Improved garin river channel base flow segmentation method and system
By using an improved Kalinin method for baseflow segmentation, data preprocessing and automation technologies are employed to address the problem of reliance on manual experience in the traditional Kalinin method. This method achieves efficient and automated baseflow segmentation, adapts to different watershed characteristics, and supports batch processing of multiple sites and generation of high-precision baseflow data.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ANHUI PROVINCIAL HYDROLOGY BUREAU (ANHUI PROVINCIAL SOIL & WATER CONSERVATION MONITORING STATION)
- Filing Date
- 2026-06-05
- Publication Date
- 2026-07-31
AI Technical Summary
The traditional Kalinin method relies on human experience in river baseflow segmentation, which leads to complex data processing, making it difficult to achieve proceduralization and automation, and thus failing to meet the needs of high-frequency, large-volume runoff analysis.
An improved baseflow segmentation method for the Kalinin River channel, including data preprocessing, delay correlation analysis, composite flood peak identification, recession process screening, and penalty mechanism, combined with an error tolerance mechanism, enables automated baseflow segmentation.
It enables one-click generation of baseflow segmentation results from raw runoff data, improves processing efficiency, eliminates human experience differences, supports batch processing of long sequences and multiple stations, provides high-precision baseflow data, and adapts to different watershed characteristics and data quality.
Smart Images
Figure CN122332879B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of hydrological and water resources analysis technology in water conservancy engineering, and relates to an improved method and system for baseflow division in the Kalinin River channel. Background Technology
[0002] Baseflow refers to the relatively stable portion of river runoff replenished by groundwater, and is a key element in maintaining river ecological health and ensuring the sustainable use of water resources. Baseflow segmentation, as a fundamental task in hydrological analysis, directly impacts the reliability of decisions in key areas such as water resource assessment, ecological flow determination, drought early warning, and water environment capacity calculation. Accurately quantifying river baseflow has become a technical prerequisite for watershed water resource rights confirmation, the establishment of ecological compensation mechanisms, and the guarantee of river and lake ecological flows.
[0003] However, baseflow segmentation involves a complex interaction between surface water and groundwater, influenced by multiple factors such as watershed geological structure, soil properties, precipitation patterns, and human activities. Scientifically separating baseflow components has always been a classic challenge in hydrology. Especially under changing environments, the demand for high-frequency, large-scale runoff analysis is increasingly urgent, highlighting the inefficiencies and subjectivity of traditional manual segmentation methods. Therefore, developing programmed, autonomous, intelligent, and standardized baseflow segmentation technologies is of significant practical importance.
[0004] Current baseflow segmentation methods mainly include the linear cutting method, the oblique cutting method, and the numerical analysis method represented by Kalinin. The linear cutting and oblique cutting methods are simple and convenient, but their physical mechanisms are weak, resulting in low accuracy and insufficient reliability. The segmentation results are significantly affected by subjective point selection. The Ministry of Water Resources' "National Water Resources Comprehensive Planning Technical Outline" and its accompanying "Supplementary Details on Groundwater Resources and Extractable Quantities" list the Kalinin trial-and-error method as a recommended method for baseflow segmentation, emphasizing that for regions with continuous peak-shaped flow processes, the Kalinin method, based on linear reservoir theory and with clearly defined physical meanings of its parameters, has greater technical advantages. The Kalinin method has significant advantages such as calibrable parameters, clear and reproducible physical processes, and transferable results. By tracking the baseflow recession process through mathematical iteration, it can better characterize the baseflow recovery dynamics during flood remission periods.
[0005] While the Kalinin method possesses theoretical advantages, its data analysis and processing are complex, requiring hydrologists to have certain analytical and computational skills and experience in baseflow segmentation. In particular, the processing involves several steps requiring human judgment, which is the main reason for its limited adoption and failure to be automated. These manual processing steps include: accurately identifying flood events from long-series runoff data, extracting the recession hydrographs of these flood events, selecting recession processes with minimal precipitation interference, and superimposing and aligning multiple recession processes to determine key parameters. Traditionally, these steps rely on the experience and judgment of hydrologists. These manual steps are not only extremely labor-intensive but also introduce difficult-to-quantify cognitive biases, leading to a lack of comparability in the segmentation results of the same river at different times and from different stations. This also hinders its automation and severely restricts the practical application and promotion of the Kalinin method. Summary of the Invention
[0006] The technical solution of this invention is used to solve the problems of the traditional Kalinin method in river baseflow segmentation, which involves complex data processing, reliance on human experience for flood identification, recession process extraction and screening, resulting in strong subjectivity in manual operation, difficulty in achieving programmed and automated processing, and inability to meet the needs of high-frequency and large-volume runoff analysis.
