CN-121682452-B - Correction method and device for multi-classification precipitation phase states

CN121682452BCN 121682452 BCN121682452 BCN 121682452BCN-121682452-B

Abstract

The invention discloses a correction method and a correction device for multi-classification precipitation phases, and relates to the technical field of weather forecast, wherein the correction method comprises the steps of determining the probability of primarily corrected multi-phase precipitation by adopting a stepwise conditional two-classification precipitation phase probability correction algorithm for a full amount of space-time samples; and carrying out spatial downscaling treatment on the probability of the primarily corrected multiphase precipitation by adopting a precipitation phase probability variable rate spatial downscaling algorithm to obtain a high-resolution precipitation phase probability forecast correction product. The correction effect is not good due to unbalance among event sample amounts can be eliminated to the maximum extent. When facing to the reduction of the kilometer resolution, the calculation efficiency is improved, and meanwhile, the spatial continuity of the precipitation phase probability can be ensured.

Inventors

DONG QUAN
DAI KAN
ZHAO HUIXIA

Assignees

国家气象中心(中央气象台、中国气象局气象导航中心)

Dates

Publication Date: 20260512
Application Date: 20260210

Claims (9)

1. A method of correcting for multi-classification precipitation phases, comprising: determining the probability of the primarily corrected multiphase precipitation by adopting a gradual condition two-classification precipitation phase probability correction algorithm aiming at a full amount of space-time samples, wherein all events are divided into two types of non-precipitation and precipitation-containing precipitation, and correction is carried out by adopting a two-classification correction method to obtain corrected precipitation-containing probability P1; dividing the non-rainfall event into snowfall and non-solid rainfall, correcting by adopting a two-class correction method to obtain a snowfall probability P_snort under the condition of the rainfall, and determining a corrected snowfall probability P_rain based on the snowfall probability P_snort, the rainfall probability P_1 and the rainfall probability P_rain, dividing the non-solid rainfall event into a rain-clip snow and a freezing rain, correcting by adopting a two-class correction method to obtain a snowfall probability P_ sleet under the condition of the rainfall and the non-rainfall and a freezing rain probability P_ freezing, and determining a corrected snowfall probability P_snort based on the snowfall probability P_snort, the rainfall probability P_snort and the rainfall probability P_rain, and the freezing probability P_rain based on the correction probability P_37 and the freezing probability P_rain; And carrying out spatial downscaling treatment on the probability of the primarily corrected multiphase precipitation by adopting a precipitation phase probability variable rate spatial downscaling algorithm to obtain a high-resolution precipitation phase probability forecast correction product.
2. The correction method for multi-classification precipitation phases according to claim 1, wherein the spatial downscaling of the probability of the primarily corrected multi-phase precipitation using a precipitation phase probability variability spatial downscaling algorithm comprises: Calculating a difference between the terrain height of the coarse resolution product and the terrain height of the high resolution product, and calculating an air temperature variability dT of the high resolution relative to the coarse resolution based on the wet adiabatic reduction rate and combining the difference; Fitting the occurrence probability of rain and snow at different air temperatures by adopting a logistic regression function to obtain a occurrence probability function of rain and a occurrence probability function of snow, and respectively deriving the occurrence probability function of rain and the occurrence probability function of snow to obtain a variation dPr of rainfall probability and a variation dPs of snowfall probability at different air temperatures; And respectively calculating a rainfall probability correction value and a snowfall probability correction value according to the air temperature change rate dT, the rainfall probability change rate dPr and the snowfall probability change rate dPs and a high-resolution air temperature forecast product, further obtaining a rain and snow probability correction value, respectively adding the rainfall probability correction value, the snowfall probability correction value and the rain and snow probability correction value to the probabilities of the primarily corrected multiphase precipitation to obtain the down-scale rain, snow, rain and snow probability and freezing rain probability, and forming the high-resolution rainfall phase state probability forecast correction product.
3. The correction method for multi-classification precipitation phases according to claim 1, wherein determining a corrected preliminary probability of precipitation based on the probability of precipitation and the probability of precipitation comprises: Based on a conditional probability formula, the rainfall preliminary probability after final correction is calculated as P_rain×P1.
4. The correction method for multi-classification precipitation phases according to claim 1, wherein determining the corrected snowfall probability based on the snowfall probability p_snorw, the precipitation probability P1, and the precipitation probability p_rain comprises: the snowfall probability is determined to be P_snorow× (1-P_rain) ×P1 according to the conditional probability formula.
5. The correction method for multi-classification precipitation phases according to claim 1, wherein determining the corrected rain and snow probability based on the rain and snow probability p_ sleet, the snowfall probability p_snor, the rainfall probability p_rain, the precipitation probability P1, and determining the corrected freezing and rain probability based on the freezing and rain probability p_ freezing, the snowfall probability p_snor, the rainfall probability p_rain, the precipitation probability P1 comprises: The probability of the finally corrected rain and snow is calculated according to a conditional probability formula as follows: P_sleet×(1-P_snow)×(1-P_rain)×P1, the freezing rain probability after final correction is calculated according to a conditional probability formula is as follows: P_freezing×(1-P_snow)×(1-P_rain)×P1。
6. a correction device for multi-classification precipitation phases, comprising: The probability determination unit of the multiphase precipitation is used for determining the probability of the multiphase precipitation after preliminary correction by adopting a step-by-step condition two-class precipitation phase probability correction algorithm aiming at a full-quantity space-time sample, and comprises the steps of dividing all events into two classes of precipitation-free precipitation and precipitation-free precipitation, correcting by adopting a two-class correction method, and obtaining corrected precipitation-free probability P1; dividing the non-rainfall event into snowfall and non-solid rainfall, correcting by adopting a two-class correction method to obtain a snowfall probability P_snort under the condition of the rainfall, and determining a corrected snowfall probability P_rain based on the snowfall probability P_snort, the rainfall probability P_1 and the rainfall probability P_rain, dividing the non-solid rainfall event into a rain-clip snow and a freezing rain, correcting by adopting a two-class correction method to obtain a snowfall probability P_ sleet under the condition of the rainfall and the non-rainfall and a freezing rain probability P_ freezing, and determining a corrected snowfall probability P_snort based on the snowfall probability P_snort, the rainfall probability P_snort and the rainfall probability P_rain, and the freezing probability P_rain based on the correction probability P_37 and the freezing probability P_rain; The high-resolution precipitation phase probability prediction correction unit is used for performing spatial scale reduction treatment on the probability of the primarily corrected multiphase precipitation by adopting a precipitation phase probability variability spatial scale reduction algorithm to obtain a high-resolution precipitation phase probability prediction correction product.
7. A computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of any one of claims 1-5.
8. An electronic device comprising at least one processor and a memory communicatively coupled to the at least one processor, wherein the memory stores a computer program executable by the at least one processor to cause the at least one processor to perform the method of any of claims 1-5.
9. A computer program product, characterized in that the computer program, when being executed by a processor, implements the method of any of claims 1-5.

