CN112345499A

CN112345499A - International boundary lake transparency inversion method based on multi-source remote sensing satellite

Info

Publication number: CN112345499A
Application number: CN202011067772.9A
Authority: CN
Inventors: 房冲; 辛卓航; 李昱; 宋长春; 张弛
Original assignee: Dalian University of Technology
Current assignee: Dalian University of Technology
Priority date: 2020-10-07
Filing date: 2020-10-07
Publication date: 2021-02-09
Anticipated expiration: 2040-10-07
Also published as: CN112345499B

Abstract

An international boundary lake transparency inversion method based on a multi-source remote sensing satellite belongs to the technical field of environmental science and remote sensing monitoring. Firstly, uniformly distributing sampling points in domestic water areas of international boundary lakes, and acquiring a large amount of actually measured SDD data in different periods; secondly, acquiring the water surface remote sensing reflectivity of the relevant wave bands of different remote sensing satellites; then, dividing the well matched point pairs into building modules and verification groups according to the proportion; thirdly, respectively constructing SDD remote sensing inversion models suitable for remote sensing images of different satellites by using the related wave band water surface remote sensing reflectivity of the building module and different remote sensing satellites, and evaluating the model precision by using verification group data; and finally, selecting different remote sensing images on the same day, respectively carrying out remote sensing inversion, and carrying out consistency evaluation on results obtained by inversion of different sensors. The invention can obtain the SDD of the whole lake at one time by researching the SDD through the remote sensing satellite image, has strong timeliness and is simple and convenient to operate; the application range of the remote sensing image inversion method can be widened.

Description

International boundary lake transparency inversion method based on multi-source remote sensing satellite

Technical Field

The invention belongs to the technical field of environmental science and remote sensing monitoring, and relates to a method for estimating transparency of a lake reservoir based on multi-source remote sensing images, in particular to a method for inverting and verifying model precision of the lake reservoir positioned at the national boundary of China based on multi-source remote sensing satellite data.

Background

The water transparency (SDD) of lake reservoirs is defined as: a circular dish of approximately 30cm diameter, white or black and white alternating, was placed vertically in water to a depth just barely visible to the human eye, namely the SDD of the water body [ Tyler, J.E. the secchi disc, Limnology and oceanography13,1-6(1968) ]. The SDD of the lake reservoir is one of basic parameters for describing the optical characteristics of the water body of the lake reservoir, can visually reflect the clarity and turbidity degree of the water body, is an important index for evaluating the eutrophication of the water body and measuring the quality of water quality [ the Yudine peak, the Chen predecessor and the Shi Ping. According to the definition of the lake and reservoir SDD, the SDD does not have the obvious absorption curve characteristic of chlorophyll a, but is determined by various optical components in the water body, the remote sensing reflectivity of the water surface is increased due to the back scattering generated by suspended particles such as suspended silt, phytoplankton, clay, microorganisms, organic and inorganic debris and the like in the water body, the transmittance of sunlight is disturbed, and further the visible depth of a visible light band is reduced, and the lake SDD is reduced [ Jianghui, Poyang lake water quality parameter inversion and analysis based on multi-source remote sensing, Nanchang university, (2011) ]. However, with the acceleration of the urbanization process and the stress of global climate change, the ecological safety of water in lakes and reservoirs in China faces a great threat, the SDD in lakes and reservoirs is reduced, and the water quality is worsened, so that the monitoring of the SDD in lakes and reservoirs is realized rapidly and in a large range, and the monitoring is urgent.

Compared with the traditional lake and reservoir investigation method, the remote sensing satellite has the advantages of one-time large-range observation, strong real-time performance, free acquisition, dynamic sustainable observation and the like, so that many scholars detect the lake water body SDD by means of remote sensing satellite images, and common inversion models comprise three types according to research mechanisms: semi-analytical models, quasi-analytical models, and empirical models. Although the method has better optical interpretation and applicability in the aspect of physical mechanism compared with semi-analysis and quasi-analysis methods, the two models have more complex acquisition parameters, and many parameters cannot be acquired under the limitation of an observation instrument [ Schmidt, g., Jenkerson, c.b., mask, j.et al. However, due to political factors, the international boundary lake can only acquire a part of water area conditions or even a small part of water area conditions when the international boundary lake is acquired in the field, and the international boundary lake cannot acquire foreign water areas, so that the reliability of model construction is greatly reduced. Therefore, it is very necessary to construct a model of the SDD suitable for international lake boundary and develop a model verification method specific to international lake boundary.

Disclosure of Invention

The invention aims to solve the problems that field sampling of international boundary lakes can only obtain partial water area conditions, foreign water areas cannot be collected, and the estimation accuracy of an SDD model of the foreign water areas cannot be verified.

