A method for correcting salinity profile data of MVP system based on temperature gradient profile analysis

Abstract

This invention discloses a method for correcting salinity profile data of an MVP system based on temperature gradient profile analysis. The steps include: preprocessing the MVP profile data; correcting the response time of the preprocessed MVP profile data using an optimal sensor response time correction algorithm; finding the optimal thermal hysteresis correction coefficient for each thermocline and non-thermocline using an enumeration optimization algorithm; calculating the corrected temperature data using the optimal thermal hysteresis correction coefficient; and calculating the salinity using conductivity data, pressure data, and the corrected temperature data. This method significantly reduces the salinity difference between the uplink and downlink MVP measurement profiles, essentially eliminating salinity peaks, particularly showing significant improvement at thermoclines. The average absolute salinity difference between the uplink and downlink profiles decreases from 0.04 psu to 0.014 psu, and the overall salinity error is reduced by 65%.

Description

Technical Field

[0001] This invention relates to a method for salinity profile correction of an MVP system, and more particularly to a method for salinity profile correction of an MVP system based on temperature gradient profile analysis. Background Technology

[0002] The Moving Vessel Profiler (MVP) is a highly integrated and automated rapid temperature, salinity, and depth (TDM) measurement platform. It can perform high-resolution TDM and salinity profile observations in both vertical and horizontal directions while the ship is moving at high speed. It has significant advantages in observing the fine TDM and salinity structure of ocean mesoscale processes, such as mesoscale eddies and fronts.

[0003] The MVP probe, equipped with a pumpless CTD, employs a "free-fall" method for profile measurement, relying on the instrument's motion to generate the flow rate through the conductivity cell. As the instrument's descent speed changes, the flow rate through the conductivity cell is not constant. Due to the rapid and non-uniform descent speed of the MVP probe, using traditional pump-type CTD salinity correction methods to correct the salinity data measured by the MVP will produce significant salinity spikes. This is mainly caused by two factors: the mismatch between the temperature and conductivity sensor response times and the thermal hysteresis effect of the conductivity cell.

[0004] Firstly, the mismatch in response time between temperature and conductivity sensors is due to differences in the physical characteristics (working principle and physical properties) of temperature and conductivity sensors, different measurement ranges, and design or manufacturing defects of the sensors themselves. These differences in response time result in temperature and conductivity data being either ahead or behind.

[0005] Secondly, the thermal hysteresis effect of the conductivity cell is mainly due to the fact that the conductivity cell material of the CTD itself has the ability to store heat. The conductivity cell wall will slowly heat / cool the seawater flowing through the conductivity cell. The conductivity sensor is not as sensitive to changes in the surrounding environment as the temperature sensor, so a time lag will occur, namely the thermal inertia / thermal hysteresis effect. Summary of the Invention

[0006] Purpose of the invention: To provide a method for salinity profile correction of the MVP system based on temperature gradient profile analysis, which can significantly reduce the salinity difference between the uplink and downlink measurement profiles of the MVP.

[0007] Technical solution: The MVP system salinity profile data correction method based on temperature gradient profile analysis described in this invention includes the following steps:

[0008] Step 1: Preprocess the MVP profile data to obtain effective and smooth MVP profile data;

[0009] Step 2: Use the optimal sensor response time correction algorithm to correct the response time of the preprocessed MVP profile data;

[0010] Step 3: Divide the temperature profile into various thermoclimate layers and non-thermoclimate layers, and use an enumeration optimization algorithm to find the optimal thermal hysteresis correction coefficient for each layer's thermal inertia correction algorithm. Then, use the optimal thermal hysteresis correction coefficient to calculate the corrected temperature data.

[0011] Step 4: Calculate the salinity using conductivity data, pressure data, and corrected temperature data.

[0012] Furthermore, in step 1, the specific steps for preprocessing the MVP profile data are as follows:

[0013] Step 1.1: Remove MVP profile data within the range of 6-10 dBar after the CTD instrument is started and activated;

[0014] Step 1.2: Eliminate reverse pressure data to obtain stable vertical MVP profile data;

[0015] Step 1.3: Remove outliers from the MVP profile data based on the set conductivity threshold, pressure threshold, temperature threshold, and salinity threshold.

[0016] Step 1.4: Use low-pass filtering to remove noisy data from the MVP profile data.

[0017] Furthermore, in step 1.2, the reverse pressure data includes data collected when the instrument moves in the opposite direction during the instrument's descent or ascent, as well as data collected when the instrument's ascent and descent speed is less than the speed threshold.

[0018] Furthermore, in step 1.3, when removing outliers from the MVP profile data, the conductivity threshold is set to 0-6 S / m, the pressure threshold is set to 1.1 times the pressure value of the maximum profile depth, the temperature threshold is set to -2.5 to 40℃, and the salinity threshold is set to 0 to 41 psu. If the conductivity data is not within the range of 0-6 S / m, it is removed; if the pressure data is greater than 1.1 times the maximum profile depth, it is removed; if the temperature data is not within the range of -2.5 to 40℃, it is removed; if the salinity data is not within the range of 0 to 41 psu, it is removed.

