Large lake water level time series encryption method based on altimetry satellite multi-track foot point fusion

Abstract

The application discloses a large lake water level time series encryption method based on multi-track satellite foot point fusion. Multi-track laser altimetry satellite foot point data located in a lake to be measured is extracted, and lake water level elevation values of all foot points are calculated based on a unified geodetic datum elevation coordinate system to obtain daily observation water levels of the multi-track. The track with the most observation times of each lake to be measured is taken as a reference track, and the remaining tracks are taken as correction tracks, offset between the reference track and the correction tracks is calculated, the correction track is translated and corrected, and multi-track fusion is realized. The water level time series after multi-track correction and fusion is optimized through a filtering method to obtain a dense lake water level time series after multi-track fusion. The method does not depend on measured water level data, can be popularized to lakes covered by multi-track altimetry satellites all over the world, and has important application significance for management and protection of lake water resources.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of lake hydrology remote sensing, and particularly relates to a large lake water level time series encryption method based on multi-track foot point fusion of altimetry satellites. BACKGROUND

[0002] Monitoring lake water level change is crucial for understanding the terrestrial water cycle. However, in-situ data of lake water level monitoring is very limited globally. In addition, these hydrological station water level data are usually not publicly available and difficult to obtain. It has become an urgent need to monitor lake water level using remote sensing technology, especially satellite altimetry technology. As a result, more and more radar or laser altimetry satellites have been launched for studying the dynamic change of water level.

[0003] Among these altimetry satellites, the Ice, Cloud and Land Elevation 2 satellite (ICESat-2) launched by NASA in September 2018 provides more refined footprint measurements that can be used to monitor smaller bodies of water, such as lakes, reservoirs and ponds. In addition, ICESat-2 also carries the Advanced Topographic Laser Altimeter System (ATLAS), which can obtain more accurate elevation and geographic positioning, bringing great help to the water level monitoring of small and medium-sized lakes. The observation advantage of ICESat-2 altimetry satellite makes it more concerned in the study of lake water level dynamics. Although there have been successful cases of using ICESat-2 altimetry satellite to extract lake water level in a region or even globally, there are still challenges in extracting lake water level using altimetry technology. For example, the time interval of the extracted lake water level is not dense enough and the data accuracy is uncertain. Moreover, for large lakes, it is more likely to be repeatedly observed by multiple orbits of ICESat-2 in one revisit cycle. However, when processing the water level elevation of ICESat-2 original altimetry data, we usually do not specifically distinguish different measurement orbits in one observation cycle. The common processing method is to simply take the average value of the lake water level extracted by different orbits in the whole cycle as the final lake water level. However, due to the different geodetic height of different orbits, there is a systematic bias in the extraction of large lake water level by different orbits. This simple calculation may introduce uncertainty in the final water level result. In addition, the surface elevation of a large lake has obvious slope difference at different positions. Another common processing method is to extract water level from one orbit, such as the orbit with the most observation dates. However, this processing method means that the height measurements of other orbits are discarded, thereby reducing the effective observation dates.

[0004] In summary, how to integrate the multi-orbit altimetry data of ICESat-2 is the key to accurately obtain the dense water level time series of large lakes. The existing water level extraction method of altimetry satellite cannot consider the density of observation time and the accuracy of water level extraction. Under this background, a large lake water level time series encryption method based on the fusion of multi-orbit foot point observation of laser altimetry satellite is proposed, which has important scientific and practical significance for scientifically understanding the hydrological cycle of land and the management and protection of water resources of land lakes. SUMMARY

[0005] The purpose of the present application is to address the shortcomings of the existing lake water level extraction method based on altimetry satellite, and to propose a large lake water level time series encryption method based on the fusion of multi-orbit foot points of altimetry satellite. This method uses publicly available laser altimetry satellite foot point data to achieve fine extraction of water level of large lakes with multi-orbit observation.