[0007] The present invention solves the above-mentioned technical problems through the following technical solutions:
[0008] This invention provides an improved method for baseflow separation in the Kalinin River channel, comprising the following steps: Acquire long-sequence daily river runoff data and concurrent daily precipitation data, and perform standardized preprocessing; Based on the preprocessed data, the average runoff time and maximum runoff time were determined using a delay correlation analysis between precipitation and runoff. The daily runoff sequence is smoothed to eliminate data oscillations, local extreme points are extracted, composite flood peaks are identified and merged to obtain the flood process, and the peak and trough points of the flood process are reviewed and corrected based on the measured runoff data. Typical drainage processes are extracted from flood events based on preset drainage process screening conditions, and drainage sections without precipitation are screened in conjunction with precipitation data to obtain a set of clean drainage sections. All clean water drainage curves are plotted on the same coordinate system with their tails overlapping. The lower envelope of the drainage curve cluster is fitted with an exponential drainage empirical formula. A penalty mechanism is introduced to allow a very small number of measured flow values to be lower than the lower envelope. The fitted lower envelope is the standard drainage curve, and the exponential parameter is the drainage coefficient. The measured runoff was divided into sections using the standard drainage curve. A flow measurement error tolerance mechanism was introduced, and the allowable proportion of the base flow exceeding the measured runoff was set. The proportion coefficient was determined by trial calculation. After the decline coefficient and proportionality coefficient are determined, the measured runoff is divided into baseflows according to the Kalinin method baseflow division formula to obtain the baseflow process of the measured runoff sequence at the same time and the proportion of the overall baseflow to the total runoff.
[0009] Furthermore, the method for determining the average runoff time and maximum runoff time using the delay correlation analysis of precipitation and runoff is as follows: The delayed sequence is constructed as follows:
[0010] in, The watershed average precipitation series calculated based on the correlation coefficient. The river cross-sectional runoff sequence is calculated using correlation coefficients. The river cross-sectional runoff is the length of the trial period, dT, which is the delay time. The correlation coefficients under different confluence time conditions were calculated separately. The formulas for calculating the correlation coefficients under different confluence time conditions are as follows:
[0011] in, The mean of the watershed average precipitation series is calculated using the correlation coefficient. The mean of the river cross-sectional runoff sequence calculated for the correlation coefficient; The correlation coefficients under different confluence time conditions were calculated, and the confluence time corresponding to the maximum correlation coefficient was taken as the average confluence time, denoted as . ; 2× +1 is set as the maximum confluence time, denoted as... .
[0012] Furthermore, the identification and merging of the composite flood peak must simultaneously meet the following conditions: The time difference between the next peak and the previous trough is less than the maximum confluence time; The runoff corresponding to the next trough is less than the runoff corresponding to the previous trough. The difference between the peak runoff and the trough runoff is less than 20% of the amplitude of the current recession process; If the conditions are met, the peak-valley process is merged, the current valley point and the next peak point are deleted from the extreme point sequence, the two peak-valley processes are merged into one, the sequence is updated and iteratively processed until there is no composite flood peak that meets the conditions.
[0013] Furthermore, the preset screening conditions for the dewatering process include: The duration of the receding process shall not be less than the maximum confluence time; The peak runoff is greater than the annual maximum monthly average runoff that was the smallest in previous years. The trough runoff was less than the historical minimum monthly average runoff. For the drainage process that meets the above conditions, threshold filtering is used to truncate the tail end of the drainage and remove the head end which contains significant surface runoff characteristics.
[0014] Furthermore, the method for screening the receding water section without precipitation is as follows: for each receding water tail period, select the corresponding precipitation period, and the corresponding precipitation period is the average confluence time of the entire receding water period shifted forward; calculate the average precipitation within the precipitation period, and if it is greater than a preset multiple of the annual average precipitation, then the receding water section is removed.
[0015] Furthermore, the penalty mechanism specifically involves calculating a penalty value for measured flow points that are lower than the standard drainage curve. and cumulative penalty value ,in, Let F be the penalty value at the measured point i, and let F be the cumulative penalty value. The number of typical receding water processes. Let i be the measured runoff value at the measured point. The calculated value of the lower envelope corresponding to the measured point i. This is the penalty coefficient; it is adjusted so that the cumulative penalty value does not exceed 1, allowing a small number of measured points to lie below the standard recession curve. The flow measurement error tolerance mechanism is as follows: if the measured runoff error is 5% to 10%, then the allowable error range shall not exceed half of the measured error.
[0016] Furthermore, the empirical formula for exponential evaporation is as follows: The Kalinin method's basic current segmentation formula is as follows: ;in, The flow rate is t days after the start of the receding water level. The flow rate at the start of the receding water level. The decay coefficient is denoted by t, where t is time. To calculate the aquifer reserves at the end of the time period, This is the proportionality coefficient. To measure the total runoff, To generate traffic, This is the calculation period.
[0017] The present invention also provides a river baseflow segmentation system based on the above-mentioned improved Kalinin River baseflow segmentation method, comprising: a data preprocessing module, a confluence time analysis module, a flood process identification module, and a baseflow segmentation module; The data preprocessing module is used to acquire and standardize daily runoff and precipitation data; The runoff time analysis module is used to determine the average runoff time and the maximum runoff time through precipitation-runoff delay correlation analysis; The flood process identification module is used to smooth the runoff sequence, extract extreme points and perform composite flood peak merging and peak-valley verification, and extract complete clean receding sections without effective precipitation according to set conditions. The baseflow segmentation module is used to determine the fading coefficient by combining the clean water effluent section with the penalty mechanism, calculate the proportional coefficient by combining the error tolerance mechanism, and output the baseflow segmentation result based on the water balance calculation.
[0018] The present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a program that supports the processor in executing the above-described improved Kalinin River baseflow separation method, and the processor is configured to execute the program stored in the memory.
[0019] The present invention also provides a storage medium storing a computer program, which, when executed by a processor, performs the steps of the improved Kalinin River baseflow separation method described above.
