Single-point rainfall forecast optimization method and system for multi-period independent calibration

Abstract

The invention provides a multi-period independent calibration single-point rainfall forecast optimization method and system. The method comprises the following steps: acquiring forecast data and observation data, and preprocessing the forecast data and the observation data to obtain a forecast observation matching sample; dividing the forecast observation matching sample into a plurality of independent time periods; setting a cold start period; setting an attenuation coefficient, and dynamically updating the forecast CDF and the observation CDF; presetting a multi-gradient rainfall threshold value and calculating to obtain a corresponding optimization threshold value; and performing interpolation calculation on the forecast rainfall intensity to obtain optimized rainfall intensity. The method has the beneficial effects that independent optimization is carried out according to time periods, different aging error characteristics are accurately adapted, and the forecasting precision of each time period is comprehensively improved; in the cold start stage, the CDF is initialized through sample accumulation without depending on prior parameters; data space-time difference is eliminated through multi-step preprocessing, space matching of forecast and observation data is ensured, and sample dislocation is avoided; and the attenuation coefficient is associated with an attenuation window, new and old data weights are dynamically allocated, and the stability of historical data and the timeliness of recent data are considered.

Description

Technical Field

[0001] This invention belongs to the field of rainfall prediction technology, and in particular relates to a method and system for optimizing single-point precipitation forecasts with multi-period independent calibration. Background Technology

[0002] Numerical weather forecasting is the core of modern meteorological operations. The global numerical weather prediction products of the European Centre for Medium-Range Weather Forecasts (ECMWF) are widely used for forecasting meteorological elements such as precipitation due to their wide spatial coverage and long forecast lead time. The China Regional Precipitation Analysis Product (CMPA) of the China Meteorological Administration, with its high spatial resolution and hourly updates, has become a key observational data source for precipitation forecast bias correction and effectiveness verification.

[0003] Single-point precipitation forecasting is a core requirement of meteorological services, directly impacting decision-making in agriculture, flood control, transportation, and other fields. However, ECMWF precipitation forecasts suffer from systematic biases such as missed light rain and false heavy rain, stemming from model parameterization schemes, initial field errors, and underlying surface characteristics. Furthermore, the error characteristics of different forecast lead times vary significantly. Existing precipitation forecast optimization techniques suffer from the following shortcomings: insufficient generalization across multiple time periods, lack of independent mechanisms for different lead times; low reliability during the cold start phase due to a lack of historical samples; rigid historical data update mechanisms that fail to respond to seasonal weather changes; and a single, simplistic precipitation intensity threshold design. Therefore, a precipitation forecast optimization method that addresses these issues is urgently needed. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for optimizing single-point precipitation forecasts with independent calibration over multiple time periods, which is particularly suitable for optimizing single-point precipitation forecasts with independent calibration over multiple time periods.

[0005] The technical solution adopted in this invention is as follows: Firstly, a method for optimizing single-point precipitation forecasting with independent calibration over multiple time periods is provided, comprising the following steps: Acquire forecast and observation data and preprocess them to obtain a forecast-observation matching sample; The predicted observation matching samples are divided into multiple independent time periods; A cold start period is set, the forecast observation matching samples of each independent time period are accumulated, and the daily forecast CDF and daily observation CDF are calculated respectively. When the cold start period meets the standard, the average value is calculated to obtain the forecast CDF and observation CDF of each independent time period. Set an attenuation coefficient and dynamically update the predicted CDF and the observed CDF; A multi-gradient precipitation threshold is preset, and the corresponding optimized threshold is determined by probability matching between the forecast CDF and the observed CDF. Based on the multi-gradient precipitation threshold and the optimized threshold, the predicted rainfall intensity is interpolated to obtain the optimized rainfall intensity.

[0006] Further, the preprocessing includes the following steps: A rectangular area with the observation distance as the radius is defined, centered on the target station, and the forecast data and the observation data are then cropped. The forecast data and the observation data are spatiotemporally aligned to eliminate grid misalignment; The forecast data and the observation data are spatially resampled using a linear interpolation algorithm to ensure that the number of grid cells is consistent. The grid points around the target site are located as the calculation input for bilinear interpolation. The predicted and observed values ​​of the target site location are calculated to form the predicted-observation matching sample.

[0007] Furthermore, setting an attenuation coefficient and dynamically updating the predicted CDF and the observed CDF includes the following steps: A preset attenuation window is used to calculate the attenuation coefficient. Based on the matching sample of the forecast observations for the day, the daily forecast CDF and the daily observation CDF for the day are calculated; Historical and recent weights are set, and the forecast CDF and observation CDF are updated separately through dynamic weighting.

