CN106202878A - A kind of long sequential remote sensing soil moisture NO emissions reduction method - Google Patents

Abstract

The invention discloses long sequential remote sensing soil moisture NO emissions reduction method, it includes: long sequential low resolution remotely-sensed data, according to the acquisition situation of high-resolution Land Surface Parameters, is divided into two parts by (one)；(2) according to the statistical relationship between soil moisture and Land Surface Parameters, obtaining Land Surface Parameters can high-resolution soil moisture data in acquisition period；(3) data obtained with step (two) are as reference data, according to soil moisture " time stability " feature, obtain soil moisture conversion parameter between different scale, and then the Part II data product that step () division obtains is converted to high-resolution soil moisture data.There is advantages that the present invention can effectively solve the soil moisture NO emissions reduction dependency to high-resolution data, make under limited data support, obtain long sequential high-resolution soil moisture and become a reality；Inventive algorithm simple in construction, it is to avoid the problems such as physical model structure is complicated, input parameter is many.

Description

A kind of long sequential remote sensing soil moisture NO emissions reduction method

Technical field

The present invention relates to a kind of soil moisture NO emissions reduction method, particularly a kind of coupling Land Surface Parameters auxiliary is changed and different Between yardstick, the long sequential NO emissions reduction method of statistical relationship, belongs to remote sensing technology field.

Background technology

Soil moisture is the state variable that earth system is important, and its change in time and space is handed over aqueous vapor for ground vapour system capacity Change and play an important role, water cycle process and agricultural production etc. all can be caused corresponding impact, be that the important of hydrology ecosphere is ground Study carefully object.In recent years, remote sensing (especially microwave remote sensing) obtains in the monitoring soil moisture of basin with features such as its Spatial continual Extensively application.But, owing to development of remote sensing history is shorter, the restricted lifetime of different platform in addition, existing Remote Sensing Products Time range is shorter.Additionally, microwave remote sensing soil moisture spatial resolution is relatively low, it is difficult to the spatial detail of reflection soil moisture is special Levy.NO emissions reduction method is the effective way obtaining high-resolution soil moisture data set, and many scholars have carried out extensively in this respect Research, is broadly divided into Land Surface Parameters auxiliary spatial scaling and based on soil moisture statistical relationship two between different scale from method Kind.

The former mainly utilizes soil moisture and other high spatial resolution Land Surface Parameters (such as ground mulching, surface temperature etc.) Between physics or empirical statistical relationship carry out spatial scaling.

Second method is based primarily upon soil moisture " time stability (Temporal stability) " principle, i.e. different Between yardstick, soil moisture has consistent variation tendency on long-term sequence, by the mutual pass of soil moisture between different scale Low yardstick data are changed to higher spatial resolution by system.

Being limited by factors such as observation condition and technology, ground auxiliary parameter and high-resolution soil moisture data product lead to It is commonly present the situations such as shortage of data, causes said method still to face many when building long sequential soil moisture data set and choose War, becomes basin soil moisture and the most precisely monitors problem demanding prompt solution.

Summary of the invention

For deficiencies of the prior art, it is an object of the invention to provide a kind of coupling Land Surface Parameters auxiliary conversion And the long sequential NO emissions reduction method of statistical relationship between different scale.

For achieving the above object, the present invention uses following technological means:

A kind of long sequential remote sensing soil moisture NO emissions reduction method, it is characterised in that it comprises the following steps:

(1) divide according to acquisition situation sequential long to the CCI remote sensing soil moisture data product of MODIS data:

Long for CCI sequential remote sensing soil moisture data product is divided into Part I CCI data product and Part II CCI Data product；Described Part I CCI data product is that MODIS data can obtain the long sequential remote sensing soil water of the CCI in the period Divided data product；Described Part II CCI data product is the CCI long sequential remote sensing soil water in the MODIS shortage of data period Divided data product；

(2) shown that calendar year 2001 is the soil moisture data of 1Km to intrinsic resolution in 2013 by MODIS data:

