribution result of the accumulated rainfall within the observation time. According to the invention, the accuracy of regional rainfall monitoring is effectively improved.

The invention discloses a multi-source data fusion method based on a water conservancy rain measuring radar, and belongs to the technical field of meteorological observation. Historical rainfall inversion, short temporary rainfall forecast and short-term rainfall forecast are realized based on multi-source data fusion, and a space-air-ground integrated rainfall observation and prediction system is constructed. According to the method, innovative technologies such as multi-source data, multi-parameter inversion, AI intelligent optimization and manual correction are fused, the limitation of a traditional rainfall inversion method is broken through, and the precision and service efficiency of quantitative rainfall estimation are remarkably improved; through multi-modal fusion and full-link integration, the method has significant technical advancement and business value in application in the fields of water conservancy and meteorology.

An early warning method for debris flow disaster in a small watershed, and a disaster reduction method for debris flows in a small watershed is disclosed. An early warning method for debris flow disaster in a small watershed is provided, in which parameters of a hydrological model are calibrated by rainfall, evaporation and runoff data of the watershed based on water balance of the watershed, and an early monitoring and warning scheme for debris flows, which takes water storage in the watershed as a core monitoring and evaluation indicator, is established. The GR4J hydrological model is improved by taking into account a watershed area, specific yield of an aquifer and several geological parameters, as well as an evapotranspiration effect of vegetation. A disaster reduction method by means of regulation of a drainage channel can provide a basis for mathematical model simulation studies of disaster reduction in small watersheds.

The invention provides an intelligent flood forecasting method and device based on an FQI-DMTSNet self-adaptive generative adversarial mechanism and a readable storage medium thereof.According to the method, data are preprocessed through multi-source data fusion and assimilation, time sequence features are extracted through a dynamic hierarchical multi-scale time sequence sensing network (DMTSNet), a scheduling strategy is optimized by combining fitting value iteration (FQI) reinforcement learning, and the real-time flood forecasting method based on the FQI-DMTSNet self-adaptive generative adversarial mechanism is obtained. And data enhancement and data-free area parameter migration are realized by means of an adaptive feature-gated generative adversarial network (AFG-GAN), so that a closed-loop collaborative forecasting system is formed. The problems that a traditional model is high in dependence on data, poor in extreme scene adaptability and the like are solved, data-free area parameter generation, extreme scene sample enhancement and regulation and control strategy optimization are achieved, the accuracy, adaptability and practicability of flood forecasting are improved, and the method is suitable for flood forecasting, hydrological dispatching, urban inland inundation early warning and other scenes.

The present invention relates to a high temperature and drought composite disaster monitoring and early-warning method and system, belonging to the technical fields of disaster risk assessment and early warning. Internal correlation features and abnormality information of high temperature and drought events are input into a model, multi-time-space scale features of the high temperature and drought events can be identified accurately, high event identification accuracy and space resolution are achieved, the progress can be predicted progressively, the drought and high temperature threshold change can be monitored closely, and fine forecasting and early warning can be performed in different periods, regions and intensities, thereby ensuring that indicators are in the same time scale, avoiding the complication of the high temperature and drought process caused by frequent time and space discontinuities of the indicators in a single point or small region, and ensuring the suitability for any periods of the process.

The invention provides a hydrological forecasting method and system based on an adaptive correction chain, and relates to the field of data processing, and the method comprises the steps: generating a first reservoir inflow forecasting result according to a rainfall deviation traceability-dynamic correction chain and rainfall observation data; multi-dimensional features of the current scene are determined, and a second reservoir inflow forecasting result is generated according to the historical similar sample matching-rule application chain and the multi-dimensional features of the current scene; according to the confluence and production knowledge-threshold constraint chain, the underlying surface condition, the rainfall spatial distribution and the basin state, a third reservoir inflow forecasting result is generated; and determining the dynamic weights of the first reservoir inflow forecasting result, the second reservoir inflow forecasting result and the third reservoir inflow forecasting result, and generating a final reservoir inflow forecasting result, thereby improving the accuracy of the reservoir inflow forecasting result.

