• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Data Quality Control Method of a New Drifting Observation Technology Named Drifting Air-Sea Interface Buoy

    2024-03-12 11:14:52LIShuoWANGBinDENGZenganWUBaoqinZHUXiandeandCHENZhaohui
    Journal of Ocean University of China 2024年1期

    LI Shuo , WANG Bin , , DENG Zeng’an , WU Baoqin, ZHU Xiande,and CHEN Zhaohui

    1) National Ocean Technology Center, Tianjin 300112, China

    2) School of Marine Science and Technology, Tianjin University, Tianjin 300072, China

    3) Key Laboratory of Ocean Observation Technology, Ministry of Natural Resources, Tianjin 300112, China

    4) People’s Liberation Army 92859, Tianjin 300061, China

    5) Physical Oceanography Laboratory, Ocean University of China, Qingdao 266100, China

    Abstract An integral quality control (QC) procedure that integrates various QC methods and considers the design indexes and operational status of the instruments for the observations of drifting air-sea interface buoy was developed in the order of basic inspection followed by targeted QC. The innovative method of combining a moving Hampel filter and local anomaly detection complies with statistical laws and physical processes, which guarantees the QC performance of meteorological variables. Two sets of observation data were used to verify the applicability and effectiveness of the QC procedure, and the effect was evaluated using the observations of the Kuroshio Extension Observatory buoy as the reference. The results showed that the outliers in the time series can be correctly identified and processed, and the quality of data improved significantly. The linear correlation between the quality- controlled observations and the reference increased, and the difference decreased. The correlation coefficient of wind speed before and after QC increased from 0.77 to 0.82, and the maximum absolute error decreased by approximately 2.8 m s?1. In addition, air pressure and relative humidity were optimized by 10?3 – 10?2 orders of magnitude. For the sea surface temperature, the weight of coefficients of the continuity test algorithm was optimized based on the sea area of data acquisition, which effectively expanded the applicability of the algorithm.

    Key words drifting air-sea interface buoy; quality control; oceanic variables; meteorological variables; continuity test

    1 Introduction

    The drifting air-sea interface buoy (DrIB), a small-sized,lightweight, and cost-effective mini-buoy, is mainly used to measure basic variables at the air-sea interface by expendable and drifting methods in real time. The large-scale application of the DrIB can realize the high-frequency acquisition of key meteorological (wind speed, wind direction, and air pressure at approximately 3 m and air temperature and relative humidity at approximately 2.8 m above the sea surface) and oceanic (sea surface temperature (SST)and wave parameters) variables in the global ocean (Centurioniet al., 2019), which is beneficial to understanding ocean dynamics, air-sea interaction, and global climate change and forecasting of modal accuracy.

    The external structure of the DrIB consists of two parts:the above-water and underwater units (Fig.1). The above water unit includes a hemispherical float and a mast, and the underwater unit consists of two mutually perpendicular water sails and an underwater counterweight. The DrIB is composed of a meteorological observation module, a hydrographic observation module, a measurement and control system, a communication module, and a power supply module. After the DrIB is deployed, it carries out observations of key meteorological and hydrological variables at the air-sea interface by drifting with the currents.After preliminary online calibration, the observations are transmitted to the shore/ship-based data management system, which can receive, process, store, and distribute observations in real time through the satellite communication system. The DrIB has a sampling frequency of 1 Hz,and the output data are the average of 120 samples with optional sampling intervals of 10 min / 30 min / 1 h. Table 1 shows the observational performance of the DrIB and the instruments installed on it.

    Table 1 Observation indexes and associated instruments of the DrIB

    Fig.1 (a) Dimensional view of the drifting air-sea interface buoy and (b) the site of drifting air-sea interface buoy observation in the Western Pacific. The abbreviations AT and RH in (a) represent air temperature and relative humidity,respectively.

    Marine science is an observation-based discipline (Wuet al., 2020). Observations are the critical foundation for scientific research, which is directly affected by data quality. However, operation errors, marine environment, unstable communication transmission,etc., can interfere with the observation results and cause data anomalies, and most of the observations cannot be applied directly. Therefore,rigorous and effective quality control (QC) of the received data and removal of the error and abnormal values hidden in the time series are necessary to improve the accuracy and reliability of observations.

    For the QC research of marine environmental observations, many institutions and experts have shared their valuable experiences. Foreign research has been carried out earlier, and after continuous improvement and development,QC technology is more mature nowadays. For example,the drifting buoy of the Surface Velocity Program (SVP)can transmit back its position and SST data, and the SVP data collection center conducts QC based on whether the value violates the set continuity criterion (Hansen and Herman, 1989; Hansen and Poulain, 1996). The National Data Buoy Center of America manages 50 ocean stations and more than 100 moored buoys. It has developed an automated QC program to perform transmission error checks,range inspections, time continuity checks, correlation checks,etc.(Jarraud, 2008; National Data Buoy Center, 2009). The Kuroshio Extension Observatory (KEO) buoy was first deployed in 2004. Currently, the time series and statistical properties of hydrometeorological variables, including wind,air temperature, air pressure, relative humidity, precipitation, long- and short-wave radiation, seawater temperature,salinity,etc., can be independently checked and identified for data quality by a complete QC program (Andersonet al., 2021). In addition, the glider toolbox based on Matlab is specially used to correct the thermal lag bias of the salinity observed by underwater gliders (Troupinet al.,2015).

    Comparatively, China’s research is relatively late. Verified by the observations of weather stations and ocean observatories, real-time meteorological and hydrological data can be quality-controlled by determining the range threshold and physical relationships between elements (Li and Li, 1997). The QC software developed by the National Marine Data and Information Service realizes the automatic QC of the data collected by ocean observatories (Yanget al.,2017); Wuet al.(2019)used a laboratory high-precision salinometer to compare and evaluate the observations of the ship-borne Conductivity, Temperature, Depth (CTD)instrument for complete QC. Liuet al.(2020)proposed a real-time QC procedure including thermal lag correction and 12 tests based on the temperature-salt profile obtained by the domestic Sea Wing underwater glider. Renet al.(2021) innovatively integrated the spike inspection and PauTa criterion, effectively eliminating the abnormal values in sea temperature and salinity observations.