[0006] In order to achieve the above technical purpose, the present application adopts the following technical scheme:

[0007] The large lake water level time series encryption method based on the fusion of multi-orbit foot points of altimetry satellite comprises:

[0008] Extracting multi-orbit laser altimetry satellite foot point data located in the lake to be measured, and recalculating the lake water level elevation values of all foot points based on a unified geodetic datum elevation coordinate system to obtain daily observation water level of multi-orbit;

[0009] Taking the orbit with the most observation times of each lake to be measured as the reference orbit, and the remaining orbits as the correction orbits, calculating the offset between the reference orbit and the correction orbit, and performing translation correction on the correction orbit to realize multi-orbit fusion;

[0010] Optimizing the water level time series after multi-orbit correction and fusion by filtering method to obtain dense time series of lake water level after multi-orbit fusion.

[0011] As a preferred embodiment, the extraction method of multi-orbit laser altimetry satellite foot point data in the lake to be measured is:

[0012] Extracting the two-dimensional table of altimetry satellite attribute information, converting the two-dimensional table into spatial vector coordinate point data, and cutting the coordinate point data generated by the altimetry satellite foot point in combination with the lake water area range vector data to obtain the altimetry satellite foot point data in the lake to be measured.

[0013] As a preferred embodiment, all altimetry satellite foot point data intersecting with the boundary of the lake to be measured at all times are obtained by spatial intersection judgment, and are merged to serve as the laser altimetry satellite foot point data in the lake to be measured.

[0014] As a preferred embodiment, the method further comprises denoising the obtained multi-track daily observation water level, and removing outliers to be used in multi-track fusion calculation. Preferably, the normalized median absolute deviation (NMAD) method is used for denoising to remove errors of cloud, snow or saturated reflection signals; and observation points with less than n observations per observation day are excluded, and points with a difference from the average elevation greater than m times the standard deviation of the water level are removed to avoid trend deviation caused by insufficient time sampling, n and m being preset constants. The NMAD denoising method can automatically generate a threshold range for data constraint according to the water level time series of different lakes.

[0015] As a preferred embodiment, the ellipsoidal height of the EGM2008 geoid reference coordinate system is used to calculate the lake water level elevation value of all the extracted points. The spherical harmonic coefficients of the EGM2008 gravity field model are extended to 2190 orders, with a resolution of 9 km, and have higher accuracy and resolution.

[0016] As a preferred embodiment, when multiple tracks simultaneously satisfy the maximum number of observations (for example, ICESat-2 satellite, i.e., the track with the maximum number of cycles in the 91-day revisit period of the ICESat-2 satellite), the first track in the multiple tracks is used as the reference track.

[0017] As a preferred embodiment, the average deviation between the reference track and the track to be corrected is calculated as the offset of the track to be corrected.

[0018] As a preferred embodiment, the average value of the corrected daily observation water level of multiple tracks on the same day in the lake range is calculated as the water level value of the observation date, and then the water level time series after multi-track correction and fusion is optimized by filtering method.

[0019] As a preferred embodiment, the ensemble Kalman filtering method is used to optimize the water level time series after multi-track correction and fusion, which overcomes the limitation of other filtering methods that can only handle linear problems.

[0020] The present application has the following advantages:

[0021] (1) The present application does not rely on measured data, but only uses publicly available laser satellite altimetry data to complete the water level extraction of large lakes;

[0022] (2) The principle of the present application is simple and easy to operate, and realizes the automatic processing of obtaining the fused and optimized water level time series by only inputting multi-track altimetry water level data;

[0023] (3) The application is suitable for all lakes covered by multi-orbit altimetry satellites, and can be popularized to other laser or radar altimetry satellites with multi-orbit observation to extract lake water level. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 is the algorithm flowchart of the application.

[0025] Figure 2 is the multi-orbit observation schematic diagram of the application example.

[0026] Figure 3 is the multi-orbit fusion concept and effect diagram in the application example.

[0027] Figure 4 is the lake water level verification result of the application example.

[0028] Figure 5 is the optimization comparison result of the ordinary Kalman filter and the ensemble Kalman filter method in the application example.