(1) reception calendar year 2001 is the MODIS data of 1Km to intrinsic resolution in 2013, and described MODIS data include: vegetation Exponent data NDVI and surface temperature data T；

(2) data NDVI described in vegetation index and surface temperature data T are normalized, eliminate in number of levels Diversity:

T^*=(T-T_min)/(T_max-T_min) (1)

NDVI^*=(NDVI-NDVI_min)/(NDVI_max-NDVI_min) (2)

Wherein, T_maxFor the maximum of surface temperature data T, T_minMinima for surface temperature data T；NDVI_maxFor returning One maximum changing vegetation index data NDVI, NDVI_minFor the minima of normalized differential vegetation index data NDVI, T^*For normalizing Surface temperature data after change, NDVI^*For the vegetation index data after normalization；

(3) to surface temperature data T after normalization^*With vegetation index data NDVI after normalization^*It is resampled to Resolution is 25Km；

Based on surface temperature data T after the normalization in the calendar year 2001 to 2013 that resampling obtains^*After normalization Vegetation index data NDVI^*Set up the statistical relationship of first stage CCI soil moisture data:

$SSM=\underset{i=0}{\overset{n}{\Σ}}\underset{j=0}{\overset{n}{\Σ}}{a}_{ij}\·{\mathrm{NDVI}}^{*\left(i\right)}\·T^{*\left(j\right)}---\left(3\right)$

Wherein a_ijRepresenting weight coefficient, i, j represent the dimension of independent variable respectively, and SSM is the soil in calendar year 2001 to 2013 Earth moisture data；

(4) it is the earth's surface of 1Km by vegetation index data NDVI that resolution is 1Km and the resolution that obtain in step (1) Temperature data T data substitute in formula (1), (2) and (3), it is achieved the NO emissions reduction of soil moisture processes, and show that calendar year 2001 is to 2013 Resolution in Nian is the soil moisture data of 1Km；

(3) data obtained with step (two) are as reference data, obtain soil moisture spatial scaling parameter, and then will step Suddenly (one) divides the Part II CCI data product that obtains and is converted to the soil moisture data that resolution is 1Km:

(1) matching conversion coefficient is determined:

Between different scale, the relation of soil moisture is

${\mathrm{SSM}}_{x,y,t}^F=a_{x,y}\·{\mathrm{SSM}}_{x,y,t}^C+b_{x,y}---\left(4\right)$

Wherein,WithIt is expressed as t (x, y) high-resolution and corresponding low resolution thereof on position Soil moisture in rate pixel；a_x,yAnd b_x,yRepresent matching conversion coefficient；

The soil moisture data that resolution is 1Km in the calendar year 2001 to 2013 obtain step (two) substitute into formula (4) InStep (one) divides the Part I CCI data obtained substitute in formula (4)Show that matching turns Change coefficient a_x,yAnd b_x,yValue；

(2) step () is divided the Part II CCI data obtained to substitute in formula (4)By step (1) the matching conversion coefficient a obtained in_x,yAnd b_x,yValue substitute into a in formula (4) respectively_x,yAnd b_x,y, obtain 1979 to 2000 Resolution in Nian is the soil moisture data of 1Km.

Further, described MODIS data can obtain the period is that calendar year 2001 was to 2013.

Compared to existing technology, there is advantages that

Being limited by factors such as observation condition and technology, ground auxiliary parameter and high-resolution soil moisture data product lead to It is commonly present the situations such as shortage of data, causes normally used two kinds of methods in prior art building long sequential soil moisture data Many challenges are still faced during collection.The present invention comprehensive existing methods feature and superiority-inferiority, it is proposed that coupling Land Surface Parameters auxiliary The long sequential remote sensing soil moisture NO emissions reduction method of statistical relationship between conversion and different scale, compared to existing technology, the present invention has There is following advantage: (1) present invention can effectively solve the soil moisture NO emissions reduction dependency to high-resolution data, makes limited Obtain long sequential high-resolution soil moisture under data support to become a reality；(2) inventive algorithm simple in construction, has well Operability, it is to avoid the problems such as physical model structure is complicated, input parameter is many, has stronger practicality and a generalization.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of the present invention；