[0019] Furthermore, in step 1.4, when using low-pass filtering to remove noise data from the MVP profile data, temperature and conductivity data are filtered using a 4-point low-pass filter with a cutoff time of 0.16s; pressure and vertical velocity data are filtered using a 32-point low-pass filter with a cutoff time of 1.28s.

[0020] Furthermore, in step 2, the specific steps for response time correction of the preprocessed MVP profile data are as follows:

[0021] Step 2.1: Calculate the first-order temperature difference dT and the first-order conductivity difference dC for each data point in each set of profile data. Then, use the xcorr function in Matlab to calculate the lag correlation coefficient r and lag time difference lags of the first-order temperature difference dT and the first-order conductivity difference dC.

[0022] Step 2.2: Find the extreme data point corresponding to the maximum value of the lag correlation coefficient r. Then, establish a binomial fitting equation with the lag correlation coefficient r as the dependent variable and the lag time difference lags as the independent variable. Then, use the lag correlation coefficient r and the lag time difference lags of the extreme data point, the point before the extreme data point, and the point after the extreme data point to fit the binomial fitting equation. The maximum value of the binomial fitting equation is the maximum lag correlation coefficient r. The independent variable corresponding to the maximum lag correlation coefficient r is the corrected lag time difference lags between the temperature data and the conductivity data.

[0023] Step 2.3: Correct the acquisition time of conductivity data according to the correction lag time difference of 1ags, and then interpolate the corrected conductivity data to the same acquisition time as the temperature data to achieve response time correction; or correct the acquisition time of temperature data according to the correction lag time difference of 1ags, and then interpolate the corrected temperature data to the same acquisition time as the conductivity data to achieve response time correction.

[0024] Furthermore, in step 3, the specific steps for dividing the temperature profile into various thermoclimate layers and non-thermoclimate layers are as follows:

[0025] The up-line and down-line profile data in each set of profile data are interpolated and converted according to the set height interval, so that the up-line and down-line profile data are unified to the same height interval.

[0026] The temperature gradient values ​​for the uplink and downlink profiles were calculated separately:

[0027]

[0028] In the formula, G k Z represents the temperature gradient value at the k-th data point. k T represents the measured depth value of the k-th data point. k This represents the measured temperature value of the k-th data point;

[0029] If G k If the temperature is ≤0.05℃ / m, then the kth data point is determined to be within the thermocline. If G k If the value is greater than 0.05℃ / m, then the kth data point is determined to be a point within the non-jump layer.

[0030] Furthermore, after determining the thermocline and non-thermocline at each data point, if the thickness of a thermocline is less than the thickness threshold (10m), the thermocline is merged into the adjacent thermocline and becomes part of the adjacent thermocline.

[0031] Furthermore, in step 3, the specific steps for finding the optimal thermal hysteresis correction coefficient for each thermal inertia correction algorithm in each thermoclimate and non-thermoclimate layer using an enumeration optimization algorithm are as follows:

[0032] Define the range of variation for the thermal hysteresis correction coefficients α and τ, and form M×N pairs of α and τ coefficients;

[0033] For each thermocline and non-thermocline, traverse M×N pairs of α and τ coefficients, and calculate the M×N temperature corrections T for each thermocline and non-thermocline using the following formula. T for:

[0034] T T (n)=-bT T (n-1)+a[T(n)-T(n-1)]

[0035] In the formula, T(n) is the measured temperature value at the nth point, and the formulas for calculating a and b are:

[0036]

[0037]

[0038] In the formula, α and τ are the thermal hysteresis correction coefficients, i.e., the amplitude of the error and the time constant, respectively, and f n The sampling frequency;

[0039] Then, using the measured temperature values ​​T(n) at each point in each thermocline and non-thermocline layer, along with the temperature correction amount T... T The difference is used to calculate the corrected temperature value T at each point. X Then use the corrected temperature value T X The corrected salinity values ​​at each point were calculated using electrical conductivity and pressure.

[0040] Then calculate the absolute value of the difference between the salinity of the ascending profile and the descending profile at each height in each thermocline and non-thermocline layer. Find the smallest absolute value of the difference in each layer in each thermocline and non-thermocline layer. Then find the α and τ coefficient pairs corresponding to the smallest absolute value of the difference, as well as the corresponding temperature correction, correction temperature and correction salinity in each layer.

[0041] Finally, the corrected salinity values ​​within each layer are combined to form the corrected salinity profile data for the entire profile.