[0008] Furthermore, a correction factor is introduced into the calculation of the optimized rainfall intensity, and a range interval for the correction factor is set.

[0009] Secondly, a single-point precipitation forecast optimization system with independent multi-time period calibration is provided, including: The data processing module is used to acquire forecast data and observation data and perform preprocessing to obtain forecast-observation matching samples; The multi-period segmentation module includes multiple dedicated optimization units, used to segment the forecast observation matching samples according to the forecast lead time; The cold start processing module is used to set the cold start period and accumulate the forecast observation matching samples within the period, and calculate the daily forecast CDF and the daily observation CDF; The dynamic update module is used to set the attenuation coefficient and update the forecast CDF and observation CDF of each dedicated optimization unit. The precipitation optimization module is used to set gradient precipitation thresholds, determine the corresponding optimization thresholds by probability matching between the forecast CDF and the observed CDF, calculate the correction factor by logarithmic linear interpolation, and output the optimized rainfall intensity.

[0010] Furthermore, it also includes an output monitoring module for optimizing accuracy verification, anomaly monitoring, regular maintenance, and efficiency assurance.

[0011] Furthermore, the dedicated parameters of the dedicated optimization unit include time period name, current forecast CDF, current observation CDF, historical sample set, cold start status, and cold start CDF data.

[0012] Furthermore, the dynamic update module sets a sample upper limit for each dedicated optimization unit to limit the total number of historical samples. When the sample upper limit is exceeded, the earliest predicted observation matching sample is automatically deleted.

[0013] The advantages and positive effects of this invention are as follows: By adopting the above-mentioned technical solution, independent optimization is performed according to time periods, accurately adapting to different time-sensitivity error characteristics and comprehensively improving the forecast accuracy of each time period; during the cold start phase, CDF is initialized through sample accumulation, without relying on prior parameters, ensuring the stability of optimization in the early stage of operation; multi-step preprocessing eliminates the spatiotemporal differences of data, ensuring spatial matching between forecast and observation data and avoiding sample misalignment; the attenuation coefficient is associated with the attenuation window, dynamically allocating the weights of new and old data, taking into account both the stability of historical data and the timeliness of recent data. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating a precipitation forecast optimization method according to an embodiment of the present invention. Detailed Implementation

[0015] The present disclosure will now be described more fully with reference to the accompanying drawings, which illustrate exemplary embodiments of the present disclosure. The technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present disclosure, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present disclosure without creative effort are within the scope of protection of the present disclosure.

[0016] like Figure 1 As shown, this invention provides a method for optimizing single-point precipitation forecasts with multi-time period independent calibration, comprising the following steps: S100. Obtain forecast data and observation data and perform preprocessing to obtain forecast-observation matching samples; Specifically, forecast data must have clear forecast lead time and spatial resolution, and be able to output variables related to rainfall intensity; observation data must have a time dimension and spatial coverage that match the forecast data, and a resolution no lower than that of the forecast data; the data formats of both forecast and observation data must support latitude and longitude information parsing; preprocessing is performed to address the spatiotemporal mismatch between forecast and observation data, extracting forecast and observation values ​​from the target stations to obtain a forecast-observation matching sample.

[0017] Preferably, the forecast data is ECMWF (European Centre for Medium-Range Weather Forecasts) forecast data, which is a global numerical weather prediction product released by the European Centre for Medium-Range Weather Forecasts; the observation data is CMPA (China Meteorological Administration) observation data, which is a precipitation analysis product for the Chinese region released by the China Meteorological Administration. CMPA is a high-precision precipitation observation and analysis data source for the Chinese region, primarily used for verifying the effectiveness of precipitation forecasts and correcting biases. Both ECMWF forecast data and CMPA observation data are in GRIB2 format.

[0018] S200, Divide the forecast observation matching sample into multiple independent time periods; Specifically, the forecast observation matching samples are divided according to the forecast lead time. The divided units are completely isolated, and only the forecast observation matching samples of the current period are used for calculation.

[0019] Preferably, the forecast observation matching samples are divided into short-term units, medium-term units, and long-term units. The forecast lead time for short-term units is 0-24 hours, and the forecast frequency is once every 3 hours; the forecast lead time for medium-term units is 24-72 hours, and the forecast frequency is once every 3 hours; the forecast lead time for long-term units is 72-168 hours, and the forecast frequency is once every 6 hours.