[0020] The beneficial effects of this invention are as follows: The technical solution of this invention enables one-click generation of baseflow segmentation results from raw runoff data, completely eliminating manual intervention and greatly improving processing efficiency. Unified data processing and parameter identification standards eliminate differences in human experience, ensuring strict spatiotemporal comparability of segmentation results. It supports batch processing of long sequences and multiple stations, and can be embedded into hydrological forecasting and simulation systems to meet operational needs. The introduction of penalty coefficients and error tolerance mechanisms adapts to different watershed characteristics and data quality, enhancing the method's practicality and stability. It provides high-precision baseflow data for water resource assessment, ecological flow verification, and drought early warning, contributing to the refined and intelligent transformation of water resource management. Attached Figure Description
[0021] Figure 1 This is a flowchart of the improved baseflow separation method of the Kalinin River according to Embodiment 1 of the present invention; Figure 2 This is a comparison chart of precipitation and runoff data from a certain hydrological station; Figure 3 This is a graph showing the calculated average confluence time of the watershed; Figure 4a This is a selection map of all extreme points in the runoff sequence; Figure 4b It is a magnified view of all extreme points in the runoff sequence; Figure 5a This is a diagram of a clean drainage section without precipitation disturbance; Figure 5b This is a magnified view of a clean drainage section without precipitation disturbance; Figure 6 It is a standard drainage curve fitting graph; Figure 7a This is a diagram showing the base current segmentation results of the entire sequence; Figure 7b This is a magnified view of the base current segmentation results. Detailed Implementation
[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0023] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments: Example 1 like Figure 1 As shown, this embodiment provides an improved baseflow segmentation method for the Kalinin River, including the following steps: Step 1: Data Preparation and Preprocessing (1) Data preparation Prepare long-term series of river runoff process data and concurrent precipitation data, unify the source data series process, and standardize them into daily data.
[0024] The long series of river runoff process data and concurrent precipitation data are presented as follows:
[0025] in, This represents the average precipitation in the basin, in mm. The corresponding river cross-section runoff is expressed in cubic meters (m³). 3 / s; the time resolution is 1 day, and T is the total number of days in the sequence.
[0026] (2) Data standardization processing The data was standardized to remove interference fluctuations from low precipitation and runoff. Specifically, the following methods were used: invalid precipitation was removed based on an effective precipitation of 0.5 mm; the minimum baseflow value was estimated based on the average of the second lowest monthly values over many years, the minimum baseflow value was removed, and non-negative processing was performed.
[0027] Step 2: Calculate the average convergence time and then calculate the maximum convergence time based on the average convergence time. The delayed sequence is constructed as follows:
[0028] in, The watershed average precipitation series calculated based on the correlation coefficient. The river cross-sectional runoff sequence is calculated using correlation coefficients. The river cross-sectional runoff is the length of the trial period, dT, which is the delay time.
[0029] The correlation coefficients under different confluence time conditions were calculated separately. The formulas for calculating the correlation coefficients under different confluence time conditions are as follows:
[0030] in, The mean of the watershed average precipitation series is calculated using the correlation coefficient. The mean of the river cross-sectional runoff sequence calculated using the correlation coefficient.
[0031] The correlation coefficients under different confluence time conditions were calculated, and the confluence time corresponding to the maximum correlation coefficient was taken as the average confluence time, denoted as . Considering the flattening and tailing delay effects of runoff processes relative to precipitation events, when automatically extracting precipitation events and corresponding runoff event periods, it is also necessary to consider the maximum runoff event end time that the precipitation event can influence. Therefore, in addition to analyzing and calculating the average runoff time, it is also necessary to estimate the maximum runoff time. Based on hydrological experience, the surface runoff tailing effect at piedmont hydrological stations generally does not exceed twice the average runoff time. Furthermore, considering that the runoff time may be less than one day, 2× +1 is set as the maximum confluence time, denoted as... .
[0032] Step 3: Identification and extraction of secondary floods and typical drainage processes Typical drainage curves are crucial for analyzing and fitting standard drainage curves, and are key parameters for the fitting process. A standard drainage curve can be understood as a scenario where all groundwater runoff from any point on the curve (at a given flow rate) is groundwater runoff, with no subsequent precipitation. Therefore, the selection of typical drainage curves should prioritize runoff processes with a relatively small proportion of surface runoff and virtually no subsequent precipitation until the most recent low point (end point). Since the analysis typically involves long series and numerous drainage processes each year, manual selection would be time-consuming and labor-intensive. To improve efficiency and lower the technical barrier to this method, this invention proposes an automatic analysis method for selecting drainage curves, replacing manual selection. The specific process is as follows: Based on flood characteristics, runoff data exhibits alternating troughs and peaks. Ideally, one alternation constitutes one flood. However, in reality, due to the continuous and intermittent nature of precipitation, multiple flood peaks within a runoff event are the norm. Furthermore, due to the intermittent nature of continuous precipitation and the inherent errors in flow measurement, when actual flow changes are not significant, the flow path exhibits localized oscillations, further complicating flood event identification. To achieve automatic identification of flood events, this embodiment employs the following processing: (1) The runoff process was smoothed by using the mid-moving average method, and the minimum and maximum monthly average runoff volumes for each year were statistically analyzed.
[0033] The purpose of smoothing the runoff process using the mid-moving average method is to eliminate objectively existing flow measurement errors and the oscillations inherent in the runoff from the mountain piedmont. The window parameter for smoothing the runoff process is the maximum confluence time. After smoothing, the smoothed value of the runoff is obtained. .