[0042] Furthermore, the range of variation for the thermal hysteresis correction coefficients α and τ is defined, and M×N pairs of α and τ coefficients are formed as follows:

[0043] First, set the range of α to be 0.001 to 0.035, with a step size of 0.001;

[0044] Then set the range of τ to 1 to 18, with a step size of 0.5s;

[0045] Finally, using the 35 α and 35 τ obtained according to the step size, 35×35 pairs of α and τ coefficients are formed.

[0046] Compared with the prior art, the beneficial effects of this invention are: the MVP system salinity profile data correction method of this invention significantly reduces the salinity difference between the uplink and downlink measurement profiles of the MVP system, the salinity peaks are basically eliminated, and the improvement is particularly significant at the thermocline. The average absolute salinity difference between the uplink and downlink profiles is reduced from 0.04 psu to 0.014 psu, and the overall salinity error is reduced by 65%. Attached Figure Description

[0047] Figure 1 This is a flowchart of the correction method of the present invention;

[0048] Figure 2 This is a diagram showing the salinity curves of three pre-treated profiles according to the present invention.

[0049] Figure 3 This is a layered diagram of the STA06 group profile data of the present invention;

[0050] Figure 4 This is a layered diagram of the STA70 group profile data of the present invention;

[0051] Figure 5 This is a diagram showing the layering of the STA12 group profile data of the present invention;

[0052] Figure 6 This is a layered diagram of the STA30 group profile data of the present invention;

[0053] Figure 7 This is a graph showing the hysteresis of the relative conductivity of the temperature in the upward profile of the MVP as a function of the rising rate.

[0054] Figure 8 This is a graph showing the hysteresis of the relative conductivity of the MVP down-sectional temperature as a function of the descent rate.

[0055] Figure 9 A comparison chart of salinity with and without response time correction for the MVP downlink profile of this invention;

[0056] Figure 10A comparison of salinity in the uplink and downlink profiles after sensor time response correction and thermal inertia correction for the MVP of this invention;

[0057] Figure 11 This is a map showing the average absolute salinity error of 80 sets of up and down profiles after all the correction processing of this invention has been completed.

[0058] Figure 12 This is a scatter plot of temperature and salinity of the STA49 cross section after pretreatment according to the present invention;

[0059] Figure 13 This is a scatter plot of the temperature and salinity of the STA49 profile after response time correction according to the present invention.

[0060] Figure 14 This is a scatter plot of temperature and salinity of the STA49 profile after thermal inertia correction according to the present invention. Detailed Implementation

[0061] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the embodiments described.

[0062] Example 1:

[0063] like Figure 1 As shown, the MVP system salinity profile data correction method based on temperature gradient profile analysis disclosed in this invention includes the following steps:

[0064] Step 1: Preprocess the MVP profile data to obtain effective and smooth MVP profile data;

[0065] Step 2: Use the optimal sensor response time correction algorithm to correct the response time of the preprocessed MVP profile data;

[0066] Step 3: Divide the temperature profile into various thermoclimate layers and non-thermoclimate layers, and use an enumeration optimization algorithm to find the optimal thermal hysteresis correction coefficient for each layer's thermal inertia correction algorithm. Then, use the optimal thermal hysteresis correction coefficient to calculate the corrected temperature data.

[0067] Step 4: Calculate the salinity using conductivity data, pressure data, and corrected temperature data.

[0068] Furthermore, in step 1, the specific steps for preprocessing the MVP profile data are as follows:

[0069] Step 1.1: Remove MVP profile data within the range of 6-10 dbar after the CTD instrument is activated, because according to the working principle of the CTD instrument, the data between the instrument's activation and stable operation is unavailable.

[0070] Step 1.2: Eliminate reverse pressure data to obtain stable vertical MVP profile data;

[0071] Step 1.3: Remove outliers from the MVP profile data based on the set conductivity threshold, pressure threshold, temperature threshold, and salinity threshold.

[0072] Step 1.4: Use low-pass filtering to remove noisy data from the MVP profile data.

[0073] Furthermore, in step 1.2, the reverse pressure data includes data collected when the instrument moves in the opposite direction during the instrument's descent or ascent (generally due to instrument vibration) and data collected when the instrument's ascent and descent speed is less than the speed threshold.

[0074] Furthermore, in step 1.3, when removing outliers from the MVP profile data, the conductivity threshold is set to 0-6 S / m, the pressure threshold is set to 1.1 times the pressure value of the maximum profile depth, the temperature threshold is set to -2.5 to 40℃, and the salinity threshold is set to 0 to 41 psu. If the conductivity data is not within the range of 0-6 S / m, it is removed; if the pressure data is greater than 1.1 times the maximum profile depth, it is removed; if the temperature data is not within the range of -2.5 to 40℃, it is removed; if the salinity data is not within the range of 0 to 41 psu, it is removed.