[0020] S300. Set a cold start period, accumulate forecast observation matching samples for each independent time period and calculate the daily forecast CDF and daily observation CDF respectively. When the cold start period meets the standard, calculate the mean value to obtain the forecast CDF and observation CDF for each independent time period. Specifically, CDF stands for Cumulative Distribution Function, which is used to fully describe the probability distribution of a random variable. A cold start period is set, during which only samples are accumulated and no formal optimization is performed, i.e., the forecast rainfall intensity is not optimized. Each day, based on the forecast observation matching samples for this period, a set of "daily forecast CDF" and "daily observation CDF" are calculated respectively (i.e., the daily forecast CDF and daily observation CDF are calculated for different independent time periods). When the number of cold start days reaches the target, the mean of all daily forecast CDFs and the mean of daily observation CDFs during the cold start period are calculated respectively, and the two are assigned to the current forecast CDF and the current observation CDF respectively, thus formally entering the normal optimization phase.

[0021] The calculation steps for the daily forecast CDF are as follows: Extract all forecast values ​​from the matched samples of the daily forecast observations and count the total number of valid forecast samples; if there are multiple time-leading forecast samples for that period of the day, count the number of samples whose forecast values ​​are less than or equal to the corresponding precipitation threshold for each precipitation threshold, and then take the average to smooth the result; count the number of valid forecast values ​​less than or equal to the precipitation threshold for each precipitation threshold; calculate the forecast cumulative probability corresponding to each precipitation threshold; pair all precipitation thresholds with their corresponding forecast cumulative probabilities to form the daily forecast CDF. The calculation steps for the daily observation CDF are completely consistent with those for the daily forecast CDF.

[0022] Preferably, the cold start period is set to 50 days.

[0023] S400: Set the attenuation coefficient and dynamically update the forecast CDF and observed CDF; S500. Based on the preset multi-gradient precipitation threshold, the corresponding optimized threshold is determined by probability matching between the forecast CDF and the observed CDF. Specifically, multiple gradient precipitation thresholds are preset (preset at the start of the scheme) to cover different rainfall ranges. All forecast values ​​in the current period's forecast observation matching samples are extracted, and the total number of samples N is counted. For each precipitation threshold, the number n of samples whose forecast values ​​are less than or equal to the gradient threshold is counted. The cumulative probability n / N corresponding to the precipitation threshold is calculated, and the precipitation threshold is paired with its corresponding forecast cumulative probability to form a complete historical forecast CDF. The same method is used to form a complete historical observation CDF. The forecast cumulative probability corresponding to the precipitation threshold is extracted from the historical forecast CDF. If its observation cumulative probability is not sorted, it is first sorted from smallest to largest according to the preset precipitation threshold to ensure that the probability monotonically increases with the threshold. In the sorted historical observation CDF, the observation cumulative probability that is equal to or closest to the forecast cumulative probability is found. The observation threshold corresponding to this observation cumulative probability is extracted, and this value is the optimized threshold corresponding to the precipitation threshold (which also corresponds to the number of precipitation thresholds).

[0024] Preferably, 14 precipitation thresholds are fixed according to the rainfall intensity, from low to high. .

[0025] S600: Based on the multi-gradient precipitation threshold and the optimization threshold, the forecast precipitation intensity is interpolated to obtain the optimized precipitation intensity.

[0026] Specifically, a natural logarithmic transformation is performed on all multi-gradient precipitation thresholds, optimization thresholds, and forecasted rainfall intensity (original forecast value); the interpolation interval is located; the logarithm of the target optimization threshold corresponding to the forecasted rainfall intensity is calculated through linear interpolation; the target optimization threshold is obtained through inverse logarithmic transformation, and the target optimization threshold is output as the optimized rainfall intensity (i.e., the forecasted rainfall intensity is optimized).

[0027] Using the above methods, independent optimization is performed for each time period to accurately adapt to different time-leadership error characteristics and comprehensively improve the forecast accuracy for each time period. During the cold start phase, CDF is initialized through sample accumulation without relying on prior parameters, ensuring the stability of initial optimization. The closed-loop design of the entire process takes into account spatiotemporal adaptation, dynamic updates, and full-intensity coverage, making the forecast more in line with actual weather conditions.

[0028] To address the issues of poor spatiotemporal data adaptability in existing technologies and the failure to resolve latitude and longitude alignment problems, which lead to sample space misalignment and data distortion, this embodiment provides an implementation method.

[0029] In one embodiment, the preprocessing includes the following steps: A rectangular area with the observation distance as the radius is drawn around the target station, and the forecast data and observation data are cropped accordingly; Spatiotemporal alignment of forecast and observation data is performed to eliminate grid misalignment; Spatial resampling of forecast and observation data is performed using a linear interpolation algorithm to ensure consistent grid size. The grid points around the target site are used as the calculation input for bilinear interpolation to calculate the predicted and observed values ​​of the target site location, forming a predicted-observation matching sample.