The invention provides a rainfall sampler for meteorological monitoring, and relates to the technical field of water sampling, the rainfall sampler comprises a placing box, and a water quality collecting mechanism is arranged in the placing box; the water collecting mechanism comprises a shaft rod, a circular plate, a turntable, an up-down moving part and an annular concave-convex plate; the shaft rod is connected into the containing box, the circular plate is connected with the shaft rod in the containing box, a gear disc is arranged on the shaft rod, arc-shaped grooves are formed in the gear disc and the circular plate, the rotating disc is in threaded connection with the shaft rod, the up-down moving part is slidably connected to the rotating disc and is opposite to the arc-shaped grooves, the annular concave-convex plate is fixedly connected into the containing box, and the up-down moving part annularly rotates on the annular concave-convex plate. The inner top wall of the placing box is communicated with an elastic plugging pipe mechanism, under the assistance of an annular concave-convex plate, up-down moving pieces are inserted into arc-shaped grooves in a circular plate and a gear disc at intervals to control the circular plate to rotate and ascend, when the circular plate rotates, respective sampling of a plurality of sampling bottles is achieved, and when the circular plate ascends, water overflow is avoided under the assistance of the elastic plugging pipe mechanism.

The invention discloses a multi-mode set heavy rainfall forecasting method fusing space loss function deep learning, which comprises the following steps: acquiring rainfall site observation data, meteorological element data and various physical factor data to form multivariate meteorological factor data; processing the multivariate meteorological factor data into equal-resolution lattice point data and preprocessing the lattice point data; screening out meteorological element and physical factor data of which the importance measurement value is greater than a threshold value, and dividing a data set according to research requirements; constructing a mixed loss function fusing precipitation spatial features; a mixed loss function is developed to train the U-NET deep learning neural network model, and the performance of the model is evaluated; and inputting the multi-element meteorological factor data of the multi-mode output real-time forecast into the model, and generating the real-time heavy rainfall forecast of the multi-mode set. According to the method, precipitation space structure characteristics are integrated into a deep learning model, the problems of deep learning forecast averaging and peak loss caused by a traditional point-to-point strength loss function are solved, and the method aims at improving the precision of heavy precipitation forecast.

The invention discloses a rainfall intensity rapid identification method and system based on multi-sensor data fusion, and relates to the technical field of hydro-meteorological monitoring and urban drainage scheduling, and the method comprises the steps: obtaining a historical rainfall data set, drawing a peak curve, and extracting an evolution sequence feature before a peak value; correlation analysis is carried out on the data and complete rainfall process characteristics to establish a historical rule model; rainfall intensity data are collected in real time through multiple sensors, and space-time dimension features are fused; matching the fusion features with historical features, and screening target historical events; constructing a rainfall reproduction prediction model to reconstruct a current rainfall evolution process; and finally, calculating an inflow load based on a reconstruction result, and generating a dynamic drainage scheduling scheme. According to the invention, accurate prediction of the rainfall process and intelligent scheduling of the drainage system are realized, and the urban waterlogging prevention capability is improved.

The invention is suitable for the technical field of rainfall forecast correction, and provides a regional short-term rainfall quantitative forecast machine learning correction method integrated with weather system information, and the method comprises the steps: determining an initial forecast factor based on meteorological elements in an ECMWF mode, the meteorological elements comprise a precipitation field, ground transformation, convection effective potential energy, UV wind fields of 500hPa, 850hPa and 925hPa, relative humidity, temperature and vertical speed; taking the dichotomy labels corresponding to the 500hPa slot line and the tangent line between 850hPa and 925hPa as newly added forecast factors, and constructing a four-dimensional training set according to the initial forecast factors and the newly added forecast factors; converting the four-dimensional training set into a three-dimensional training set; training a random forest regression model through the three-dimensional training set, and outputting a rainfall forecast field within a preset time; according to the CMASH3 mode rainfall forecast data and a preset judgment logic, dynamically correcting the rainfall forecast field within a preset time, and generating a target rainfall forecast product; and the accuracy of heavy rainfall forecasting is effectively improved.

This disclosure relates generally to system and method for estimating crop water requirement using multi-sensor data fusion. Increasing global population is imparting pressure on both agriculture for food demand and limited freshwater resources for consumption. Estimating crop water requirement reduces water demand for crop production. The method divides soil and crop into multiple vertical and horizontal profiles to estimate water balance thereby reducing the errors in estimation of crop water requirement. Additionally, the method has capability to interlink the multiple data sets such as satellite based earth observations, weather observations from IoT sensors, Weather forecasts from global circulation models, and crop knowledge base for crop water requirement estimation. The method is based on spatio-temporal modeling for multi-layer crop and soil water balance and helps to generate the additional insights on crop water requirement like moisture at different levels in soil profile, crop canopy growth at different locations.