In order to achieve the purpose, the technical scheme adopted by the invention is as follows:

a multisource remote sensing satellite-based international boundary lake SDD inversion method comprises the steps of firstly, uniformly distributing sampling points in domestic water areas of the international boundary lake to obtain a large amount of actually measured SDD data in different periods; secondly, matching all the actually measured sampling points with the multi-element remote sensing satellite in a planet ground to obtain the water surface remote sensing reflectivity of the related wave bands of different remote sensing satellites; then, dividing the well matched point pairs into building modules and verification groups according to the proportion of 2: 1; thirdly, respectively constructing SDD remote sensing inversion models suitable for remote sensing images of different satellites by using the related wave band water surface remote sensing reflectivity of the building module and different remote sensing satellites, and evaluating the model precision by using verification group data; and finally, selecting different remote sensing images on the same day, respectively carrying out remote sensing inversion, and carrying out consistency evaluation on results obtained by inversion of different sensors. The method specifically comprises the following steps:

the method comprises the steps of firstly, selecting a proper international boundary lake as a research area, selecting clear and cloudless Landsat satellite transit time, uniformly distributing sampling points, collecting SDDs of different sampling points by using black and white disks as measuring tools, and simultaneously recording longitude and latitude information of the sampling points.

And secondly, acquiring the longitude and latitude information of all sampling points obtained by field cruise measurement, acquiring the water surface remote sensing reflectivity of visible light-near infrared bands of Landsat ETM +/OLI and MODIS images passing within 48 hours, removing invalid sampling points influenced by cloud layers, noise and the like, and respectively preparing Landsat effective arrays and MODIS effective arrays with matched SDD.

And thirdly, dividing effective array sampling points made by MODIS and Landsat into a building module and a verification module according to the quantity distribution of 2:1, ensuring that the two groups of data are uniformly distributed in the domestic water area part, and ensuring that the two groups of data contain sampling points in different stages.

Fourthly, obtaining an estimation model of the SDD based on the MODIS satellite image and the Landsat satellite image respectively:

and 4.1) taking SDD as a dependent variable, obtaining band ratios by arranging and combining single-band and multi-band data of different bands of MODIS as independent variables, sequentially carrying out regression analysis, screening the band or band ratio with the largest decision coefficient as a preselected SDD inversion model, then applying the preselected model to a verification data set, and selecting the SDD estimation model with the best evaluation result.

4.2) constructing and verifying a Landsat model in the same step as 4.1), specifically, taking SDD as a dependent variable, obtaining wave band ratios as independent variables respectively by arranging and combining single-wave band and multi-wave band data of Landsat ETM +/OLI different wave bands, sequentially carrying out regression analysis, screening the wave band or wave band ratio with the largest decision coefficient as a preselected SDD inversion model, then applying the preselected model to a verification data set respectively, and selecting the SDD estimation model with the best evaluation result.

And fifthly, selecting MODIS and Landsat remote sensing images on the same day, obtaining two SDD inversion results by applying the model constructed in the fourth step, resampling the SDD obtained by Landsat image inversion to 500m, deriving SDD data point pairs obtained by Landsat and MODIS inversion, and evaluating the consistency of the international boundary lake whole lake inversion results, namely realizing the evaluation of the model construction effect of the non-sampling area of the international boundary lake.

Compared with the existing SDD research method, the beneficial effects of the invention are embodied in the following aspects:

(1) compared with the traditional SDD monitoring method, the SDD of the whole lake can be obtained at one time through researching the SDD by the remote sensing satellite image, the timeliness is strong, the cost is low, the operation can be finished by one notebook or a common computer, the operation is simple and convenient, the historical data can be obtained, and the lake area which cannot be obtained due to political, terrain and other reasons can also be monitored through the remote sensing satellite image.

(2) Compared with the existing SDD inversion method, the existing SDD research targets are complete lakes within the national boundary such as Taihu lake, the sample point distribution of the lakes is uniformly distributed throughout the whole lake, the bright point of the method breaks the traditional thinking, the research target is positioned as the international boundary lake, and the application range of the remote sensing image inversion method is greatly improved.

(3) Compared with the existing SDD remote sensing inversion model verification method, the traditional method for verifying the model precision based on the actual measurement sampling points fails to be effective for the international boundary lake because a considerable part of the international boundary lake cannot obtain the actual measurement sampling points.