[0030] Using the above method, multi-step preprocessing eliminates spatiotemporal differences in data, ensures spatial matching between forecast and observation data, and avoids sample misalignment; it accurately extracts single-point samples from target stations, providing a reliable data foundation for subsequent CDF calculation and precipitation optimization, and improving the accuracy of optimization results.

[0031] To address the problem that existing historical data update mechanisms are rigid and fixed sliding windows do not consider timeliness weights, thus failing to respond promptly to weather and seasonal changes, this embodiment provides an implementation method.

[0032] In one embodiment, setting an attenuation coefficient and dynamically updating the forecast CDF and observed CDF includes the following steps: A preset attenuation window is used to calculate the attenuation coefficient. Preferably, the decay window is 50 days and the decay coefficient is 1 / decay window, that is, the decay coefficient is 0.02.

[0033] Based on the matching samples of the forecast observations for the day, calculate the daily forecast CDF and the daily observation CDF for the day; Historical and recent weights are set, and the forecast CDF and observation CDF are updated separately through dynamic weighting.

[0034] Specifically, through equations Updated forecast CDF, in which For the updated forecast CDF, For the CDF forecast before the update, The CDF is the forecast for the day, and W is the decay coefficient (i.e., the recent weight). Historical weights; through equations Update the observed CDF, where To update the observed CDF, To update the CDF observations, For CDF observations on that day.

[0035] Using the above method, the attenuation coefficient is associated with the attenuation window, and the weights of new and old data are dynamically allocated, taking into account both the stability of historical data and the timeliness of recent data; the CDF is updated in real time, enabling the model to quickly adapt to changes in weather systems and improve the accuracy of forecasts in real-time weather characteristics; the weighted update method is scientific and reasonable, avoiding optimization lag caused by rigid updates and ensuring the continuity of forecast accuracy.

[0036] To address the issue of over-correction, under-correction, or extreme outliers that may occur during precipitation optimization, leading to distorted forecast results, this embodiment provides an implementation method.

[0037] In one embodiment, a correction factor is introduced into the calculation of optimized rainfall intensity, and a range interval for the correction factor is set.

[0038] Specifically, the correction factor is the target optimization threshold / forecast rainfall intensity. The calculated correction factor must be within the range; otherwise, it needs to be pruned. The final output optimized rainfall intensity is the correction factor multiplied by the forecast rainfall intensity.

[0039] Preferably, the range of the correction factor is as follows: .

[0040] Using the above method, the correction factor can quantify the correction magnitude, making the optimization logic clearer and more traceable, and facilitating result verification and adjustment; the range limit avoids extreme corrections, prevents forecast runaway caused by abnormal data, and ensures the rationality of the optimization results.

[0041] To facilitate the use of the multi-period independent calibration single-point precipitation forecast optimization method provided in this disclosure, this disclosure also provides a multi-period independent calibration single-point precipitation forecast optimization system, including: The data processing module is used to acquire forecast data and observation data and perform preprocessing to obtain forecast-observation matching samples; The multi-period segmentation module includes multiple dedicated optimization units, which are used to segment the forecast observation matching samples according to the forecast lead time; The cold start processing module is used to set the cold start period and accumulate forecast observation matching samples within the period, and calculate the daily forecast CDF and the daily observation CDF; Specifically, during the cold start period, the cold start processing module accumulates forecast-observation matching samples, calculates the daily forecast CDF and daily observation CDF for each dedicated optimization unit, and stores them in the data list of the cold start processing module to avoid single-day weight skew. When the number of cold start days meets the target and the data is not empty, the mean of all daily forecast CDFs and the mean of daily observation CDFs during the cold start period are calculated respectively, and these are assigned to the current forecast CDF and the current observation CDF respectively. At the same time, the cold start status is set to False.

[0042] The dynamic update module is used to set the attenuation coefficient and update the forecast CDF and observation CDF of each dedicated optimization unit; The precipitation optimization module is used to set gradient precipitation thresholds. It determines the corresponding optimization threshold by probability matching between the forecast CDF and the observed CDF, calculates the correction factor by log-linear interpolation, and outputs the optimized rainfall intensity.

[0043] In one embodiment, it further includes an output monitoring module for performing optimization accuracy verification, anomaly monitoring, regular maintenance, and efficiency assurance.

[0044] In one embodiment, the dedicated parameters of the dedicated optimization unit include time period name, current forecast CDF, current observation CDF, historical sample set, cold start status, and cold start CDF data.