[0034] The annual minimum and maximum monthly average runoff statistics are as follows:
[0035] in, The minimum monthly average runoff each year, The maximum monthly average runoff each year, The average monthly runoff over the years, For months.
[0036] (2) All extreme points of the runoff sequence are extracted as follows:
[0037] in, This is the time series corresponding to the local maxima of runoff. These are the time series corresponding to the local minimum values of runoff. To calculate the smoothed value of runoff at the specified time, To calculate the smoothed value of runoff from the time preceding the current time, This is the smoothed value of the runoff volume at the time following the calculation time.
[0038] By deleting the first or last extreme point, and The lengths are equal, and the sequence is guaranteed to start with a peak and end with a valley, with peak-valley alternation.
[0039] (3) Composite flood peak identification and valley peak merging processing Composite peak identification and valley peak merging processing require the following conditions to be met simultaneously: ① That is, the time difference between the rear front and the front valley is less than the maximum confluence time; among which, For the first The time corresponding to each peak For the first Each valley value corresponds to .
[0040] ② That is, the measured runoff in the rear valley is less than that in the front valley, where, For the first The runoff corresponding to each valley value For the first The runoff corresponding to each valley value ③ The increase in the value of the subsequent peak compared to the preceding trough is less than 20% of the amplitude of the current recession process. This refers to the amplitude of the current receding water process. For the first The runoff corresponding to each peak.
[0041] The processing of composite flood peaks adopts an iterative approach, handling only one flood peak at a time. For flood peaks that meet the conditions and are processed together, the peak-valley process needs to be deleted and the process updated. And the peak and trough values corresponding to After the update is completed, the processed sequence is substituted into the iteration for the next processing, until there are no composite flood peaks that meet the conditions in the entire sequence, and the iteration ends. The entire iteration process is carried out from front to back in time.
[0042] (4) Review and processing of flood peak process Because the flood peak precipitation process (peak-valley process) is composed of smoothed runoff values This approach avoids fluctuations caused by measurement errors, sporadic precipitation, and confluence interference, and is significant for determining the overall characteristics of the flood peak. However, the final analysis still uses the measured runoff value, not the smoothed runoff value. Therefore, it is necessary to conduct a detailed review and correction of the start and end points of the peak-valley process. For peak and trough points, the measured runoff values should be used for forward and backward correction.
[0043] Taking peak point correction as an example: Let the time point at which the i-th peak occurs be . The initial value of the peak value is Then, the following process is repeated iteratively to achieve the correction until α stops changing, i.e. Until it remains unchanged.
[0044]
[0045] in, It is the maximum value among the values at the time points before the initial peak value, the initial peak value, and the value at the time points after the initial peak value. This is the value at the time point preceding the initial peak value. This is the value at the point one time after the initial peak value.
[0046] (5) Extraction of complete dewatering process By identifying and merging composite flood peaks, individual flood events are obtained. However, not every flood event has a complete recession process. For example, during the flood season and the plum rain season in southern China, when the basin area is large, it is difficult to have a continuous period without precipitation. This can even lead to multiple floods occurring throughout the flood season or for several consecutive months. Because each recession does not continue to the complete baseflow portion before the next flood occurs, several floods do not have a complete recession process. The Kalinin method, on the other hand, deals with multiple complete recession processes. Therefore, it is necessary to identify and extract complete and clean recession processes (without precipitation interference during the process).
[0047] The following criteria are set to screen for complete dewatering processes that meet the requirements of the Kalinin method: ①The entire receding process is long enough, at least not less than the maximum confluence time; ② The peak point must be greater than the minimum annual maximum monthly average runoff to eliminate incomplete drainage processes caused by small precipitation. ③ The trough point must be less than the annual minimum monthly average runoff to eliminate the impact of the re-peak during the receding process; ④ Extraction of the tail end of the receding water: For the remaining complete receding water process, threshold filtering is used to extract the receding water process. Upper limit filtering is used to ensure that the receding water part belongs to the tail end (i.e., the receding water section) rather than the head containing significant surface runoff receding water characteristics.
[0048] (6) Screening of drainage sections without precipitation A single-peak flood hydrograph formed by a single rainfall event includes the flooding phase, the peak phase, and the recession phase. The shape of the flooding phase is mainly influenced by the characteristics of the rainfall storm. The inflection point of the recession phase is generally considered to occur when surface runoff ceases. The Kalinin method requires that the recession process begin from the inflection point of the recession phase and end at the lowest point, while also requiring that no effective precipitation with a significant impact on runoff occurs during this period. Therefore, it is necessary to perform precipitation-free screening for each tail end of the recession process. The specific method is as follows: First, extract the precipitation time period corresponding to the tail end of the receding water flow. The time difference between the precipitation time period and the tail end of the receding water flow is the average confluence time. For example, if the receding water level occurs from day m to day m+n, then the concurrent precipitation is... to Day. Areas with average precipitation for the same period exceeding 0.4 times the annual average precipitation are considered to have significant precipitation. The default value for this parameter is 0.4, but users are allowed to adjust it appropriately during actual operation. The smaller the parameter, the stricter the screening.
[0049] Step 4: Improvement of the Kalinin method and base current segmentation To adapt to one-click baseflow segmentation driven by pure data and without manual judgment, some aspects of the Kalinin method need to be adjusted. The Kalinin method is based on the characteristic that river baseflow is generally supplied by groundwater from bedrock fissures, and assumes that there is a proportional relationship between the amount of water discharged from the aquifer into the river (i.e., river baseflow) and surface runoff (including overland runoff and soil runoff). A reasonable proportional coefficient is determined through trial calculations, and the annual river baseflow is obtained through repeated calculations of the basic water balance equation.