    In summary, extensive research has been carried out on QC technologies of weather and ocean observatories, fixed buoy observations, and data from underwater sensors on mobile apparatus. However, studies on simultaneous observations of key meteorological and hydrological variables at the air-sea interface through drifting ways are relatively scarce. Moreover, with the rapid development of the DrIB observation network, observations have accumulated,and QC research has become urgent. In addition, drifting or mobile apparatuses, such as wave gliders and unmanned ships, are currently emerging, representing a significant trend in the development of marine observation technologies (Grareet al., 2021). However, no systematic data QC method has been formed yet, and the reliability of observations is uncertain. The research on the QC of DrIB observations and related theories and methods also applies to other mobile observation equipment, which can effectively compensate for shortcomings and solve the problem of poor quality of original data.

    With a focus on the air-sea interface scenarios, an integral QC procedure for DrIB observations was established in conjunction with the instrument’s design indexes and operational records. Basic inspection was performed first,and the observed variables were categorized for targeted QC. Two sets of observations from the Northwest Pacific were used as examples to analyze the effect of QC and verify the applicability, effectiveness, and advancement of QC methods and the related procedure.

    2 QC Procedure

    QC is mainly used for two types of errors in long-term continuous hydrometeorological observations. The gross error exceeds the observation capability of the equipment or normal variation range of the observed variables and is intuitive and easy to identify and eliminate; meanwhile,the outlier is within a reasonable range but deviates from the overall or local trend and thus needs to be screened by a variety of means (Tanet al., 2022).

    This paper applied various QC methods, such as land inspection, repetition check, incremental inspection, range inspection, and continuity inspection. Applicable continuity tests were used to carry out targeted QC on location data and meteorological and oceanic variables. The abnormal data were then eliminated or linearly interpolated based on the nearest samples. Each variable was considered as an independent one-dimensional time series. Fig.2 shows the specific steps of the QC procedure.

    Fig.2 Schematic chart of the QC procedure.

    2.1 Basic QC

    Previously, QC of observations tended to detect anomalies only from the data themselves, which might have affected efficiency. For DrIB, the observed variables are relatively abundant, and the operation of the buoy and changes in the air-sea environment are recorded well. QC should consider the instrument’s design indexes and operating attitude and maximize the advantages of operation records to identify abnormal data quickly and accurately. The basic QC mainly includes the following steps.

    1) Land inspection. Unlike mooring buoy, the DrIB may start operating before entering the sea due to land or shipbased functional test and other reasons, sending land information and directly causing abnormal SST observations.Such errors usually only exist at the initial stage of a continuous-time series. Therefore, land inspection needs to be performed first to eliminate invalid data when the DrIB is not in the sea and ensure that the observations are of the real ocean environment. By assessing whether the swing angle conforms to its in-water motion and whether a fault difference exists between the initial and stable periods of SST to make an accurate judgment of the working environment, the observations on land are removed, and a separate file of land observations is generated to facilitate subsequent inspection and verification.

    2) Repetition check. In cases of unstable satellite transmission and failed data storage modules, more than one piece of data may be stored in the land-based data receiving station simultaneously, which results in duplicate errors. The repetition check is applied to eliminate duplications and ensure a ‘one-to-one’ relationship between time and observation. A file of duplications is generated at the same time.

    3) Time increment test. Observations should be arranged in chronological order. ‘Time reversal’ is prevented by performing an incremental time test on the time records to ensure correct timing. The data detected with chaotic time are detected. If necessary, the order is adjusted, and deletion and adjustment actions are recorded in real time in the timing-related log.

    4) Range inspection. Range inspection is a practical judgment of whether values are reasonable based on the basic cognition of geography, general laws of hydrometeorological variables, and the observation capability of the equipment itself. For example, the SST generally varies within ?4℃ to 44℃ (Fenget al., 1999), and the maximum relative humidity does not exceed 100%. The design observation indexes of DrIB are shown in Table 1, and the data beyond the range are considered as set W.

    2.2 Targeted QC

    After basic QC, evident errors in the time series are eliminated. The working characteristic of ‘drifting with the tide’ determines the continuity and regularity of the observed variables. Next, the observations are divided into three categories: location data, oceanic variables, and meteorological variables. Then, the corresponding continuity criterion is set to identify the values that violate the continuing trend of the series for further targeted QC.

    1) Location data. For the sudden abnormal change (spike)in the drifting trajectory of DrIB, the spike test is used to determine whether the latitude and longitude at the spike are correct (Gouretski, 2018), which is determined by the following:

    whereln?1,ln, andln+1represent three consecutive longitude or latitude values, andαis a critical parameter, which is considered as 0.1 when the data sampling interval is 1 h.If the formula holds, thenlnis abnormal, and the abnormal values screened are set to G.

    2) Oceanic variables. The SVP data collection center integrates the spike and continuity tests and sets a weighted averagemiof iterative changes as the SST ‘spike’ detector to check the continuity of the SST time series (Hansen and Poulain, 1996). The determination ofmiis based on the recent changes in SST experienced by the buoy, and the nearest samples of the detected data are given the highest weight. The test is as follows: the first valid SST ism1, and the observations are tested sequentially by the following:

    The SST is anomalous if the formula cannot be formed andmiis not updated. The SVP data collection letsCbe 1/4 and ? be 0.5 K and applies the method to the detection of SST data in tropical Pacific Ocean.

    Compared with the relatively balanced and stable environment in tropical oceans, mid-latitude oceans exhibit more complex dynamics and a stronger air-sea nonlinear instability interaction. Thus, the SST observations are more dependent on the actual change processes of seawater at the time of acquisition. Moreover, for the improved applicability of the method, the contribution weight ofin situobservations should be appropriately increased in the calculation of the SST ‘spike’ detectormifor DrIB in mid-latitude oceans (Yuet al., 2010). In other words, the value of parameterCshould be increased. After experimental verification, the detection performance is the best when it is raised to 1/3. Correspondingly, ? is set to 0.5℃.The QC results may be incorrect when the DrIB crosses a strong front or when the sampling interval is large. Thus,the data are tested in both forward and reverse directions.The values identified as anomalous in both directions are taken as set S.