Drawings

FIG. 1 is a distribution diagram of sampling points of Xingkai lake of International boundary lake;

FIG. 2 is an inversion model and accuracy evaluation diagram constructed by the Xingkai lake SDD based on the Landsat satellite images; fig. 2(a) is a model construction diagram of modeling group data, and fig. 2(b) is a model evaluation diagram of verification group data;

FIG. 3 is an inversion model and accuracy evaluation diagram constructed by the SDD of Xingkai lake based on MODIS satellite images; fig. 3(a) is a model construction diagram of modeling group data, and fig. 3(b) is a model evaluation diagram of verification group data;

FIG. 4 is a diagram illustrating consistency verification of inversion results of Landsat and MODIS; fig. 4(a) is a diagram showing the consistency verification of the inversion results of the OLI sensors in MODIS and Landsat8, and fig. 4(b) is a diagram showing the consistency verification of the inversion results of the ETM + sensors in MODIS and Landsat 7.

Detailed Description

The present invention is further illustrated by the following specific examples.

Firstly, as shown in fig. 1, the invention selects a famous international lake of China-xingkai lake as a research area, samples are carried out in non-ice sunny and windless weather in 2013 and 2018, sample points are uniformly distributed, discs with alternate black and white are used as measuring tools, SDDs of different sample points are collected, longitude and latitude information of the sample points is recorded at the same time, and 153 effective sampling points are obtained in all sampling times.

And secondly, acquiring the longitude and latitude information of all sampling points obtained by field cruise measurement, acquiring the water surface remote sensing reflectivity of visible light-near infrared bands of Landsat ETM +/OLI and MODIS images passing within 48 hours, removing invalid sampling points influenced by cloud layers, noise and the like, and respectively preparing Landsat effective arrays and MODIS effective arrays with matched SDD. The multi-source remote sensing image data source mainly comprises ETM + and OLI images of a Landsat satellite and MODIS images of a Terra satellite and an Aqua satellite. ETM +/OLI image data in Landsat series sensors used in the invention is a land remote sensing reflectivity product (product) released by the United States Geological Survey (USGS)https://www.usgs.gov/land-resources/nli/Landsat/Landsat-surface-reflection), which uses the LEDAPS method to perform atmospheric correction on the raw data provided by Landsat; the data can be downloaded through USGS official website (https:// Earth. explorer. USGS. gov /) and also can be accessed and obtained through Google Earth Engine website (https:// code. Earth. Google. com /). The MODIS satellite image data used in the invention are MOD09GA product and MYD09GA product issued by NASA, and the MODIS satellite image data and the MYD09GA product are generated by respectively carrying out atmospheric correction on data acquired by Terra of the morning star and Aqua of the afternoon starThe product data of (1). MOD09GA and MYD09GA are commodity products that can be downloaded and retrieved from the MODIS data official website (https:// ladssweb. modaps. eosdis. nasa. gov/search /), then undergo clipping and batch transformation projection based on MRT, and then undergo band fusion in batches by means of the IDL program.

And thirdly, dividing effective array sampling points made by MODIS and Landsat into a building module and a verification module according to the quantity distribution of 2:1, ensuring that the two groups of data are uniformly distributed in the domestic water area part, and ensuring that the two groups of data contain sampling points in different stages. Among the 153 effective SDD sampling points, 103 effective sampling points which can be matched with the transit Landsat satellite image within 48 hours of a time window are available, 69 point pairs of the effective sampling points are used for model construction, and the other independent 34 point pairs are used for model verification; there are 128 effective sampling points matched with the MODIS satellite images passing within 48 hours of the time window, wherein 84 point pairs are used for model construction, and other independent 44 point pairs are used for model verification.

And fourthly, taking the SDD as a dependent variable, obtaining wave band ratios by arranging and combining single-wave band and multi-wave band data of Landsat different wave bands as independent variables, sequentially carrying out regression analysis, screening the wave band or the wave band ratio with the largest decision coefficient as a preselected SDD inversion model, then applying the preselected model to a verification data set respectively, and selecting the SDD estimation model with the best evaluation result. The MODIS model construction and verification steps are as above.

The finally obtained estimation model of the SDD base MODIS satellite image is shown in a formula (1):

SDD_MODIS＝0.4554×exp(4.1823×Green/Red) (1)

in the formula, Green and Red are Green band and Red band of the remote sensing satellite image of MODIS, respectively, and correspond to 4 band and 1 band of MODIS, respectively, and specific model construction and estimation effect evaluation thereof are shown in FIG. 3.

The estimation model of the finally obtained SDD based on the Landsat satellite image is shown in a formula (2):

SDD_ETM+/OLI＝0.1631×exp(5.1092×Green/Red) (2)

in the formula, Green and Red are respectively a Green band and a Red band of the Landsat remote sensing satellite image, which respectively correspond to 2 bands and 3 bands of ETM +, and 3 bands and 4 bands of a sensor of an OLI sensor, and specific model construction and estimation effect evaluation are shown in fig. 2.