[0075] Furthermore, in step 1.4, since the instrument itself is affected by external factors and its own electrical signal disturbances, noise errors will appear in the measurement data. Therefore, in order to ensure the accuracy of the data, when using low-pass filtering to remove noise data in the MVP profile data, since the sampling frequency of the MVP is 25Hz, temperature and conductivity data are filtered using 4-point low-pass filtering, i.e., the cutoff time is 0.16s; pressure and vertical velocity data are filtered using 32-point low-pass filtering, i.e., the cutoff time is 1.28s.

[0076] Furthermore, in step 2, the specific steps for correcting the response time of the preprocessed MVP profile data are as follows:

[0077] Step 2.1: Calculate the first-order temperature difference dT and the first-order conductivity difference dC for each data point in each set of profile data. Then, use the xcorr function in Matlab to calculate the lag correlation coefficient r and lag time difference lags for the first-order temperature difference dT and the first-order conductivity difference dC. In this invention, every 80 points (sampling time 3.2s) are considered as a group. The first-order temperature difference is expressed as dT = T i+1 -T i T i The first-order difference of conductivity is expressed as dC = C, where dC represents the temperature at the i-th data point. i+1 -C i C iThis represents the conductivity at the i-th data point;

[0078] Step 2.2: Find the extreme data point corresponding to the maximum value of the lag correlation coefficient r. Then, establish a binomial fitting equation with the lag correlation coefficient r as the dependent variable and the lag time difference lags as the independent variable. Then, use the lag correlation coefficient r and the lag time difference lags of the extreme data point, the point before the extreme data point, and the point after the extreme data point to fit the binomial fitting equation. The maximum value of the binomial fitting equation is the maximum lag correlation coefficient r. The independent variable corresponding to the maximum lag correlation coefficient r is the corrected lag time difference lags between the temperature data and the conductivity data.

[0079] Step 2.3: Correct the acquisition time of conductivity data according to the correction lag time difference of 1ags, and then interpolate the corrected conductivity data to the same acquisition time as the temperature data to achieve response time correction; or correct the acquisition time of temperature data according to the correction lag time difference of 1ags, and then interpolate the corrected temperature data to the same acquisition time as the conductivity data to achieve response time correction.

[0080] Furthermore, in step 3, the specific steps for dividing the temperature profile into various thermoclimate layers and non-thermoclimate layers are as follows:

[0081] The up-line and down-line profile data in each set of profile data are interpolated and converted according to the set height interval, so that the up-line and down-line profile data are unified to the same height interval.

[0082] The temperature gradient values ​​for the uplink and downlink profiles were calculated separately:

[0083]

[0084] In the formula, G k Z represents the temperature gradient value at the k-th data point. k T represents the measured depth value of the k-th data point. k This represents the measured temperature value of the k-th data point;

[0085] If G k If the temperature is ≤0.05℃ / m, then the kth data point is determined to be within the thermocline. If G k If the value is greater than 0.05℃ / m, then the kth data point is determined to be a point within a non-jump layer.

[0086] Furthermore, after determining the thermocline and non-thermocline at each data point, if the thickness of a thermocline is less than a thickness threshold (preferably set to 10m), the thermocline is merged into the adjacent thermocline and becomes part of the adjacent thermocline.

[0087] Furthermore, in step 3, the specific steps for finding the optimal thermal hysteresis correction coefficient for each thermal inertia correction algorithm in each thermoclimate and non-thermoclimate layer using an enumeration optimization algorithm are as follows:

[0088] Define the range of variation for the thermal hysteresis correction coefficients α and τ, and form M×N pairs of α and τ coefficients;

[0089] For each thermocline and non-thermocline, traverse M×N pairs of α and τ coefficients, and calculate the M×N temperature corrections T for each thermocline and non-thermocline using the following formula. T for:

[0090] T T (n)=-bT T (n-1)+a[T(n)-T(n-1)]

[0091] In the formula, T(n) is the measured temperature value at the nth point, and the formulas for calculating a and b are:

[0092]

[0093]

[0094] In the formula, α and τ are thermal hysteresis correction coefficients, i.e., the amplitude of the error and the time constant, respectively, and f n The sampling frequency;

[0095] Then, using the measured temperature values ​​T(n) at each point in each thermocline and non-thermocline layer, along with the temperature correction amount T... T The difference is used to calculate the corrected temperature value T at each point. X Then use the corrected temperature value T X The corrected salinity values ​​at each point were calculated using electrical conductivity and pressure.

[0096] Then calculate the absolute value of the difference between the salinity of the ascending profile and the descending profile at each height in each thermocline and non-thermocline layer. Find the smallest absolute value of the difference in each layer in each thermocline and non-thermocline layer. Then find the α and τ coefficient pairs corresponding to the smallest absolute value of the difference, as well as the corresponding temperature correction, correction temperature and correction salinity in each layer.

[0097] Finally, the corrected salinity values ​​within each layer are combined to form the corrected salinity profile data for the entire profile.