    3) Meteorological variables. Compared with the ocean,the atmosphere is more active and fast-changing, and the changing patterns of key meteorological variables, such as temperature, air pressure, wind speed, and relative humidity, are more significant. When the QC method mentioned above is used for oceanic variables to test meteorological variables, a growing gap exists between the observations and the detectormi, which results in most normal data being judged as abnormal and prevented from being directly applied.

    According to statistical theory, meteorological variables observed continuously at fixed positions tend to obey a probability distribution; therefore, statistical tests, such as the PauTa criterion (3δ criterion), which considers data with a residual error of more than three times the standard deviation from the arithmetic mean as outliers, can be used to check data (Luet al., 2019). However, if the arithmetic mean and standard deviation are used as judgment criteria,the extreme outliers are prone to biased results. Hampel filter is a filtering algorithm for identifying anomalies in a time series, and its operating principle is similar to that of the PauTa criterion, except that the criteria are the median and converted median absolute deviation, which can effectively avoid the interference of extreme values (Gholamiet al., 2022). In addition, as the spatial location of DrIB is dynamically changing, and the atmospheric environment varies significantly in different seas, the study of the statistical characteristics of its long-term time series has little significance. Based on the continuity and regularity of temporal and spatial changes of drifting observation, a moving Hampel filter has been proposed to assess the degree of data departure from daily variability using days as the time window unit to find outliers in the time series of meteorological variables observed by DrIB (Liu, 2017; Renet al., 2021). Suppose a time series of observations,q1,q2, ···,qn, with a data acquisition interval of 1 h. The time window for sampling is one day; that is,q1,q2, ···,q25are taken as a group, and the median isqme. Then, the following are obtained:

    wheredmstands for the local conversion ofMAD, and theqithat is more than three times ofdmfromqmeis abnormal.After the detection of the first group, the time window moves one piece of information in sequence and conducts detection again until the last group of the time series has been identified. The detected anomalies are set as Q1.

    The data in set Q1 are anomalous in terms of mathematical statistics. If the atmospheric environment is relatively stable, the fluctuation of variables is balanced, the local conversionMADis small, and correct data can be easily misclassified as abnormal. Local anomaly detection is further applied to avoid the risk of misjudgment by the moving Hampel filter. Theqsof any observations in the set Q1 is correct if the difference between it and the observation of any adjacent moment does not exceed the set maximum change threshold. The data misjudged as abnormal are considered set Q2. The change threshold should be larger than the design observation error of the DrIB to prevent the instrument’s error from interfering with the judgment.

    Given the rapid and instantaneous change, poor data continuity, and complex influencing factors of wind direction,only basic QC is performed. For the abnormal location data set Wg + G, abnormal oceanic observations set Ws + S,and abnormal meteorological observations (except wind direction) set Wq + Q1 ? Q2, for the retention of data continuity based on not breaking the statistical characteristics and changing trends of the time series, the results of linear interpolation of the nearest observations are used to replace one or two separate abnormal data. Wind direction anomalies and other variables with more than two consecutive anomalies are deleted.

    In summary, the QC procedure of DrIB observations developed in this study focuses on data classification for QC based on the full use of measurements. For oceanic variables, by increasing the weight ofin situobservations in the SST continuity test algorithm, the applicability of the algorithm is extended from tropical waters to higher latitudes, and the accuracy of outlier detection is improved.For meteorological variables, a QC method combining the moving Hampel filter and local anomaly detection is proposed. The median andMADare chosen as determination parameters to avoid interference by extreme values when applying the PauTa criterion. Moving detection performed in one-day steps is more consistent with the working characteristics of DrIB rather than using long-term statistical characteristics of fixed observations. The local anomaly detection method considers the actual physical process of meteorological variables, which effectively reduces the probability of misjudgment in identifying anomalies based on mathematical features only.

    3 Application and Validation of QC Procedure

    In this study, two sets of DrIB observations, namely,5840 and 8990, were used as examples to systematically analyze the QC performance and verify the methods’ effectiveness and advancement. The results showed that the QC scheme is applied well in the processing of DrIB observations.

    3.1 Data Description

    DrIB 5840 continued its drift and observation in the Kuroshio Extension from August 30, 2018, to January 1,2019. The sampling interval was 1 h. A total of 2909 records, including longitude, latitude, SST, air temperature,air pressure, wind speed, wind direction, relative humidity,and buoy operation records, were obtained. DrIB 8990 continued to drift and observe in eastern Japan from November 4, 2019, to November 6, 2019. The sampling interval was 10 min, and 251 pieces of data were collected with the same categories as DrIB 5840.

    3.2 Analysis of QC Results

    Table 2 shows the QC results of DrIB 5840 observations. The roll and pitch angles of the first six observation records were all zero, which was inconsistent with the swaying attitude of the buoy in seas. Fig.3(b) shows a steep descent at the beginning of the SST curve, and the significant difference in SST was more than 3℃. Therefore,these data were judged to be sent by the DrIB on land by mistake. Usually, DrIB functions well, and the observations are reliable and do not exceed the normal range. Thus, the range check focuses on the missing data stored in ?999.999 format. Figs.3 and 4(a) show the time series of the ob-served variables and drifting trajectory plotted using the original observations, respectively. The abrupt straight lines on the figures are precisely due to the missing data,that is, the anomalies detected by the range inspection, in which all observed variables were missed twice on December 18, 2018.

    Table 2 QC results of observations from DrIB 5840

    Fig.3 Time series of DrIB 5840 observations (before QC).