[0050] (1) Basic water balance equation
[0051] in, To calculate the aquifer reserves at the end of the time period, To calculate the initial aquifer reserves for the time period, This is the proportionality coefficient. To calculate surface runoff over a given period, To calculate the river base flow rate within the specified time period.
[0052] (2) Equation of the receding water curve The receding water section represents the underground water reservoir replenishing river runoff. In this embodiment, the receding water section without precipitation, which has been screened, can be considered as the release process of underground runoff. In the Kalinin method, the underground reservoir equation is often used to describe the receding water section without precipitation. The groundwater has a free surface, and the underground water storage capacity... With outflow There is a linear relationship between them, that is:
[0053] in, Let be the runoff time of groundwater. When there is no infiltration recharge or evaporation loss of groundwater during the receding period, the differential relationship between the base outflow and time is as follows:
[0054] As can be seen from the differential equation, for an undisturbed, naturally flowing baseflow process, once the initial baseflow rate is determined, the subsequent changes in the baseflow rate over time are also determined, and they are related as follows:
[0055] in, The flow rate is t days after the start of the receding water level. This refers to the flow rate at the start of the receding water level. t represents the decay coefficient; t represents time, in days.
[0056] (3) Base current division calculation equation By transforming and substituting into the water balance equation, the final baseflow division calculation equation is obtained as follows:
[0057] in, The calculation period is generally one day; This represents the measured total runoff.
[0058] When the extinction coefficient is determined and proportionality coefficient Then, the river base flow can be calculated using the above formula, and the recession coefficient can be calculated. and proportionality coefficient Determined by the actual regional hydrogeological conditions, and at the same time, the proportional coefficient It is also affected by the actual annual precipitation pattern, including the recession coefficient. It can be considered constant over shorter timescales. Although the decay coefficient... and proportionality coefficient The coefficient of regression is determined by the actual hydrogeological conditions of the region, but cannot be directly measured; it needs to be calculated through data inversion. Once the receding process curves are determined, they need to be shifted and superimposed to obtain a cluster of receding process curves. The receding coefficient is then determined from the lower envelope of these clusters. Value. Proportionality coefficient. This needs to be determined through trial calculations, the process of which involves assuming a relatively large scaling factor. The value is used to calculate a large baseflow process. It is then determined whether this baseflow process curve is larger than the total runoff process curve. If so, the proportionality coefficient is gradually decreased. The value is maintained until the baseflow does not exceed the measured total runoff throughout the entire process.
[0059] (4) Identification process of extinction coefficient and proportionality coefficient The calculation principle and process of the entire method have been clearly explained above. Finally, it is only necessary to determine the fading coefficient. and proportionality coefficient This allows for the calculation of the groundwater (baseflow) process based on the measured total runoff process. Parameters or The calculations are all related to the measured total runoff. However, the measured runoff is objectively subject to non-trend distortions caused by measurement errors, omissions, sporadic precipitation, and runoff confluence. These local distortions may affect the calculated parameters. or Outliers have occurred. Therefore, to improve the universality and adaptability of the method of this invention, adaptive processing is still needed for certain steps in actual use. This mainly includes the self-identification process of the recession curve equation (recession coefficient). Identification process), base current segmentation trial calculation process (proportional coefficient B identification process).
[0060] ① Extinction coefficient Identification process Calculate the extinction coefficient At this time, it is necessary to translate and superimpose the receding process lines to obtain a cluster of receding process curves, and determine the lower envelope of the curves. Theoretically, the measured total runoff during all recession processes should lie entirely above the lower envelope. However, in practice, due to various errors caused by different factors in the measured runoff, the measured total runoff may have a minimum value or even a value of 0. If the lower envelope is drawn strictly according to the measured runoff value and the recession coefficient is determined... This would obviously lead to distorted calculation results. Therefore, this method introduces a penalty coefficient, meaning it does not strictly require all measured runoff points to be above the lower envelope; it allows some points to be below the lower envelope. The more points below the lower envelope and the higher the percentage below the lower envelope, the greater the penalty, and the larger the calculated penalty value, with the cumulative penalty value not exceeding 1. In other words, it does not require all points to be strictly above the lower envelope; it only requires that the vast majority of points be above the lower envelope. The specific calculation method for the penalty value is as follows: Suppose a measured point i is below the lower bound line. Then the penalty value and cumulative penalty value of measured point i are calculated according to the following formula:
[0061]
[0062] in, Let F be the penalty value at the measured point i, and let F be the cumulative penalty value. The number of typical receding water processes. Let i be the measured runoff value at the measured point. The calculated value of the lower envelope corresponding to the measured point i. This is the penalty coefficient.
[0063] The penalty coefficient can be selected by the user. It is defined as applying a fitting penalty term if the measured flow rate is lower than the standard recession curve. A larger penalty coefficient value indicates a more stringent requirement, meaning that the data from each flow measurement process is considered highly reliable and error-free, and the standard recession curve is always below the typical recession curve, and vice versa. The penalty coefficient is adjustable from 1 to 100, with a default value of 10 based on hydrological experience.