[0098] Furthermore, the range of variation for the thermal hysteresis correction coefficients α and τ is defined, and M×N pairs of α and τ coefficients are formed as follows:

[0099] First, set the range of α to be 0.001 to 0.035, with a step size of 0.001;

[0100] Then set the range of τ to 1 to 18, with a step size of 0.5s;

[0101] Finally, using the 35 α and 35 τ obtained according to the step size, 35×35 pairs of α and τ coefficients are formed.

[0102] Experimental verification:

[0103] To verify the data correction effect of the MVP system salinity profile data correction method based on temperature gradient profile analysis disclosed in this invention, this invention uses 80 sets of MVP profile data to analyze the correction results.

[0104] I. Data Preprocessing

[0105] Table 1 shows the distribution of the average absolute salinity difference (|ΔS|) of the 80 profiles after preprocessing. It can be seen that after data preprocessing, the average value of |ΔS| for the 80 profiles is 0.04 psu. 34 profiles have a value exceeding 0.04 psu, concentrated between profiles STA01 and STA33; 56 profiles have a value below 0.04 psu, concentrated between profiles STA34 and STA80.

[0106] Figure 2 The display shows the salinity curves of three downlink profiles, STA49, STA58, and STA70, after data preprocessing. The thermocline ranges of the three downlink profiles are 43.6–123.3 dbar, 50.5–111.6 dbar, and 40.9–100.8 dbar, respectively. The black box area represents the approximate range of the thermocline. Obvious salinity spikes, or salinity peaks, can be seen within the thermocline ranges of the three downlink profiles.

[0107] Table 1. Distribution of mean absolute salinity difference between uplink and downlink profiles after pretreatment.

[0108]

[0109]

[0110] The 80 sets of MVP profile data were stratified into four types, as shown in Table 2. One representative profile from each stratification type is presented here. Figure 3-6The diagrams show the stratification of the STA06, STA70, STA12, and STA30 profiles. Based on temperature gradient-depth data, the STA06 profile is divided into 2 layers with depths ranging from 10 to 154.7 dbar and 154.7 to 305 dbar; the STA70 profile is divided into 3 layers with depths ranging from 10 to 41.6 dbar, 41.6 to 92.9 dbar, and 92.9 to 305 dbar; the STA12 profile is divided into 4 layers with depths ranging from 10 to 24.9 dbar, 24.9 to 106.7 dbar, 106.7 to 284.4 dbar, and 284.4 to 305 dbar; and the STA30 profile is divided into 5 layers with depths ranging from 10 to 31.8 dbar, 31.8 to 64.7 dbar, 64.7 to 100.3 dbar, 100.3 to 137.5 dbar, and 137.5 to 305 dbar. Based on the temperature distribution with depth, the stratification results are basically reasonable.

[0111] Table 2. Sectional Stratification

[0112]

[0113] Figure 3-6 The left image of each plot is the TP plot, and the right image is the temperature gradient distribution with depth. The circular scatter plots are the actual temperature gradient scatter plots, and the star-shaped scatter plots are the local medians of all points within 10 units of the black scatter plots. The vertical lines from left to right are the temperature gradient threshold lines of -0.05℃ / m and 0.05℃ / m, respectively, and the horizontal lines are the stratification boundaries.

[0114] II. Perform sensor response time calibration

[0115] Figure 7 The hysteresis of the downlink profile temperature sensor response time relative to the conductivity sensor response time varies with the MVP descent rate. Figure 8 The lag of the temperature sensor response time relative to the conductivity sensor response time in the uplink profile is expressed as a function of the MVP rise rate. Using the time it takes for one MVP data acquisition as a unit (0.04 s), the lag of temperature T relative to conductivity C is closely related to the average probe descent rate over 80 time intervals. A positive lag is defined as temperature T lagging behind conductivity C. Figure 7 and 8The small and medium dots represent lag values ​​determined by a single 80 time interval (data with a maximum lag correlation coefficient less than 0.5 have been removed), while the large dots represent the average within a range of 0.25 dbar / s. The lag time values ​​used in this invention are the average values ​​within this range. Tables 3 and 4 quantitatively provide the values ​​of the lag time between temperature and conductivity in the up and down profiles as a function of the instrument speed. It can be seen that during instrument descent, temperature lags behind conductivity, with the lag time fluctuating around 2 scans; during instrument ascent, temperature precedes conductivity, and the precedence time increases with increasing instrument speed, but the increase gradually decreases and tends to stabilize.

[0116] Table 3 Relationship between hysteresis time and rate of decrease of temperature relative conductivity

[0117]

[0118]

[0119] Table 4. Relationship between hysteresis time and rate of rise of temperature relative conductivity

[0120] Rate of ascent (dbar / s) 2.5 2.75 3 3.25 3.5 3.75 4 4.25 Lag time (scans) -0.299 -0.363 -0.752 -1.416 -1.767 -1.801 -1.838 -2.158

[0121] Figure 9 This display shows a comparison of salinity in downlink profiles with and without sensor response time correction. The solid line represents salinity with sensor response time correction, and the dashed line represents salinity without. It can be observed that the data processed using the optimal sensor response time correction algorithm based on the descent rate and 7-point median filtering is significantly smoother, with salinity peaks largely eliminated. Particularly, the elimination of salinity peaks is most pronounced within the thermocline region.