Fifthly, however, the construction and verification of the model are based on the fact that the Xingkai lake is developed at a water sampling point in the interior of China, and the model precision cannot be evaluated through actually measured sampling data in Russia because actually measured sampling data cannot be obtained due to political factors. In order to solve the usability verification evaluation of the model in international lake, the research creatively provides a new idea of mutual verification through a multi-source remote sensing satellite, namely, ETM + and MODIS images, OLI and MODIS images in the same period are respectively inverted through the constructed Landsat and MODIS inversion models, and if the model estimation is accurate and the usability is good, the sample point pairs of the ETM + and the MODIS, the sample point pairs of the OLI and the MODIS are uniformly distributed on two sides of a 1:1 line. It should be noted that, the spatial resolution of the image of the Landsat series sensor is 30m, and the spatial resolution of the MODIS image is 500m, in order to verify the consistency of the inversion effects of the two sensors, the SDD image obtained by inverting Landsat ETM + and OLI needs to be resampled to 500m spatial resolution first.

Based on the above assumptions, the OLI image and MOD09GA data of 2016, 9, 26 and the like of the contemporaneous transit are randomly selected, the obtained verification effect is shown in FIG. 4a, the sampling points are uniformly distributed on two sides of a 1:1 line, and the RMSE and the MRE are both very low, so that the OLI and MODIS inversion results are very good in consistency; similarly, based on the data of ETM + and MODIS passing through the same period of 9/21/2017, the obtained verification effect is as shown in fig. 4b, the sampling points are also uniformly distributed on two sides of the 1:1 line, and the RMSE and the MRE are both very low, so that the inversion results of ETM + and MODIS are very good in consistency.

The above-mentioned embodiments only express the embodiments of the present invention, but not should be understood as the limitation of the scope of the invention patent, it should be noted that, for those skilled in the art, many variations and modifications can be made without departing from the concept of the present invention, and these all fall into the protection scope of the present invention.

Claims

1. A multisource remote sensing satellite-based SDD inversion method of international boundary lakes is characterized in that firstly, sample points are uniformly distributed in domestic water areas of the international boundary lakes, and a large amount of actually measured SDD data in different periods are obtained; secondly, matching all the actually measured sampling points with the multi-element remote sensing satellite in a planet ground to obtain the water surface remote sensing reflectivity of the related wave bands of different remote sensing satellites; then, dividing the well matched point pairs into building modules and verification groups according to the proportion; thirdly, respectively constructing SDD remote sensing inversion models suitable for remote sensing images of different satellites by using the related wave band water surface remote sensing reflectivity of the building module and different remote sensing satellites, and evaluating the model precision by using verification group data; finally, different remote sensing images of the same day are selected, remote sensing inversion is respectively carried out, and consistency evaluation is carried out on results obtained by inversion of different sensors; the method specifically comprises the following steps:

selecting a proper international boundary lake as a research area, selecting clear and cloudless Landsat satellite transit time, uniformly distributing sampling points, collecting SDDs (software development description) of different sampling points by using a black-white disc as a measuring tool, and simultaneously recording longitude and latitude information of the sampling points;

secondly, acquiring longitude and latitude information of all sample points obtained by field cruise measurement, acquiring the water surface remote sensing reflectivity of visible light-near infrared bands of Landsat ETM +/OLI and MODIS images passing within 48 hours, removing invalid sample points, and respectively preparing Landsat effective arrays and MODIS effective arrays with matched SDDs;

thirdly, dividing effective array sampling points made by MODIS and Landsat into a building module and a verification module according to the quantity distribution of 2:1, ensuring that the two groups of data are uniformly distributed in the domestic water area part, and the two groups of data comprise sampling points at different stages;

4.1) taking SDD as a dependent variable, obtaining wave band ratios by arranging and combining single-wave band and multi-wave band data of different wave bands of MODIS as independent variables respectively, sequentially carrying out regression analysis, screening and determining the wave band or wave band ratio with the maximum coefficient as a preselected SDD inversion model, then applying the preselected model to a verification data set respectively, and selecting the SDD estimation model with the best evaluation result;

4.2) taking SDD as a dependent variable, obtaining wave band ratios by arranging and combining single-wave band and multi-wave band data of Landsat ETM +/OLI different wave bands as independent variables respectively, sequentially carrying out regression analysis, screening the wave band or the wave band ratio with the largest decision coefficient as a preselected SDD inversion model, then applying the preselected model to a verification data set respectively, and selecting the SDD estimation model with the best evaluation result;