[0064] ② Identification process of proportional coefficient B proportionality coefficient The trial calculation process assumes a large scaling factor. The value is used to calculate a large baseflow process. It is then determined whether this baseflow process curve is larger than the total runoff process curve. If so, the proportionality coefficient is gradually decreased. The value is calculated until the baseflow does not exceed the measured total runoff throughout the entire process. For the same reason, the measured total runoff may have a minimum value or a value of 0. If the measured runoff value is strictly used for judgment, the proportionality coefficient will also be affected. The calculation results are distorted. For example, some hydrological stations may experience flow interruptions, with measured flow rates of 0, while the calculated baseflow value is always positive. Therefore, regardless of how the B value is selected, it cannot be guaranteed that the baseflow value at that hydrological station is less than the measured runoff. Thus, an allowable range of error between the trial calculation results and the measured runoff is set.
[0065] Based on experience with the measured runoff accuracy of basic hydrological stations in my country, assuming an error of approximately 5% to 10%, the allowable error range should, in principle, not exceed half of the measured error, i.e., 2.5% to 5%. In this embodiment, a default value of 3% (0.03) is set. Users can adjust this parameter based on experience, with an adjustment range of 0.01 to 0.1. A larger value indicates a greater allowable error in measuring low flow rates (multi-year average minimum monthly average flow), resulting in a larger calculated groundwater runoff and more instances where the calculated baseflow exceeds the total runoff, and vice versa. (This is related to the recession coefficient.) The approach is similar, which means that the calculated baseflow is not required to be strictly less than the measured runoff throughout the entire process. Instead, a default measurement error is set for the measured runoff process, as long as the baseflow does not exceed the error range of the measured runoff.
[0066] Application Cases The method of this invention will be described using a hydrological station as an example. This hydrological station is a relatively typical piedmont hydrological station, with no significant human activities or engineering impacts upstream of the control section, and the measured runoff can be considered as natural runoff.
[0067] (1) Data preparation and preprocessing Prepare runoff and concurrent precipitation data for this hydrological station, unify the source data series, and standardize them into daily data. Empirically process the source data, calculating effective precipitation at 0.5 mm, estimating the minimum baseflow based on the second-lowest monthly runoff over multiple years, removing the minimum baseflow value, eliminating invalid precipitation, and performing non-negative processing. The precipitation data is the watershed average precipitation after processing, which is the area-weighted average (in mm) of all associated rainfall stations within the watershed. Figure 2As shown, the horizontal axis is in days. For ease of calculation, the entire process uses the day number instead of year, month, and day. In this case, the first day is January 1, 1992, and the last day is December 4, 2022, with a total sequence length of 11,296 days. When the sequence is long enough, specific years of data can be selected as needed; generally, a sequence of at least 5-10 years is required.
[0068] (2) Calculation of average confluence time and maximum confluence time Trial calculations were performed for different confluence times. In this case, the trial period was three months. Generally, it is acceptable as long as the trial period is significantly longer than the empirical confluence time. For typical small and medium-sized watersheds, this usually does not exceed 1 to 3 months. The calculation results are as follows: Figure 3 As shown. (Through) Figure 3 The calculation results show that the correlation coefficient is highest when the average confluence time of the site is 2 days, i.e., the average confluence time and the maximum confluence time are respectively... , .
[0069] (3) Identification and extraction of secondary floods and typical water recession processes First, select all extreme points and process them as a peak-valley process, see... Figure 4a and Figure 4b As can be seen from the figure, the selected local peak-valley process obviously cannot represent the entire recession process, as it is affected by data oscillations, repeated peaks, sporadic precipitation, and confluence interference. Therefore, it is necessary to identify and extract typical recession curves for each flood event. The method of this invention can achieve automated extraction of typical recession processes, as detailed in step 3. According to step 3, each process is identified and processed to extract typical baseflow recession processes without precipitation interference. After processing, a total of 92 recession segments are obtained, averaging about 3 segments per year. See [link to relevant documentation]. Figure 5a and Figure 5b .
[0070] (4) Analysis of fading coefficient Theoretically, the standard recession curve is defined as groundwater runoff. Any recession process selected lies on the standard recession curve. Therefore, determining the actual standard recession curve should first involve fitting multiple typical recession processes to make their tails overlap as much as possible, and then calculating a suitable recession coefficient. This causes the standard backwater curve to lie below all typical backwater processes, exhibiting a form similar to a lower envelope.
[0071] However, due to measurement errors and the influence of some human activities and water conservancy projects, in practice, it cannot be guaranteed that all drainage sections are completely free of surface runoff, and that baseflow drainage sections will not unexpectedly decrease due to interference from other factors. Therefore, a penalty coefficient is provided as a factor to account for the influence of flow measurement errors and other factors. The penalty coefficient can be selected by the user. It is defined as applying a fitting penalty term if the measured flow is lower than the standard drainage curve. A larger penalty coefficient value indicates a more stringent requirement, meaning that the data from each flow measurement process is considered highly reliable and error-free, and the standard drainage curve is always below the typical drainage process; conversely, the opposite is true. The penalty coefficient is adjustable from 1 to 100, with a default value of 10 based on hydrological experience. The fitting results using the default value can be found in [link to relevant documentation]. Figure 6 As can be seen from the figure, the fitted lower envelope does not completely guarantee that all flow values lie above the lower envelope. This is the result after processing with a penalty coefficient, which is consistent with actual hydrological experience. The calculation yielded... The value is 0.1008 (d) -1 ).