    After the basic check, the DrIB 5840 observations were classified for targeted QC. After the spike test, two abnormal data were screened out from the longitude and latitude time series, and the two spikes near 32?N, 144?E on the drifting trajectory were removed. The drifting trajectory after QC (Fig.4(b)) was clear compared with the previous observation, and the curve was naturally connected without abnormal spikes and easily visualized the dynamic changes of the observed sea areas. In addition, the QC performance was evident and intuitive. The SST observations were of high quality with a few anomalies, and seven records, which included both true anomalies and misjudgment results, were detected by the method proposed by the SVP data collection center (Fig.5(a)). According to the local diagram of SST QC results (Fig.5(b)), three anomalies from 2:00 to 4:00 on September 7, 2018 were consistent with the overall and local developing trends of the time series, which were misjudgment results. After adjusting the coefficients weights in the algorithm, three anomalies were detected and manually checked to confirm that the results were correct. The findings proved that the improved method effectively improved the accuracy of discrimination while expanding the applicability. For meteorological variables, with air pressure used as an example,the air pressure QC results were obtained by checking the SST (Fig.6(a)). The outliers were numerous and primarily distributed in patches, and all the observations from November 22 to 24, 2018 were abnormal, which was inconsistent with the facts. A total of 73 anomalies violating the daily variability of air pressure were detected using the moving Hampel filter, and the percentage of anomalies was 2.5%. By comparison, seven anomalies exceeding 10 hPa of air pressure at adjacent moments were detected from these data after the addition of local anomaly detection (Fig.6(b)). The percentage of anomalies was reduced to 0.24%, which indicates that as the physical processes significantly reduced the probability of misjudgment, the effect of QC was more guaranteed.

    Fig.4 Drifting trajectory of DrIB 5840.

    Fig.5 QC results of SST observed by DrIB 5840.

    Fig.6 QC results of air pressure observed by DrIB 5840.

    Fig.7 shows the DrIB 5840 observations after QC. Abrupt values and outliers out of the developing trends were identified and deleted or replaced by interpolation. The continuity and regularity of the time series were intact, the characteristics of the developing trends were evident, the data quality was significantly improved, and the data can be used for subsequent applications. The DrIB 8990, which had a short observation time, fewer observation data, and better quality, showed no anomaly in the detection by the basic QC. All anomalies in Table 3 came from the targeted QC. Given that the observation interval of DrIB 8990 was 10 min, the criteria of the continuity test for variables were adjusted adaptively. Fig.8 shows the comparison of the time series before and after QC, in which abnormal values were detected from air pressure, wind speed, SST, and relative humidity observations, and the sharp points on the curve were eliminated after QC.

    Table 3 QC results of observations from DrIB 8990

    Fig.7 Time series of DrIB 5840 observations (after QC).

    Fig.8 QC results of DrIB 8990 observations.

    3.3 Validation of QC Effect

    To better assess the effect of QC, we compared and evaluated the observations of DrIB 8990 before and after QC using the internationally recognized KEO buoy observations as the evaluation standard. The observation of DrIB 8990 covered the area of Kuroshio Extension and was adjacent to the KEO buoy (Fig. 9(a)). The KEO buoy was deployed in the southern area of the Kuroshio Extension, and the nominal location was 32.3?N, 144.6?E. Its observations were widely used in the research related to the Kuroshio Extension.

    Given the drifting characteristics of DrIB, spatiotemporal matching of the evaluated data was necessary. Fig.9(b)shows the distance variation between the DrIB 8990 and the KEO buoy with time. The paired data were the observations obtained at the same time within 10 km of the two buoys, that is, 164 records from 18:50 on November 4,2019, to 22:00 on November 5, 2019. A 10 km distance was considered because the KEO buoy observes in a circular scope with a diameter of more than 10 km, and the air-sea environment within its 10 km can be considered relatively consistent (Zhanget al., 2016; Andersonet al.,2020). As the KEO buoy did not provide 10 min observations of SST for the selected periods, the comparison of variables only included air pressure, wind speed, and relative humidity. In addition, DrIB and KEO buoys observe meteorological variables at different heights. The former observes relative humidity at approximately 2.8 m and air pressure and wind at approximately 3 m, and the latter pro-vides wind at a nominal height of 4 m and air pressure and relative humidity at a nominal height of 2.5 m. Therefore,the data must be unified to the same height first. By applying the COARE 3.0 formula and inputting variables, including SST, air pressure, air temperature, relative humidity, wind speed, and the corresponding observation heights,the meteorological field at the height of 10 m above the sea surface can be outputted, thus unifying the heights of the two sets of meteorological variables (Fairallet al.,2003). Given the lack of 10 min SST of the KEO buoy, the quality-controlled SST of DrIB 8990 was used to perform the height transformation of the KEO buoy observations instead.

    Fig.9 Position (a) and distance (b) of DrIB 8990 relative to the KEO buoy.

    Evaluation of the effect of QC mainly relied on two indicators, namely, data correlation and data difference, and the statistics used included correlation coefficientr, mean absolute error (AE), maximum absolute error (MAE), and root mean square error (RMSE), which are defined as follows.

    wherenis the number of paired data,biis the DrIB observation,kiis the KEO buoy observation,represents the arithmetic mean of the DrIB observations, anddenotes the arithmetic mean of the KEO buoy observations.The comparison results are shown in Fig.10 and Table 4.The solid curves in Fig.10 correspond to the left ordinate,and dashed and dotted curves refer to the right ordinate.

    Table 4 Comparison between the observations of DrIB 8990 and KEO buoy

    Fig.10 QC effect of DrIB 8990 observations.

    The results showed that the paired air pressure was the most consistent among the three variables. The correlation coefficient can intuitively reflect the degree of linear correlation between the two air pressures. The correlation coefficient between the original air pressure of DrIB 8990 and the KEO buoy was over 0.98, which was highly correlated at the 95% confidence level, and the changes in the two curves were almost the same. All three difference statistics were less than 0.5 hPa, which shows that the quality of DrIB observations reached the advanced international level. The quality of air pressure can only be slightly improved due to the high correlation and presence of a few anomalies. Ther, AE, andRMSEafter QC were optimized by about 10?3orders of magnitude, andMAEwas reduced by about 0.1 hPa. Contrary to the air pressure, the correlation coefficient between the relative humidity of DrIB 8990 and the KEO buoy was the smallest (approximately 0.4), and the difference statistics also showed weak data conformity. The relative humidity observed by the DrIB 8990 was approximately 3% higher than that of the KEO buoy on average, and the maximum exceeded 9%. Fig. 10(c)shows a few anomalies of relative humidity data detected by the QC procedure, and the data quality improved overall.The correlation coefficient increased, the deviation decreased, and the optimizations were mostly between 10?3– 10?2orders of magnitude. The time series of wind speed fluctuated strongly, but the developing trends of the two were generally consistent. Before QC, ther, AE, andRMSEwere approximately 0.77, 1 m s?1, and 1.2 m s?1, respectively. After QC, the extreme outliers out of the time series were eliminated. As a result, the correlation coefficient increased to greater than 0.8, andMAEwas reduced from approximately 5.6 m s?1to 2.8 m s?1. QC of the wind speed was the most effective, and the reliability of wind speed observations significantly improved.