[0122] Table 5 shows the distribution of the average absolute salinity difference between the uplink and downlink profiles after sensor response time correction. It can be seen that the average value of |ΔS| for the 80 profiles is 0.0192 psu. 32 profiles exceed 0.0192 psu, concentrated between STA01 and STA32; 48 profiles are below 0.0192 psu, concentrated between STA33 and STA80. Compared with the data in Table 5, the number of profiles with |ΔS| greater than 0.07 psu decreased from 4 to 1, leaving only profile STA71, whose |ΔS| decreased from 0.084 psu to 0.077 psu. The number of profiles below the average increased from 46 to 48, with most profiles showing a reduction of approximately 50% in |ΔS|. Overall, after sensor response time correction, the average absolute salinity difference between the uplink and downlink profiles... It decreased from 0.04 psu to 0.0192 psu, a reduction of 52%.

[0123] Table 5. Distribution of mean absolute salinity difference between uplink and downlink profiles after sensor response time correction.

[0124] Mean absolute salinity difference |ΔS| (psu) Section number number |ΔS|>0.07 STA71 1 0.05 ≥ |ΔS| > 0.04 STA01 1 0.04 ≥ |ΔS| > 0.03 STA05, STA06, STA65 3 0.03 ≥ |ΔS| > 0.0192 STA02-STA04, STA07-STA10, STA12-STA25, STA27-STA32 27 0.0192 ≥ |ΔS| > 0.01 STA11, STA26, STA33-STA64, STA66-STA70, STA72-STA80 48

[0125] III. Perform thermal inertia correction

[0126] Figure 10 The study presents a comparison of salinity in the ascending and descending profiles with and without thermal inertia correction. It can be observed that the effect of thermal hysteresis correction is not significant near the mixing layer. Below the mixing layer, near and below the thermocline, the effect of thermal hysteresis correction is more pronounced, mainly manifested in a reduction in thermocline strength. The salinity difference between the upper and lower profiles decreases slightly, and the minor salinity spikes in the overall profile are also well corrected.

[0127] Table 6 shows the distribution of the average absolute salinity difference between the upper and lower profiles after thermal inertia correction for the entire layer. It can be seen that the average value of |ΔS| for the 80 profile groups is 0.015 psu. 34 profile groups have a value exceeding 0.015 psu, concentrated between STA01 and STA32; 46 profile groups have a value below 0.015 psu, concentrated between STA33 and STA80. Among these, 23 profile groups have a value less than 0.01 psu, accounting for 28.75% of the total number of profiles.

[0128] Table 7 shows the distribution of the average absolute salinity difference between the upstream and downstream profiles after stratified thermal inertia correction. It can be seen that the average value of |ΔS| for the 80 profile groups is 0.014 psu. 34 profile groups have a value exceeding 0.014 psu, concentrated between STA01 and STA32; 46 profile groups have a value below 0.014 psu, concentrated between STA33 and STA80. Among these, 30 profile groups have a value less than 0.01 psu, accounting for 37.5% of the total. Compared with the data in Table 6, the number of profile groups with a value less than 0.01 psu increased from 23 to 30, and the value of |ΔS| decreased for all profiles, with the difference between the upstream and downstream profiles... The |ΔS| decreased from 0.015 psu to 0.014 psu, a reduction of approximately 6.7%, indirectly verifying that layered correction of the profile is more reasonable than full-layer correction. Compared with the data in Table 5, profiles with |ΔS| exceeding 0.03 psu have disappeared. The STA71 profile, where the |ΔS| reduction was not significant after sensor response time correction, saw the largest reduction after thermal inertia correction, decreasing from 0.077 psu to 0.008 psu. The reductions in |ΔS| for the remaining profiles were mostly around 26%. Overall, after thermal inertia correction, the |ΔS| of the uplink and downlink profiles... It further decreased from 0.0192 psu to 0.014 psu, a decrease of 26.7%.

[0129] Table 6 shows the distribution of the average absolute salinity difference between the up and down profiles after thermal inertia correction for the entire floor.

[0130]

[0131] Table 7 shows the distribution of the average absolute salinity difference between the up and down profiles after segmented thermal inertia correction.

[0132]

[0133] IV. Salinity Profile Correction Effect

[0134] Figure 11 The average absolute salinity error of 80 uplink and downlink profiles after all corrections are shown. It can be observed that the average absolute salinity difference between the uplink and downlink profiles is significantly reduced after correction. Combining Tables 1, 5, and 7, the |ΔS| of the uplink and downlink profiles shows the largest reduction after sensor response time correction. The value decreased from 0.04 psu to 0.0192 psu, a reduction of 52%. After thermal inertia correction, |ΔS| in both the up and down profiles further decreased. It further decreased from 0.0192 psu to 0.014 psu, a decrease of 26.7%.