[0072] (5) Basestream segmentation Once the receding water curve is determined, the surface runoff can be calculated and segmented to obtain the optimal solution. Theoretically, the segmented groundwater runoff should always be less than the measured runoff, that is, less than the sum of surface and groundwater runoff. However, due to the objective existence of flow measurement errors, especially at low flow rates, the measured runoff value may be less than the actual groundwater runoff. Therefore, an allowable range of error between the calculated result and the measured runoff is set.
[0073] The baseflow segmentation of this hydrological station was calculated using a default value of 0.03. The calculation results are shown below. Figure 7a and Figure 7b As shown. Furthermore, the theoretical standard receding flow curve can approach zero infinitely, but will not equal zero. Considering that in actual processes, due to factors such as river evaporation and underground flow, there are indeed instances of interrupted flow in the measured runoff, to avoid the influence of extremely small flow values on the calculation results, the lowest monthly runoff over many years was first calculated. This flow rate is generally very close to zero (taking a certain hydrological station as an example, the flow rate at this station is 1.32 m³ / h). 3 The baseflow is calculated per second ( / s). Values less than this are all considered groundwater runoff. In practice, various situations may occur where the runoff drops instantly or even stops at a low value. Therefore, the calculated groundwater runoff will inevitably exceed the measured value, resulting in a trial-and-error error. In this case, the calculated groundwater runoff value is automatically adjusted to the measured runoff. In this case, the final trial-calculated baseflow segmentation result for the hydrological station is as follows: Figure 7a and Figure 7b As shown, the base diameter ratio for the entire series is 0.1494, meaning that the multi-year average base flow accounts for 14.94% of the total runoff.
[0074] This invention integrates hydrological experience into the algorithm, enabling accurate identification of secondary floods in long-sequence runoff data, extraction of secondary flood recession hydrographs, and screening and analysis of these hydrographs. Based on this, the classic Kalinin method is adaptively improved, optimizing the convergence conditions and initial value settings for baseflow iterative calculations, thus enabling fully automated processing while maintaining the physical mechanisms. This method completely eliminates the traditional operating mode of using transparent grid paper and manual visual interpretation, achieving one-click intelligent generation of baseflow segmentation data sequences from raw runoff data. The operational application of this method will bring significant benefits. The time required for single-station annual runoff baseflow segmentation has been reduced from several person-days to minutes, enabling watershed-scale batch analysis and supporting rapid updates to national water resources surveys and assessments. Unified empirical criteria eliminate individual subjective differences, ensuring strict spatiotemporal comparability of segmentation results and providing objective evidence for management work such as ecological flow assessment and water resources rights confirmation. The method can be embedded in hydrological analysis, forecasting, and simulation systems for real-time operation, providing dynamic baseflow monitoring for drought early warning and supporting high-resolution simulations of baseflow evolution under climate change scenarios. This method is not only an innovation in hydrological technology but also a significant practice in the intelligent and standardized transformation of hydrological data compilation and analysis, with profound implications for improving the scientific and refined level of water resources management in my country.
[0075] This invention algorithmizes hydrological empirical rules, achieving fully automated processing and completely eliminating manual intervention. The processing time per station is reduced from several person-days to minutes. By using unified identification and parameter calibration standards, it eliminates subjective human differences and ensures the spatiotemporal comparability of baseflow segmentation results. By introducing penalty coefficients and error tolerance mechanisms, it adapts to different watershed data quality, making the method highly universal and stable. It supports batch processing of long sequences and multiple stations and can be seamlessly embedded into hydrological forecasting and simulation systems. It provides high-precision, standardized baseflow data for water resource assessment, ecological flow verification, and drought early warning, effectively promoting the transformation of hydrological data compilation towards intelligence and standardization.
[0076] Example 2 This embodiment provides a river baseflow segmentation system, including: a data preprocessing module, a confluence time analysis module, a flood process identification module, and a baseflow segmentation module; The data preprocessing module is used to acquire and standardize daily runoff and precipitation data; The runoff time analysis module is used to determine the average runoff time and the maximum runoff time through precipitation-runoff delay correlation analysis; The flood process identification module is used to smooth the runoff sequence, extract extreme points and perform composite flood peak merging and peak-valley verification, and extract complete clean receding sections without effective precipitation according to set conditions. The baseflow segmentation module is used to determine the fading coefficient by combining the clean water effluent section with the penalty mechanism, calculate the proportional coefficient by combining the error tolerance mechanism, and output the baseflow segmentation result based on the water balance calculation.
[0077] Example 3 An electronic device includes a memory and a processor, the memory being used to store a program that supports the processor in executing the improved Kalinin River baseflow separation method of Embodiment 1, the processor being configured to execute the program stored in the memory.
[0078] Example 4 A storage medium storing a computer program, which, when executed by a processor, performs the steps of the improved Kalinin River baseflow separation method in Embodiment 1.
[0079] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. An improved Kalininskaya method of separating base flow from stream flow, characterized in that, Includes the following steps: Acquire long-sequence daily river runoff data and concurrent daily precipitation data, and perform standardized preprocessing; Based on the preprocessed data, the average runoff time and maximum runoff time were determined using a delay correlation analysis between precipitation and runoff. The daily runoff sequence is smoothed to eliminate data oscillations, local extreme points are extracted, composite flood peaks are identified and merged to obtain the flood process, and the peak and trough points of the flood process are reviewed and corrected based on the measured runoff data. Typical drainage processes are extracted from flood events based on preset drainage process screening conditions, and drainage sections without precipitation are screened in conjunction with precipitation data to obtain a set of clean drainage sections. All clean water drainage curves are plotted on the same coordinate system with their tails overlapping. The lower envelope of the drainage curve cluster is fitted with an exponential drainage empirical formula. A penalty mechanism is introduced to allow a very small number of measured flow values to be lower than the lower envelope. The fitted lower envelope is the standard drainage curve, and the exponential parameter is the drainage coefficient. The measured runoff was divided using a standard receding flow curve, and a flow measurement error tolerance mechanism was introduced. The allowable proportion of the base flow exceeding the measured runoff was set, and the proportion coefficient was determined by trial calculation. After the decline coefficient and proportionality coefficient are determined, the measured runoff is divided into baseflows according to the Kalinin method baseflow division formula.