    4 Conclusions

    Focusing on the air-sea interface, this study integrated various QC methods and established a scientific and practical QC procedure for the new drifting buoy observations in the order of an initial basic inspection and then targeted QC. The main conclusions and innovations are as follows.

    1) The QC procedure considers the swaying attitude of the buoy in the water, design observation indexes, and the error of the instrument itself, and it is more reasonable and more efficient than most QC methods that only rely on data.

    2) For different observational variables, respective methods were adopted to reduce the probability of misjudgment and significantly improve the effect of QC.

    3) The coefficient weight of the SST continuity check algorithm was optimized based on the differences in the sea area where the data came from. Thus, the algorithm was consistent with the characteristics of the actual air-sea environment, which expanded the application of the algorithm and effectively improved the accuracy of the QC.

    4) The innovative method combining moving Hampel filter and local anomaly detection ensured the reliability of QC results of meteorological variables in terms of mathematical statistics and actual physical processes.

    Although the evaluation of local and short-term observations was less reliable in demonstrating the effect of QC,overall, the outliers in the time series were effectively handled. After QC, the linear correlation between the observations from DrIB 8990 and KEO buoy as the control subjects increased, and the deviation decreased. The wind speed exhibited the best QC performance, the correlation coefficient before and after QC improved from around 0.77 to 0.82,andMAEwas reduced by approximately 2.8 m s?1. In addition, the quality of air pressure and relative humidity was optimized by 10?3– 10?2orders of magnitude.The results showed that the QC procedure for DrIB observations established in this study is scientific, effective,and practical, and it can provide references for the processing of observations from mobile platforms, promote the operational application of the new air-sea interface observation technologies, and offer data basis for marine environmental protection, ocean forecasting, air-sea interaction, and climate research. Marine observations are affected by many factors, such as the instruments themselves and changes in the ocean environment, and the data collected by DrIB are numerous and heterogeneous. Anomalous data also exhibit complex morphological characteristics. Therefore, finding a general QC method that can accurately identify and deal with all the anomalies is difficult. The QC procedure established in this study considers the circumstances that may lead to abnormalities in DrIB observations and has been experimentally verified to be feasible and practical. The QC procedure effectively detects independent and individual outliers in the time series.However, for continuous anomalous values, the QC performance needs further improvement, which will be the focus of our subsequent research.

    Acknowledgements

    This study is supported by the Natural Resources Development Special Fund Project of Jiangsu Province (No.JSZRHYKJ202009), the Taishan Scholar Funds (No. tsqn 201812022), the Fundamental Research Funds for the Central Universities (No. 202072001), the Guangxi Key Laboratory of Marine Disaster in the Beibu Gulf, Beibu Gulf University (No. 2021KF03), and the National Natural Science Foundation of China (No. 42176020). The authors are grateful for the financial support.