[0045] In one embodiment, the dynamic update module sets a sample limit for each dedicated optimization unit to limit the total number of historical samples. When the sample limit is exceeded, the earliest forecast observation matching sample is automatically deleted.

[0046] Based on embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0047] An electronic device includes at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute the multi-time period independently calibrated single-point precipitation forecast optimization method provided in this disclosure.

[0048] Electronic devices are intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0049] A non-transient computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to execute the multi-time period independently calibrated single-point precipitation forecast optimization method provided in this disclosure.

[0050] The various embodiments of this disclosure can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0051] A computer program product includes a computer program / instructions that are executed by a processor to provide a multi-time period independently calibrated single-point precipitation forecast optimization method.

[0052] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0053] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0054] The embodiments of the present invention have been described in detail above, but the content described is only a preferred embodiment of the present invention and should not be considered as limiting the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the patent coverage of the present invention.

Claims

1. A method for optimizing single-point precipitation forecasts with independent calibration over multiple time periods, characterized in that, Includes the following steps: Acquire forecast and observation data and preprocess them to obtain a forecast-observation matching sample; The predicted observation matching samples are divided into multiple independent time periods; A cold start period is set, the forecast observation matching samples of each independent time period are accumulated, and the daily forecast CDF and daily observation CDF are calculated respectively. When the cold start period meets the standard, the average value is calculated to obtain the forecast CDF and observation CDF of each independent time period. Set an attenuation coefficient and dynamically update the predicted CDF and the observed CDF; A multi-gradient precipitation threshold is preset, and the corresponding optimized threshold is determined by probability matching between the forecast CDF and the observed CDF. Based on the multi-gradient precipitation threshold and the optimized threshold, the predicted rainfall intensity is interpolated to obtain the optimized rainfall intensity.

2. The method for optimizing single-point precipitation forecasting with multi-period independent calibration according to claim 1, characterized in that, Preprocessing includes the following steps: A rectangular area with the observation distance as the radius is defined, centered on the target station, and the forecast data and the observation data are then cropped. The forecast data and the observation data are spatiotemporally aligned to eliminate grid misalignment; The forecast data and the observation data are spatially resampled using a linear interpolation algorithm to ensure that the number of grid cells is consistent. The grid points around the target site are located as the calculation input for bilinear interpolation. The predicted and observed values ​​of the target site location are calculated to form the predicted-observation matching sample.

3. The method for optimizing single-point precipitation forecasting with multi-time period independent calibration according to claim 1, characterized in that, Setting an attenuation coefficient and dynamically updating the forecast CDF and the observed CDF includes the following steps: A preset attenuation window is used to calculate the attenuation coefficient. Based on the matching sample of the forecast observations for the day, the daily forecast CDF and the daily observation CDF for the day are calculated; Historical and recent weights are set, and the forecast CDF and observation CDF are updated separately through dynamic weighting.

4. The method for optimizing single-point precipitation forecasting with multi-time period independent calibration according to claim 1, characterized in that: A correction factor is introduced into the calculation of the optimized rainfall intensity, and the range of the correction factor is set.

5. A single-point precipitation forecast optimization system with independent calibration over multiple time periods, characterized in that, include: The data processing module is used to acquire forecast data and observation data and perform preprocessing to obtain forecast-observation matching samples; The multi-period segmentation module includes multiple dedicated optimization units, used to segment the forecast observation matching samples according to the forecast lead time; The cold start processing module is used to set the cold start period and accumulate the forecast observation matching samples within the period, and calculate the daily forecast CDF and the daily observation CDF; The dynamic update module is used to set the attenuation coefficient and update the forecast CDF and observation CDF of each dedicated optimization unit. The precipitation optimization module is used to set gradient precipitation thresholds, determine the corresponding optimization thresholds by probability matching between the forecast CDF and the observed CDF, calculate the correction factor by logarithmic linear interpolation, and output the optimized rainfall intensity.

6. The multi-period independent calibration single-point precipitation forecast optimization system according to claim 5, characterized in that, Also includes: The output monitoring module is used for optimization accuracy verification, anomaly monitoring, and regular maintenance.

7. The multi-time period independently calibrated single-point precipitation forecast optimization system according to claim 5, characterized in that: The dedicated parameters of the dedicated optimization unit include time period name, current forecast CDF, current observation CDF, historical sample set, cold start status, and cold start CDF data.

8. The multi-period independent calibration single-point precipitation forecast optimization system according to claim 5, characterized in that: The dynamic update module sets a sample limit for each dedicated optimization unit to restrict the total number of historical samples. When the sample limit is exceeded, the earliest predicted observation matching sample is automatically deleted.

Images