2. The improved baseflow separation method for the Kalinin River according to claim 1, characterized in that, The method for determining the average runoff time and maximum runoff time using the delay correlation analysis of precipitation and runoff is as follows: The delayed sequence is constructed as follows: wherein, is a sequence of basin average precipitation calculated for the correlation coefficient, is a sequence of river cross-section runoff calculated for the correlation coefficient, is a river cross-section runoff with a delay of dT, dT being the trial period length. The correlation coefficients under different confluence time conditions were calculated separately. The formulas for calculating the correlation coefficients under different confluence time conditions are as follows: wherein is the mean of the sequence of the basin average precipitation calculated for the correlation coefficient, is the mean of the sequence of the river cross section discharge calculated for the correlation coefficient; The correlation coefficient under different confluence times is calculated, the confluence time corresponding to the maximum correlation coefficient is taken as the average confluence time, and is recorded as ; and 2× +1 is set as the maximum confluence time, and is recorded as .
3. The improved baseflow separation method for the Kalinin River according to claim 2, characterized in that, The identification and merging of composite flood peaks must simultaneously meet the following conditions: The time difference between the next peak and the previous trough is less than the maximum confluence time; The runoff corresponding to the next trough is less than the runoff corresponding to the previous trough. The difference between the peak runoff and the trough runoff is less than 20% of the amplitude of the current recession process; If the conditions are met, the peak-valley process is merged, the current valley point and the next peak point are deleted from the extreme point sequence, the two peak-valley processes are merged into one, the sequence is updated and iteratively processed until there is no composite flood peak that meets the conditions.
4. The improved Kalinin river channel baseflow separation method of claim 3, wherein, The preset screening conditions for the dewatering process include: The duration of the receding process shall not be less than the maximum confluence time; The peak runoff is greater than the annual maximum monthly average runoff that was the smallest in previous years. The trough runoff was less than the historical minimum monthly average runoff. For the drainage process that meets the above conditions, threshold filtering is used to truncate the tail end of the drainage and remove the head end which contains significant surface runoff characteristics.
5. The improved Kalinin River channel baseflow separation method of claim 4, wherein, The method for screening the receding water section without precipitation is as follows: for each receding water tail period, select the corresponding precipitation period, and the corresponding precipitation period is the average confluence time of the entire receding water period shifted forward; calculate the average precipitation in the precipitation period, and if it is greater than a preset multiple of the annual average precipitation, then the receding water section is removed.
6. The improved Kalinin River channel baseflow separation method of claim 5, wherein, The penalty mechanism specifically involves calculating a penalty value for measured flow points that are below the standard drainage curve. and cumulative penalty value ,in, Let F be the penalty value at the measured point i, and let F be the cumulative penalty value. The number of typical receding water processes. Let i be the measured runoff value at the measured point. The calculated value of the lower envelope corresponding to the measured point i. This is the penalty coefficient; it is adjusted so that the cumulative penalty value does not exceed 1, allowing a small number of measured points to lie below the standard recession curve. The flow measurement error tolerance mechanism is as follows: if the measured runoff error is 5% to 10%, then the allowable error range shall not exceed half of the measured error.
7. The improved Kalinin River channel baseflow separation method of claim 6, wherein, The empirical formula for exponential recession is as follows: The Kalinin method's basic current segmentation formula is as follows: ;in, The flow rate is t days after the start of the receding water level. The flow rate at the start of the receding water level. The decay coefficient is denoted by t, where t is time. To calculate the aquifer reserves at the end of the time period, This is the proportionality coefficient. To measure the total runoff, To generate traffic, This is the calculation period.
8. A channel baseflow segmentation system based on the improved Kalinin River baseflow segmentation method according to any one of claims 1 to 7, characterized in that, include: The module includes a data preprocessing module, a confluence time analysis module, a flood process identification module, and a baseflow segmentation module. The data preprocessing module is used to acquire and standardize daily runoff and precipitation data; The runoff time analysis module is used to determine the average runoff time and the maximum runoff time through precipitation-runoff delay correlation analysis; The flood process identification module is used to smooth the runoff sequence, extract extreme points and perform composite flood peak merging and peak-valley verification, and extract complete clean receding sections without effective precipitation according to set conditions. The baseflow segmentation module is used to determine the fading coefficient by combining the clean water effluent section with the penalty mechanism, calculate the proportional coefficient by combining the error tolerance mechanism, and output the baseflow segmentation result based on the water balance calculation.
9. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store programs that enable the processor to execute the improved Kalinin River baseflow separation method according to any one of claims 1 to 7, and the processor is configured to execute the programs stored in the memory.
10. A storage medium having stored thereon a computer program, characterized in that When a computer program is run by a processor, it performs the steps of the improved Kalinin River baseflow separation method according to any one of claims 1 to 7.