    91午夜精品亚洲一区二区三区| 亚洲av免费在线观看| 精品不卡国产一区二区三区| 亚洲无线观看免费| 少妇的逼水好多| 级片在线观看| 亚洲国产精品成人久久小说 | 亚洲精品久久久久久婷婷小说 | 日日撸夜夜添| 高清毛片免费观看视频网站| 麻豆av噜噜一区二区三区| 熟女人妻精品中文字幕| 久久精品夜色国产| 老司机影院成人| 国产高清视频在线观看网站| 一级毛片电影观看 | 亚洲精品亚洲一区二区| 亚洲精品亚洲一区二区| 村上凉子中文字幕在线| 婷婷亚洲欧美| 欧美激情国产日韩精品一区| 人人妻人人澡人人爽人人夜夜 | 一级毛片久久久久久久久女| 91在线精品国自产拍蜜月| 深爱激情五月婷婷| 免费看美女性在线毛片视频| 青青草视频在线视频观看| 午夜精品在线福利| 成人美女网站在线观看视频| 麻豆久久精品国产亚洲av| 久久亚洲精品不卡| 国产精品久久视频播放| 观看免费一级毛片| 亚洲欧美精品自产自拍| 免费不卡的大黄色大毛片视频在线观看 | 欧美最黄视频在线播放免费| 九九在线视频观看精品| 亚洲av.av天堂| 免费搜索国产男女视频| 欧美另类亚洲清纯唯美| 欧美高清成人免费视频www| 国产高潮美女av| 我的女老师完整版在线观看| 禁无遮挡网站| 国产在视频线在精品| 欧美不卡视频在线免费观看| 国产高清激情床上av| 丝袜美腿在线中文| 女人被狂操c到高潮| 国产在线精品亚洲第一网站| 边亲边吃奶的免费视频| 狂野欧美激情性xxxx在线观看| 51国产日韩欧美| a级一级毛片免费在线观看| 国产精品久久久久久久电影| 麻豆国产97在线/欧美| 亚洲欧美中文字幕日韩二区| 欧美成人免费av一区二区三区| 国产欧美日韩精品一区二区| 人人妻人人澡欧美一区二区| 99国产精品一区二区蜜桃av| 日韩欧美国产在线观看| 天堂中文最新版在线下载 | 色尼玛亚洲综合影院| 成年版毛片免费区| 亚洲一级一片aⅴ在线观看| 亚洲va在线va天堂va国产| 亚洲第一电影网av| 精品一区二区三区人妻视频| 国产国拍精品亚洲av在线观看| 国产老妇伦熟女老妇高清| 日韩视频在线欧美| 亚洲激情五月婷婷啪啪| 最近中文字幕高清免费大全6| 女人被狂操c到高潮| 国产探花极品一区二区| 国产 一区精品| 精品人妻视频免费看| 一夜夜www| 亚洲国产精品国产精品| 国产色婷婷99| 日本黄色视频三级网站网址| 欧美精品国产亚洲| 一区二区三区四区激情视频 | 国产一级毛片七仙女欲春2| 26uuu在线亚洲综合色| 亚洲18禁久久av| 精品欧美国产一区二区三| 国内精品久久久久精免费| 欧美激情久久久久久爽电影| 黄色日韩在线| 色综合亚洲欧美另类图片| 美女高潮的动态| 国产日韩欧美在线精品| 成人综合一区亚洲| 老女人水多毛片| 一本一本综合久久| 99riav亚洲国产免费| 一夜夜www| 国产乱人偷精品视频| 亚洲av一区综合| 久久这里只有精品中国| 男的添女的下面高潮视频| 色5月婷婷丁香| 69av精品久久久久久| 尤物成人国产欧美一区二区三区| 男人舔女人下体高潮全视频| 赤兔流量卡办理| 一本久久精品| 九九热线精品视视频播放| 人人妻人人澡欧美一区二区| 欧美日韩精品成人综合77777| 国产精品免费一区二区三区在线| 99久久精品热视频| 人体艺术视频欧美日本| 中文欧美无线码| 国产高清视频在线观看网站| 欧美在线一区亚洲| 爱豆传媒免费全集在线观看| 97热精品久久久久久| or卡值多少钱| 欧美成人a在线观看| 亚洲国产欧美在线一区| 国产精品人妻久久久影院| 天美传媒精品一区二区| 最近手机中文字幕大全| 国产精品蜜桃在线观看 | 人人妻人人看人人澡| 蜜桃久久精品国产亚洲av| 久久久久久久亚洲中文字幕| 国内精品一区二区在线观看| 女同久久另类99精品国产91| ponron亚洲| 亚洲中文字幕一区二区三区有码在线看| 熟妇人妻久久中文字幕3abv| 日本成人三级电影网站| 插逼视频在线观看| 国产免费一级a男人的天堂| 亚洲国产精品sss在线观看| 亚洲自偷自拍三级| videossex国产| 我要搜黄色片| 亚洲高清免费不卡视频| 六月丁香七月| 日韩成人av中文字幕在线观看| 久久久精品94久久精品| av视频在线观看入口| 久久韩国三级中文字幕| 亚洲国产色片| 国产精品美女特级片免费视频播放器| 美女 人体艺术 gogo| 亚洲国产精品成人综合色| 一级毛片电影观看 | 三级经典国产精品| 亚洲欧美精品自产自拍| 国产伦精品一区二区三区视频9| 日韩精品有码人妻一区| 久久婷婷人人爽人人干人人爱| 免费看a级黄色片| 高清毛片免费观看视频网站| 一边亲一边摸免费视频| 男女下面进入的视频免费午夜| 免费不卡的大黄色大毛片视频在线观看 | 嫩草影院入口| 亚洲一级一片aⅴ在线观看| 一级二级三级毛片免费看| 精品一区二区三区视频在线| 在线免费观看的www视频| 99热网站在线观看| 国产一区二区亚洲精品在线观看| 免费人成视频x8x8入口观看| 舔av片在线| 好男人视频免费观看在线| 能在线免费观看的黄片| 中文字幕免费在线视频6| 亚洲国产精品久久男人天堂| 国产91av在线免费观看| av在线老鸭窝| 大又大粗又爽又黄少妇毛片口| 国产伦在线观看视频一区| 在线a可以看的网站| 男人舔奶头视频| 国产精品无大码| 国国产精品蜜臀av免费| 深爱激情五月婷婷| 国产淫片久久久久久久久| 一本一本综合久久| 久久精品国产亚洲网站| 精品99又大又爽又粗少妇毛片| 国语自产精品视频在线第100页| 久久99热这里只有精品18| 99在线视频只有这里精品首页| 久久久久久久亚洲中文字幕| 日韩 亚洲 欧美在线| 欧美人与善性xxx| 夫妻性生交免费视频一级片| 国产精品女同一区二区软件| 麻豆av噜噜一区二区三区| 亚洲人成网站在线播放欧美日韩| 天美传媒精品一区二区| 国产乱人偷精品视频| 91午夜精品亚洲一区二区三区| 在线免费观看的www视频| 亚洲av免费高清在线观看| 老师上课跳d突然被开到最大视频| 