The invention discloses a multi-factor coupled mountain area mountain torrent dynamic partition early warning method, belongs to the technical field of mountain torrent defense, and realizes dynamic partition early warning of mountain area mountain torrent disasters by systematically integrating multi-source data, quantifying multi-factor synergistic effect and constructing a differentiated early warning model. The method comprises the following steps: firstly, collecting and preprocessing high-resolution topographic data, multi-source meteorological and hydrological observation data and soil attribute data, dividing hydrological partitions and constructing a basic database; secondly, screening small watershed units with similar parameters by adopting a control variable method, and simulating and analyzing a coupling driving mechanism of rainfall, terrain and soil through a distributed hydrological model; constructing a dynamic partition early warning model based on a multi-factor action rule, and dynamically adjusting an early warning threshold by combining real-time data; and finally, outputting the mountain torrent risk levels of different hydrological partitions, and providing quantitative support for mountain torrent early warning in a mountainous area.

The invention relates to a soil erosion amount dynamic monitoring method and system based on multi-source remote sensing data fusion. The method comprises the following steps: acquiring multi-source remote sensing data and auxiliary data, and performing geometric fine correction and atmospheric correction on an image based on high-resolution digital elevation data to obtain a data set with consistent space; performing inversion based on the data set to obtain a corrected terrain factor and a water and soil conservation measure factor; generating a time sequence rainfall erosivity factor and a time sequence vegetation coverage factor in combination with the multi-temporal satellite rainfall data; and calculating a preliminary soil erosion modulus in combination with the soil type map, and correcting by using multi-stage high-resolution digital elevation data to obtain a dynamic monitoring result of the soil erosion amount. By adopting the method, the problem of multi-source data fusion matching can be solved, and the factor inversion precision is improved, so that the accuracy of a dynamic monitoring result of the soil erosion amount is improved.

This invention relates to a rainfall observation method, apparatus, computer system, and medium, addressing the problem of high-reliability regional rainfall observation. First, reanalysis data and Earth system model output data are used to construct a first deep learning model for reconstructing a gridded long-series rainfall dataset by inverting rainfall from satellite data. Then, historical rainfall attenuation information is used as input, and different rainfall attenuation and intensity models are used to obtain corresponding historical rainfall simulation series. The weighted link method is then used to determine the regional rainfall under different rainfall attenuation and intensity models, and the optimal rainfall attenuation and intensity model is selected from the long-series gridded rainfall data. Further, a second deep learning model is constructed using satellite inversion data and rainfall data inverted from the rainfall attenuation and intensity model to achieve real-time correction of regional rainfall. Finally, the optimal rainfall attenuation and intensity model is used with real-time mobile signals as input to deduce real-time regional rainfall information and obtain the real-time corrected regional average rainfall.

The invention discloses an urban inland inundation dynamic early warning method fusing multi-source data and a high-performance numerical model, and belongs to the technical field of urban flood control and disaster reduction and disaster early warning, and the method comprises the following steps: 1, obtaining and preprocessing multi-source data; 2, constructing a high-performance urban flood coupling model; step 3, carrying out assimilation correction on the ensemble Kalman filter; 4, performing rolling prediction and dynamic updating; and step 5, risk identification and early warning release. According to the method, the advantages of multi-source data can be fully fused, and the problems of inaccurate rain condition and unclear waterlogging condition in rapid early warning of urban waterlogging are solved; through multi-source data fusion and high-performance numerical model real-time assimilation correction, minute-level rolling simulation and dynamic correction are achieved, the regional refined ponding risk can be forecasted in advance, the problems that traditional early warning is low in precision, updating lags behind, and the model lacks the self-updating capacity are solved, and the practical value is remarkable.