Description

Correction method and device for multi-classification precipitation phase states Technical Field The invention relates to the technical field of information processing, in particular to a correction method and device for multi-classification precipitation phases. Background Precipitation has a number of different phases including rain, snow, rain-snow, freezing rain, ice particles, aragonite, etc., and thus precipitation phases are typical discrete multi-classification variables. Aiming at probability forecast of different precipitation phases, a basic probability forecast product is provided by a combined numerical weather forecast mode. But the reliability of the products is generally lower, and the spatial resolution is also thicker (more than 9 km). Therefore, the statistical post-processing correction and the spatial downscaling are needed to be carried out on the rainfall phase probability prediction result output by the weather prediction mode, on one hand, the prediction deviation is eliminated, the prediction skill is improved, on the other hand, the spatial resolution of the prediction is improved, and the requirement of refined weather prediction is met. The post-processing method for discrete variables in the probability forecast of meteorological elements such as air temperature, precipitation phase and the like is mainly aimed at two classification events, such as the existence and non-existence of precipitation, the existence and non-existence of storm and the like. For multi-classification events, a plurality of two-classification processing methods are also adopted, for example, precipitation phases are sequentially classified into rain, non-rain, snow, non-snow, freezing rain, non-freezing rain and the like, and probability correction of two-classification is respectively carried out on different precipitation phases. This approach has two disadvantages. Firstly, the sum of probabilities of different precipitation phases (including no precipitation) is 1, and the constraint condition must be met before and after correction, but the existing correction method cannot guarantee the constraint, and the constraint can only be realized by carrying out normalization processing on the probabilities of the different precipitation phases after correction, so that the objective correction effect of probability forecast is affected. Secondly, the sample sizes of different precipitation phases are extremely unbalanced, the ratio of non-rain to rain sample sizes is in the order of 10:1, the ratio of non-rain, snow and snow sample sizes is in the order of 100:1, and the significantly unbalanced sample sizes can greatly influence the effect of objective correction, so that the final correction probability is biased to the event with more sample sizes. The two defects cause that the correction effect of the correction method of the multiple two-class is limited on the discretized multi-class event of precipitation phase state. In the space downscaling method of discrete variables such as precipitation phase, the characteristic development is mainly based on the linear change of air temperature and a rain and snow boundary, and the air temperature change of about 1 ℃ can cause the transformation of rain and snow in a certain area (usually tens of kilometers square) under the assumption that the air temperature direct reduction rate is 0.6 ℃ per 100 meters. Based on the above assumption, the precipitation phase is linearized in terms of spatial downscaling. However, for the probability of precipitation phase, no reliable statistical downscaling method exists at present. Disclosure of Invention The invention mainly aims to provide a correction method and device for multi-classification precipitation phases, which are used for solving the defects in the related art. In order to achieve the above object, according to a first aspect of the present invention, there is provided a correction method for multi-classification precipitation phases, including determining probabilities of primarily corrected multi-phase precipitation by using a step-wise conditional two-classification precipitation phase probability correction algorithm for a full amount of space-time samples, and performing spatial downscaling on the probabilities of primarily corrected multi-phase precipitation by using a precipitation phase probability variability spatial downscaling algorithm to obtain a high-resolution precipitation phase probability forecast correction product. The method comprises the steps of dividing all events into two types of non-precipitation and non-precipitation, correcting by adopting a two-class correction method to obtain corrected precipitation probability P1, dividing the precipitation events into rainfall and non-precipitation, correcting by adopting a two-class correction method to obtain precipitation probability P_rain under precipitation conditions, determining the rainfall preliminary probability after correctio