三级男女做爰猛烈吃奶摸视频| 国产精品,欧美在线| 99热6这里只有精品| 最近最新中文字幕大全电影3| 91久久精品电影网| 欧美一区二区亚洲| 国产在视频线在精品| 欧美精品一区二区大全| 日日干狠狠操夜夜爽| 国产色婷婷99| 99久久久亚洲精品蜜臀av| av在线播放精品| 成人美女网站在线观看视频| 亚洲第一区二区三区不卡| 18禁在线播放成人免费| 夜夜夜夜夜久久久久| 女同久久另类99精品国产91| 亚洲无线在线观看| 久久久a久久爽久久v久久| 久久99精品国语久久久| 六月丁香七月| 欧美日本视频| 久久精品国产亚洲av香蕉五月| 精品一区二区三区视频在线| 欧美日韩综合久久久久久| 又粗又爽又猛毛片免费看| 久久久久久大精品| a级毛色黄片| 国产精品乱码一区二三区的特点| 日韩中字成人| 啦啦啦啦在线视频资源| 国产精品综合久久久久久久免费| 卡戴珊不雅视频在线播放| 日本一本二区三区精品| 自拍偷自拍亚洲精品老妇| 亚洲欧美清纯卡通| 天天一区二区日本电影三级| 久久中文看片网| 久久草成人影院| 国产毛片a区久久久久| 国产黄a三级三级三级人| 亚洲久久久久久中文字幕| 国产精品99久久久久久久久| 国产精品爽爽va在线观看网站| 精品无人区乱码1区二区| 国产熟女欧美一区二区| 欧美在线一区亚洲| 国产av麻豆久久久久久久| 一区福利在线观看| 毛片女人毛片| 一个人看视频在线观看www免费| 亚洲国产欧美在线一区| 极品教师在线视频| 国产精品久久久久久亚洲av鲁大| 日本免费a在线| 级片在线观看| 日韩精品有码人妻一区| 成人av在线播放网站| .国产精品久久| 听说在线观看完整版免费高清| 成熟少妇高潮喷水视频| 神马国产精品三级电影在线观看| 91在线精品国自产拍蜜月| 国产精品野战在线观看| 大型黄色视频在线免费观看| 日韩亚洲欧美综合| 波野结衣二区三区在线| 亚洲内射少妇av| 在线播放国产精品三级| 两个人的视频大全免费| 国产亚洲av嫩草精品影院| 欧美激情国产日韩精品一区| 又粗又硬又长又爽又黄的视频 | h日本视频在线播放| 激情 狠狠 欧美| 国产精品久久久久久久久免| 久久久久久九九精品二区国产| 美女内射精品一级片tv| 联通29元200g的流量卡| 一本—道久久a久久精品蜜桃钙片 精品乱码久久久久久99久播 | 国产成人a区在线观看| 美女大奶头视频| 97热精品久久久久久| 蜜桃亚洲精品一区二区三区| 亚洲色图av天堂| 国产男人的电影天堂91| 国产成人精品一,二区 | 亚洲国产精品久久男人天堂| 在线观看一区二区三区| 狠狠狠狠99中文字幕| 国产午夜精品一二区理论片| 男女边吃奶边做爰视频| 欧美极品一区二区三区四区| 免费看a级黄色片| 午夜精品在线福利| 国产精品人妻久久久久久| 美女黄网站色视频| 一级av片app| 国产免费一级a男人的天堂| 一级毛片我不卡| 在线播放无遮挡| 麻豆国产97在线/欧美| 午夜福利成人在线免费观看| 亚洲欧美日韩东京热| 中文精品一卡2卡3卡4更新| 成年av动漫网址| 12—13女人毛片做爰片一| 一区二区三区四区激情视频 | 精品熟女少妇av免费看| 我要搜黄色片| 两性午夜刺激爽爽歪歪视频在线观看| 哪里可以看免费的av片| 久久精品久久久久久久性| 国产高潮美女av| 成人午夜高清在线视频| 成人永久免费在线观看视频| 久久精品国产清高在天天线| 51国产日韩欧美| 精品一区二区三区人妻视频| 哪里可以看免费的av片| 午夜免费激情av| 少妇高潮的动态图| 久久久精品94久久精品| 啦啦啦韩国在线观看视频| 亚洲欧美成人综合另类久久久 | 午夜久久久久精精品| 亚洲av中文av极速乱| 毛片女人毛片| 人妻夜夜爽99麻豆av| 亚洲av第一区精品v没综合| 国产精品不卡视频一区二区| 国产高清激情床上av| 精品欧美国产一区二区三| av在线老鸭窝| 亚洲最大成人av| 免费人成在线观看视频色| 国产精品一区二区三区四区久久| 精品久久久噜噜| 91av网一区二区| 乱系列少妇在线播放| 亚洲国产高清在线一区二区三| 精品久久久久久久久av| 精品国内亚洲2022精品成人| 日韩欧美一区二区三区在线观看| 人妻久久中文字幕网| 国产白丝娇喘喷水9色精品| 美女xxoo啪啪120秒动态图| 麻豆一二三区av精品| 九九久久精品国产亚洲av麻豆| 国产91av在线免费观看| 人妻夜夜爽99麻豆av| 99久久精品国产国产毛片| 日韩视频在线欧美| 欧美一区二区国产精品久久精品| 日本在线视频免费播放| 大型黄色视频在线免费观看| 精品无人区乱码1区二区| 三级男女做爰猛烈吃奶摸视频| 嫩草影院入口| 免费av不卡在线播放| 听说在线观看完整版免费高清| 免费av观看视频| 久久这里只有精品中国| 中文在线观看免费www的网站| 菩萨蛮人人尽说江南好唐韦庄 | 可以在线观看毛片的网站| 色噜噜av男人的天堂激情| 亚洲av不卡在线观看| 国产免费一级a男人的天堂| 99在线视频只有这里精品首页| 亚洲国产精品合色在线| 69av精品久久久久久| 91狼人影院| 欧美性猛交╳xxx乱大交人| 精品久久久噜噜| АⅤ资源中文在线天堂| 乱码一卡2卡4卡精品| 亚洲av.av天堂| 国产精品久久久久久久久免| 午夜a级毛片| 可以在线观看的亚洲视频| 国产高潮美女av| 午夜老司机福利剧场| 在线观看免费视频日本深夜| 久久久久九九精品影院| 少妇高潮的动态图| 最后的刺客免费高清国语| 欧美区成人在线视频| 欧美zozozo另类| 亚洲最大成人中文| 中文字幕熟女人妻在线| 欧美激情国产日韩精品一区| av在线播放精品| 国产亚洲精品av在线| 在线观看美女被高潮喷水网站| 特级一级黄色大片| 国产视频首页在线观看| 国产免费男女视频| 国产精品爽爽va在线观看网站| 啦啦啦韩国在线观看视频| 国国产精品蜜臀av免费| 精品99又大又爽又粗少妇毛片| 男女做爰动态图高潮gif福利片| 国产乱人偷精品视频| 日本色播在线视频| 亚洲国产精品国产精品| 国产精品人妻久久久影院| 