[0135] Figure 12-14 The temperature-salinity scatter plots on the STA49 profile are shown after preprocessing, response time correction, and thermal inertia correction, respectively. Solid dots represent downlink data, and hollow dots represent uplink data. Combined with... Figures 9-10 It can be seen that the salinity difference is mainly concentrated at the thermocline. After sensor response time correction, the salinity difference between the upper and lower profiles is significantly reduced, most notably within the thermocline range, manifested in the near disappearance of salinity peaks and a shift from a divergent to a concentrated distribution of temperature-salinity scattering points. After thermal inertia correction, the salinity difference between the upper and lower profiles is further reduced, but the rate of reduction is less. Some minor salinity peaks are further eliminated, the thermocline strength decreases, and salinity values ​​decrease at depths below the thermocline.

[0136] In summary, after all corrections were completed, the salinity difference between the uplink and downlink profiles was significantly reduced, becoming more concentrated within the thermocline. Minor salinity spikes were effectively eliminated, and the distribution of temperature-salinity scattering points shifted from divergent to concentrated. It decreased from 0.04 psu to 0.014 psu, a reduction of 65%.

[0137] IV. Conclusion

[0138] This invention addresses the limitation of existing thermal inertia correction methods that still exhibit salinity spikes at strong thermoclines. Combining salinity processing methods from pump-type CTD, UCTD, and gliders, and using 80 sets of MVP data profiles measured during a voyage in the western Pacific as experimental subjects, this invention proposes a salinity profile data correction method for the MVP system based on temperature gradient profile analysis. This method corrects the temperature, and consequently, the salinity profile. Results show that the correction method significantly reduces the salinity difference between the uplink and downlink profiles of the MVP measurement, essentially eliminating salinity spikes, particularly showing significant improvement at the thermocline. The average absolute salinity difference between the uplink and downlink profiles decreased from 0.04 psu to 0.014 psu, and the overall salinity error was reduced by 65%, verifying the rationality of the proposed MVP salinity profile correction method.

[0139] As described above, although the invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the invention itself. Various changes in form and detail may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method for correcting salinity profile data of an MVP system based on temperature gradient profile analysis, characterized in that, Includes the following steps: Step 1: Preprocess the MVP profile data to obtain effective and smooth MVP profile data; Step 2: Use the optimal sensor response time correction algorithm to correct the response time of the preprocessed MVP profile data; Step 3: Divide the temperature profile into various thermoclimate layers and non-thermoclimate layers, and use an enumeration optimization algorithm to find the optimal thermal hysteresis correction coefficient for each layer's thermal inertia correction algorithm. Then, use the optimal thermal hysteresis correction coefficient to calculate the corrected temperature data. Step 4: Calculate the salinity using conductivity data, pressure data, and corrected temperature data; In step 3, the specific steps for finding the optimal thermal hysteresis correction coefficient for each thermal inertia correction algorithm in each thermoclimate and non-thermoclimate layer using an enumeration optimization algorithm are as follows: Define the range of variation for the thermal hysteresis correction coefficients α and τ, and form M×N pairs of α and τ coefficients; For each thermocline and non-thermocline, traverse M×N pairs of α and τ coefficients, and calculate the M×N temperature corrections T for each thermocline and non-thermocline using the following formula. T for: T T (n)=-bT T (n-1)+a[T(n)-T(n-1)] In the formula, T(n) is the measured temperature value at the nth point, and the formulas for calculating a and b are: In the formula, α and τ are thermal hysteresis correction coefficients, i.e., the amplitude of the error and the time constant, respectively, and f n The sampling frequency; Then, using the measured temperature values ​​T(n) at each point in each thermocline and non-thermocline layer, along with the temperature correction amount T... T The difference is used to calculate the corrected temperature value T at each point. X Then use the corrected temperature value T X The corrected salinity values ​​at each point were calculated using electrical conductivity and pressure. Then calculate the absolute value of the difference between the salinity of the ascending profile and the descending profile at each height in each thermocline and non-thermocline layer. Find the smallest absolute value of the difference in each layer in each thermocline and non-thermocline layer. Then find the α and τ coefficient pairs corresponding to the smallest absolute value of the difference, as well as the corresponding temperature correction, correction temperature and correction salinity in each layer. Finally, the corrected salinity values ​​within each layer are combined to form the corrected salinity profile data for the entire profile.