The invention discloses a method for predicting reservoir inflow and under-dam water resource change trend in a future change environment. The method comprises the steps of obtaining future monthly rainfall data, future monthly temperature data, future monthly evapotranspiration data and future land utilization data based on a GCM climate mode; the future runoff process of the reservoir control section is obtained through simulation in combination with an RCCC-WBM model; based on a data mining technology and an LSTM deep learning coupling model, mining relation characteristics of future reservoir runoff and the under-dam water resource quantity, and obtaining the under-dam water resource quantity by inputting the reservoir runoff; and a nonlinear trend analysis method and the like are adopted to analyze change trend characteristics of water resources under the dam in the future. The method can fully solve the problem of insufficient consideration of climate change and land utilization in the existing method, and can significantly improve the prediction precision of the water resource under the dam in the future.

The invention discloses a water and rain condition emergency monitoring method and system based on a Beidou inspection robot dog, and relates to the technical field of water and rain condition monitoring, and the method comprises the steps: obtaining environment point cloud data and the positioning data of the inspection robot dog; generating a three-dimensional terrain model according to the environment point cloud data; constructing a three-dimensional topology constraint model based on the three-dimensional terrain model, and performing anomaly correction in combination with the three-dimensional topology constraint model and the positioning data to obtain a corrected three-dimensional terrain model; establishing a motion feasibility matrix according to the corrected three-dimensional terrain model, and performing path planning on the inspection robot dog based on the motion feasibility matrix; acquiring historical water and rain monitoring data, and acquiring real-time water and rain monitoring data of a target area in real time by using an inspection robot dog; a dynamic risk evaluation model is constructed by combining real-time water and rain monitoring data and historical water and rain monitoring data, and a regional dynamic risk early warning region is generated based on the dynamic risk evaluation model and a three-dimensional terrain model. According to the invention, the response efficiency of water and rain condition emergency monitoring can be effectively improved.

The invention discloses a rainfall prediction method based on multi-scale physical Transform, and the method comprises the steps: firstly collecting continuous historical radar observation data from a meteorological radar system, and carrying out the preprocessing of the data; secondly, according to the processed observation data, on the basis of a Transform network, capturing global semantics to obtain aggregation features; and then motion field and intensity residual field prediction is carried out based on the aggregation features to obtain a deterministic prediction sequence. And finally, constructing a generative network based on a multi-scale adapter, obtaining a rainfall probabilistic prediction map according to a deterministic prediction sequence, constructing a multi-scale space-time physical consistency loss function, and carrying out end-to-end training. According to the method, the problems of error accumulation, physical inconsistency, low calculation efficiency and the like in a traditional forecasting method are effectively solved, so that a brand new and accurate technical path is provided for real-time forecasting of the extreme rainfall event.

The invention discloses a rainstorm ground waterlogging prediction method, device, equipment and product, and relates to the technical field of urban waterlogging prediction.The rainstorm ground waterlogging prediction method comprises the steps that ephemeris and ground observation station data are obtained, the atmospheric moisture delay amount and the high-altitude water vapor conversion factor are calculated respectively, the historical water vapor content is calculated, and the future atmospheric precipitable water amount is predicted; rainfall time distribution data are obtained through rainstorm intensity decomposition; calculating the rainwater overflow amount of the catchment area in combination with the water discharge amount time distribution data of the urban drainage pipe network; and iterative calculation is carried out by using a digital elevation model grid to obtain the waterlogging depth, and waterlogging early warning information is sent out. By constraining the total rainfall amount, the rainfall amount can be effectively constrained by predicting the precipitable water amount based on the satellite ephemeris; the underground pipe network drainage is restrained, and a water flow condition in a pipe channel is simulated by adopting an SWMM model; and finally, the waterlogging accumulated water volume is calculated, constraint verification is carried out on each link of urban waterlogging, and relatively high precision and relatively long timeliness are ensured.

The invention discloses a rainfall prediction method based on LightGBM and a variable attention mechanism, and belongs to the technical field of weather prediction. The objective of the invention is to solve the problem of poor prediction effect of an existing model when noise or missing exists in data. According to the method, the LightGBM and the variable attention mechanism are combined, the variable attention mechanism is used for accurately capturing cross-variable dynamic association and multi-scale time dependency in meteorological data, the model can flexibly adapt to the relative dependency between different time steps through relative position coding, and the modeling capacity for long-time-sequence data is improved; by adopting a split weighting mechanism based on LightGBM, different types of exogenous variable data can be weighted, large-scale data can be effectively processed, and feature selection can be carried out, so that the model is helped to pay more attention to variables having great influence on meteorological prediction in the training process, and the robustness of the model to noise data and missing data is enhanced. The method can be applied to rainfall prediction.