[0029] Figure 6 is the multi-orbit altimetry water level fusion result of the lake in the application example. DETAILED DESCRIPTION

[0030] The specific embodiments of the application will be further described in detail below in combination with the drawings and examples. The following examples are used to illustrate the application, but are not used to limit the scope of the application.

[0031] The 18 large lakes with an area of more than 500 square kilometers distributed on all continents in the world are selected as the application examples of the application, including 6 lakes with measured water level for verification of the application, and 12 lakes for extrapolation of the application.

[0032] Example 1

[0033] As shown in Figure 1 is the algorithm flowchart of the application example, and the present example includes the following steps:

[0034] Step one, water level extraction.

[0035] Firstly, the latitude, longitude, time, geodetic height, and reference ellipsoid height of the ICESat-2 satellite foot point data are read. Then, the water surface mask of 18 lakes in this embodiment is used to extract the observation foot points of the ICESat-2 laser altimeter within the lake range. Specifically, the two-dimensional table of the altimeter attribute information is extracted, the two-dimensional table is converted into spatial vector coordinate point data, the coordinate point data generated by cutting the altimeter foot points is combined with the lake water area range vector data, and thus the altimeter foot point data within the lake range is obtained. The determination of the foot points within the lake range adopts the spatial intersection judgment function of the GIS spatial analysis function module, that is, all foot points intersecting with the lake water boundary at all times are merged as the observation data results of the altimeter of the lake.

[0036] Secondly, based on the ellipsoid height of the EGM2008 geoid reference coordinate system, the lake water level elevation values of all foot points extracted are recalculated.

[0037] Finally, the normalized median absolute deviation method (NMAD) is applied to remove the errors of cloud, snow or saturated reflection signals. In order to avoid the trend deviation caused by insufficient time sampling, the observation foot points with less than 5 observations per observation day are excluded, and the foot point data with a difference greater than 2.5 times the water level standard deviation from the elevation mean is removed.

[0038] Step two, water level fusion.

[0039] Using the automatic track correction processing program of the present application, the track with the most observations of each embodiment lake is found as the reference track of the respective lake. If there are multiple tracks that meet the most observations at the same time, the track with the most observations in the first traversal is used as the reference track. After the reference track is determined, the remaining tracks are used as the correction tracks. Then, the average deviation of the correction tracks and the reference track is calculated as the system error between the tracks, and the system error is used as the offset of the correction tracks. The average daily observation water level of the multiple tracks after correction in the lake range on the same day is calculated as the water level value of the observation date. The automatic multi-track processing program is repeated for all the lakes in the following embodiments, and the multi-track data fusion process of all the lakes is realized.

[0040] Step three, water level optimization.

[0041] The EnKF method is used to optimize the water level time series of the 18 sample lakes after multi-track correction and fusion. The specific implementation method of EnKF includes the water level initialization step (formula 1), the prediction step (formula 2), and the update step (formula 3-6). The specific calculation formulas are as follows:

[0042] (1)

[0043] In the initialization step (Equation 1), where is the set of size N at initial time. is randomly drawn from a normal distribution with a prior guess of mean and covariance of zero.

[0044] (2)

[0045] In the prediction step (Equation 2), where is the set member propagated forward using the stochastic model. F(X) is the stochastic model function. is randomly drawn from a normal distribution with computer model mean and covariance of zero.

[0046] (3)

[0047] (4)

[0048] (5)

[0049] (6)

[0050] The update step includes computing the Kalman gain matrix (Equations (3) and (4)), updating each set member using the perturbed water level (Equation (5)), and computing the analysis error covariance matrix (Equation (6)). is the evolving covariance of the prior guess, expressed as a spread of the evolving set at time . is the mean of the set. y(x) describes the relationship of the input water level to the current state. is randomly drawn from a normal distribution with computer model mean and covariance of zero. is the mean of the updated set at time .

[0051] The multi-orbit fused lake water level dense time series is obtained after water level optimization.