精品久久久久久久久久久久久| 99久久精品热视频| 精品久久久久久久久久免费视频| av免费在线看不卡| 国产探花在线观看一区二区| 免费不卡的大黄色大毛片视频在线观看 | 搡女人真爽免费视频火全软件| 成人亚洲精品av一区二区| 欧美一区二区精品小视频在线| 好男人在线观看高清免费视频| 久久热精品热| 麻豆成人午夜福利视频| 人人妻人人澡欧美一区二区| 久久草成人影院| 日韩欧美 国产精品| 99在线视频只有这里精品首页| 久久久久久九九精品二区国产| 国产精品久久久久久精品电影小说 | 青春草亚洲视频在线观看| 国产精品野战在线观看| 欧美一区二区亚洲| 欧美日本视频| 欧美zozozo另类| 欧美日韩综合久久久久久| 91在线精品国自产拍蜜月| 久久久精品欧美日韩精品| 99热全是精品| 又粗又爽又猛毛片免费看| 1024手机看黄色片| 日韩av不卡免费在线播放| 1024手机看黄色片| 日本与韩国留学比较| 最近2019中文字幕mv第一页| 日韩人妻高清精品专区| 亚洲av免费高清在线观看| 男女啪啪激烈高潮av片| 免费在线观看成人毛片| 久久热精品热| 久久久国产成人精品二区| 看非洲黑人一级黄片| 日本黄色片子视频| 日韩欧美一区二区三区在线观看| 蜜桃久久精品国产亚洲av| 免费不卡的大黄色大毛片视频在线观看 | 成人午夜高清在线视频| 日韩,欧美,国产一区二区三区 | 国内精品宾馆在线| 一区二区三区高清视频在线| 久久久国产成人精品二区| 日日啪夜夜撸| 欧洲精品卡2卡3卡4卡5卡区| 国产高清视频在线观看网站| 日韩大尺度精品在线看网址| 亚洲最大成人av| 国产毛片a区久久久久| 日日摸夜夜添夜夜爱| 一级黄色大片毛片| 给我免费播放毛片高清在线观看| 亚洲第一区二区三区不卡| 少妇人妻精品综合一区二区 | 亚洲婷婷狠狠爱综合网| 国产极品精品免费视频能看的| 一边摸一边抽搐一进一小说| 国产高潮美女av| 免费看av在线观看网站| 村上凉子中文字幕在线| 中文字幕人妻熟人妻熟丝袜美| 波多野结衣高清作品| 国产成人精品一,二区 | 国产精品国产高清国产av| 一夜夜www| 国产亚洲精品久久久久久毛片| 国产真实乱freesex| 99久久久亚洲精品蜜臀av| 久久99热6这里只有精品| 18禁黄网站禁片免费观看直播| 久久99热6这里只有精品| www日本黄色视频网| 最近最新中文字幕大全电影3| 九九热线精品视视频播放| 美女被艹到高潮喷水动态| 99九九线精品视频在线观看视频| 国产高清不卡午夜福利| 99riav亚洲国产免费| 亚洲精品自拍成人| 如何舔出高潮| 久久精品夜夜夜夜夜久久蜜豆| 久久久久久久亚洲中文字幕| 亚洲欧洲日产国产| 国产激情偷乱视频一区二区| 一夜夜www| 在线天堂最新版资源| 一本久久中文字幕| 中文字幕av成人在线电影| 国产精品蜜桃在线观看 | 国产成人aa在线观看| 久久久久免费精品人妻一区二区| 欧美性感艳星| 国产毛片a区久久久久| 亚洲婷婷狠狠爱综合网| 日韩成人伦理影院| 在线观看午夜福利视频| 色哟哟哟哟哟哟| 18禁在线播放成人免费| 国产探花极品一区二区| 成人午夜高清在线视频| 天天一区二区日本电影三级| 欧美色视频一区免费| 久久精品91蜜桃| 久久九九热精品免费| 最好的美女福利视频网| 春色校园在线视频观看| 久久99精品国语久久久| 国产又黄又爽又无遮挡在线| 国产亚洲5aaaaa淫片| 好男人在线观看高清免费视频| 蜜臀久久99精品久久宅男| 在线播放国产精品三级| 精品久久久久久久久亚洲| 亚洲av不卡在线观看| 亚洲av免费在线观看| 免费人成在线观看视频色| 亚洲欧美日韩卡通动漫| 日本免费一区二区三区高清不卡| 人妻少妇偷人精品九色| 搡女人真爽免费视频火全软件| 一级毛片aaaaaa免费看小| 国产av不卡久久| 在线播放无遮挡| 亚洲精品国产av成人精品| 日本五十路高清| 男人狂女人下面高潮的视频| 日韩三级伦理在线观看| 久久精品国产亚洲av涩爱 | 欧美日韩一区二区视频在线观看视频在线 | 好男人视频免费观看在线| av在线蜜桃| 少妇猛男粗大的猛烈进出视频 | 久久精品夜色国产| 男女做爰动态图高潮gif福利片| 观看免费一级毛片| 欧美性猛交黑人性爽| 小说图片视频综合网站| 一级毛片我不卡| 中文欧美无线码| 十八禁国产超污无遮挡网站| 国产视频首页在线观看| 色综合色国产| 欧美xxxx黑人xx丫x性爽| 成人毛片60女人毛片免费| 麻豆乱淫一区二区| 国产不卡一卡二| 国产精品久久久久久精品电影小说 | 亚洲,欧美,日韩| 免费观看的影片在线观看| 边亲边吃奶的免费视频| 成人亚洲精品av一区二区| 色综合站精品国产| 特级一级黄色大片| 老司机影院成人| 黄片wwwwww| 亚洲天堂国产精品一区在线| 少妇裸体淫交视频免费看高清| 人妻夜夜爽99麻豆av| 国产视频内射| 亚洲精品色激情综合| 欧美人与善性xxx| 亚洲国产精品成人综合色| 久久欧美精品欧美久久欧美| 午夜久久久久精精品| 免费看美女性在线毛片视频| h日本视频在线播放| 国产视频内射| 又粗又硬又长又爽又黄的视频 | 国产精品一区二区三区四区久久| 国产精品乱码一区二三区的特点| 男人的好看免费观看在线视频| 国产精品久久久久久亚洲av鲁大| 又黄又爽又刺激的免费视频.| 国产成人福利小说| 国产老妇女一区| 国产v大片淫在线免费观看| 亚洲内射少妇av| 精品少妇黑人巨大在线播放 | 久久精品久久久久久久性| av免费在线看不卡| 久久鲁丝午夜福利片| 青春草亚洲视频在线观看| 三级国产精品欧美在线观看| 婷婷色av中文字幕| 2021天堂中文幕一二区在线观| 黄色欧美视频在线观看| 国产精品免费一区二区三区在线| 国产伦精品一区二区三区四那| 国产精品一区二区三区四区久久| 欧美日韩国产亚洲二区| 麻豆国产av国片精品| 亚洲真实伦在线观看| 国产成人a区在线观看| 性插视频无遮挡在线免费观看| 国产一级毛片七仙女欲春2| 啦啦啦啦在线视频资源| 亚洲国产欧美人成|