2. The method for correcting salinity profile data of the MVP system based on temperature gradient profile analysis according to claim 1, characterized in that, In step 1, the specific steps for preprocessing the MVP profile data are as follows: Step 1.1: Remove MVP profile data within the range of 6-10 dBar after the CTD instrument is started and activated; Step 1.2: Eliminate reverse pressure data to obtain stable vertical MVP profile data; Step 1.3: Remove outliers from the MVP profile data based on the set conductivity threshold, pressure threshold, temperature threshold, and salinity threshold. Step 1.4: Use low-pass filtering to remove noisy data from the MVP profile data.

3. The method for correcting salinity profile data of the MVP system based on temperature gradient profile analysis according to claim 2, characterized in that, In step 1.2, the reverse pressure data includes data collected when the instrument moves in the opposite direction during the instrument's descent or ascent, as well as data collected when the instrument's ascent and descent speed is less than the speed threshold.

4. The method for correcting salinity profile data of the MVP system based on temperature gradient profile analysis according to claim 2, characterized in that, In step 1.3, when removing outliers from the MVP profile data, the conductivity threshold is set to 0-6 S / m, the pressure threshold is set to 1.1 times the pressure value of the maximum profile depth, the temperature threshold is set to -2.5 to 40℃, and the salinity threshold is set to 0 to 41 psu. If the conductivity data is not within the range of 0-6 S / m, it is removed; if the pressure data is greater than 1.1 times the maximum profile depth, it is removed; if the temperature data is not within the range of -2.5 to 40℃, it is removed; if the salinity data is not within the range of 0 to 41 psu, it is removed.

5. The method for correcting salinity profile data of the MVP system based on temperature gradient profile analysis according to claim 2, characterized in that, In step 1.4, when removing noise data from the MVP profile data using low-pass filtering, temperature and conductivity data are filtered using a 4-point low-pass filter with a cutoff time of 0.16s; pressure and vertical velocity data are filtered using a 32-point low-pass filter with a cutoff time of 1.28s.

6. The method for correcting salinity profile data of the MVP system based on temperature gradient profile analysis according to claim 1, characterized in that, In step 2, the specific steps for correcting the response time of the preprocessed MVP profile data are as follows: Step 2.1: Calculate the first-order temperature difference dT and the first-order conductivity difference dC for each data point in each set of profile data. Then, use the xcorr function in Matlab to calculate the lag correlation coefficient r and lag time difference lags of the first-order temperature difference dT and the first-order conductivity difference dC. Step 2.2: Find the extreme data point corresponding to the maximum value of the lag correlation coefficient r. Then, establish a binomial fitting equation with the lag correlation coefficient r as the dependent variable and the lag time difference lags as the independent variable. Then, use the lag correlation coefficient r and the lag time difference lags of the extreme data point, the point before the extreme data point, and the point after the extreme data point to fit the binomial fitting equation. The maximum value of the binomial fitting equation is the maximum lag correlation coefficient r. The independent variable corresponding to the maximum lag correlation coefficient r is the corrected lag time difference lags between the temperature data and the conductivity data. Step 2.3: Correct the acquisition time of conductivity data according to the correction lag time difference of 1ags, and then interpolate the corrected conductivity data to the same acquisition time as the temperature data to achieve response time correction; or correct the acquisition time of temperature data according to the correction lag time difference of 1ags, and then interpolate the corrected temperature data to the same acquisition time as the conductivity data to achieve response time correction.

7. The method for correcting salinity profile data of the MVP system based on temperature gradient profile analysis according to claim 1, characterized in that, In step 3, the specific steps for dividing the temperature profile into various thermoclimate layers and non-thermoclimate layers are as follows: The up-line and down-line profile data in each set of profile data are interpolated and converted according to the set height interval, so that the up-line and down-line profile data are unified to the same height interval. The temperature gradient values ​​for the uplink and downlink profiles were calculated separately: In the formula, G k Z represents the temperature gradient value at the k-th data point. k T represents the measured depth value of the k-th data point. k This represents the measured temperature value of the k-th data point; If G k If the temperature is ≤0.05℃ / m, then the kth data point is determined to be within the thermocline. If G k If the value is greater than 0.05℃ / m, then the kth data point is determined to be a point within a non-jump layer.

8. The method for correcting salinity profile data of the MVP system based on temperature gradient profile analysis according to claim 1, characterized in that, After determining the thermocline and non-thermocline at each data point, if the thickness of a thermocline is less than the thickness threshold (10m), the thermocline is merged into the adjacent thermocline and becomes part of the adjacent thermocline.

9. The method for correcting salinity profile data of the MVP system based on temperature gradient profile analysis according to claim 1. Its features are, The range of variation for the thermal hysteresis correction factors α and τ is set, and M×N pairs of α and τ coefficients are formed as follows: First, set the range of α to be 0.001 to 0.035, with a step size of 0.001; Then set the range of τ to 1 to 18, with a step size of 0.5s; Finally, using the 35 α and 35 τ obtained according to the step size, 35×35 pairs of α and τ coefficients are formed.

Images