[0052] Embodiment 2

[0053] In this embodiment, the scheme of Embodiment 1 is verified by selecting six lakes with different sizes and spatial distribution differences with water level monitoring stations. Figure 4). The accuracy and fusion effect of the application in fusing multi-orbit ICESat-2 observation data were evaluated by using correlation coefficient (R), root mean square error (RMSE), mean absolute error (MAE), Nash coefficient (NSE) and coefficient of variation (CV). These indicators can reflect the comparison of multi-orbit ICESat-2 laser altimetry fused water level and measured water level and the dispersion of fused water level. Specifically, the verification evaluation index of each lake in the embodiment is calculated by the following formula respectively:

[0054] (7)

[0055] (8)

[0056] (9)

[0057] (10)

[0058] (11)

[0059] wherein, is the water level calculated by ICESat-2 in the period, is the average water level during the research period; is the measured water level of each water level monitoring station in the period, is the average water level during the research period; is the observation times of ICESat-2. By comparing the performance difference of KF and EnKF in optimizing the water level time series of multi-orbit ICESat-2 data (P<0.05),

[0060] the EnKF method which is smoother and more robust is selected to optimize the fused water level time series of multi-orbit. Figure 5

[0061] Example 3

[0062] In order to verify the feasibility of the application applied to other lakes, 12 lakes with different sizes and geographical positions were selected as the extended samples of the application, which were divided into 4 groups according to the size of the lakes, and the more intensive fused water level time series optimization results (P<0.05) Figure 6 were obtained by using the application.​​

Claims

1. A large lake water level time series encryption method based on altimetry satellite multi-track foot point fusion, characterized in that, include: Extract multi-track laser altimeter satellite footpoint data located in the lake to be measured, and recalculate the lake water level elevation values ​​extracted from all footpoints based on a unified geoid elevation coordinate system to obtain the daily observed water level from multiple orbits; The orbit with the most observations for each lake to be measured is used as the reference orbit, and the remaining orbits are used as the orbits to be corrected. The average deviation between the reference orbit and the orbit to be corrected is calculated as the offset of the orbit to be corrected. The orbit to be corrected is then translated and corrected. The average daily observed water level after multi-orbit correction within the lake area on the same day is calculated as the water level value on the observation date, thus realizing the fusion of multi-orbit data. The water level time series after multi-track data fusion is optimized by filtering method to obtain the dense time series of lake water levels after multi-track data fusion.

2. The method of claim 1, wherein, The method for extracting multi-orbit laser altimetry satellite footprint data within the lake to be measured is as follows: Extract the two-dimensional table of altimeter satellite attribute information, convert the two-dimensional table into spatial vector coordinate point data, and combine it with the vector data of the lake's water area to crop the coordinate point data generated by the altimeter satellite footpoints to obtain the altimeter satellite footpoint data within the lake to be measured.

3. The method according to claim 1 or 2, characterized in that, By determining spatial intersection, all time-intersecting satellite footpoint data intersecting with the boundary of the lake to be measured are obtained, and these data are merged to form the laser altimetry satellite footpoint data within the lake to be measured.

4. The method of claim 1, wherein, It also includes denoising the daily water level observations from the acquired multi-track data, removing outliers, and then using the data for multi-track data fusion calculation.

5. The method of claim 4, wherein, The normalized median absolute deviation (NMAD) method was used for noise reduction. Observation points with fewer than n water level observations per observation day were excluded, and data points with a difference from the mean elevation greater than m times the standard deviation of the water level were removed. n and m are preset constants.

6. The method of claim 1, wherein, Based on the ellipsoidal height of the EGM2008 geoid reference coordinate system, the lake water level elevation values ​​extracted from all foot points were calculated.

7. The method of claim 1, wherein, When multiple orbitals simultaneously satisfy the condition of having the maximum number of observations, the orbital that is traversed for the first time among these multiple orbitals is used as the reference orbital.

8. The method of claim 1, wherein, An ensemble Kalman filter method is used to optimize the water level time series after multi-track data fusion.

Images