Fig. 2 is soil moisture NO emissions reduction result schematic diagram based on Land Surface Parameters auxiliary scale-transformation method；

Fig. 3 is 1979-2013 many annuals soil moisture spatial distribution map based on this algorithm

Fig. 4 is that Butut is distinguished in the checking of NO emissions reduction soil moisture space characteristics

Fig. 5 is TVDI and original CCI data, the dependency diagram of NO emissions reduction soil moisture；

Fig. 6 is NO emissions reduction calibration and Checkout result schematic diagram.

Detailed description of the invention

Below in conjunction with the accompanying drawings and embodiment, the present invention is described in further details.

One, the present invention comprehensive existing methods feature and superiority-inferiority, the CCI long sequential (1979-2013 delivered with European Space Agency Year, 25Km) as a example by soil moisture, disclose the long sequential fall of statistical relationship between coupling Land Surface Parameters auxiliary conversion and different scale Two time scales approach.

As it is shown in figure 1, the invention mainly comprises three big steps: (one) according to the acquisition situation of high-resolution Land Surface Parameters, Long sequential low resolution remotely-sensed data is divided into two stages；(2) close according to the statistics between soil moisture and Land Surface Parameters System, is drawn the high-resolution soil moisture data in its corresponding time by high-resolution Land Surface Parameters data；(3) with step (two) Obtain data be reference data, obtain soil moisture spatial scaling parameter, and then by step () divide obtain without high score The soil moisture conversion high-resolution data in resolution Land Surface Parameters period.

Hereinafter each step is described in detail.

(1) sequential remote sensing soil moisture data product long to the CCI delivered by European Space Agency divides.

Basic ideas are: according to high-resolution Land Surface Parameters data, (with 2001-2013 MODIS data instance, space is divided Resolution 1km) acquisition situation, to long sequential low resolution remote sensing soil moisture data product (with European Space Agency 1979-2013 As a example by CCI soil moisture, spatial resolution 25km) divide: Part I can obtain the period corresponding to Land Surface Parameters (2001-2013), Part II is Land Surface Parameters loss period.

MODIS starts formally to issue data from April, 2000, China the most established several receiving station and respectively at Calendar year 2001 starts around March to receive data.Accordingly, because the acquisition time of high-resolution MODIS data starts from calendar year 2001, thus Soil moisture NO emissions reduction before this cannot be applied to based on Land Surface Parameters auxiliary scale-transformation method process.

(2) shown that calendar year 2001 is the soil moisture data of 1Km to intrinsic resolution in 2013 by MODIS data:

Basic ideas are: it is known in the art that soil moisture is closely related with the factor such as vegetative coverage, surface temperature. Based on this principle, MODIS normalized differential vegetation index (the Normalized Difference that the present invention uses resolution to be 1Km Vegetation Index, NDVI) and surface temperature (Land surface temperature, LST) data product, set up with The empirical equation of soil moisture, it is achieved the NO emissions reduction of soil moisture processes.

First NDVI data and LST data that resolution is 1Km are normalized by basic process, eliminate the order of magnitude The diversity not gone up, formula is:

T^*=(T-T_min)/(T_max-T_min) (1)

NDVI^*=(NDVI-NDVI_min)/(NDVI_max-NDVI_min) (2)

Wherein, T_max, T_minIt is respectively maximum and minimum surface temperature, NDVI_max, NDVI_minIt is respectively maximum and minimum NDVI value, T^*And NDVI^*For data after normalization.

Then with LST data, normalization NDVI being resampled to resolution is that (in remote sensing, resampling is from high score to 25Km Resolution remote sensing image extracts the process of low resolution image), and statistical relationship is set up with corresponding CCI data:

$SSM=\underset{i=0}{\overset{n}{\Σ}}\underset{j=0}{\overset{n}{\Σ}}{a}_{ij}\·{\mathrm{NDVI}}^{*\left(i\right)}\·T^{*\left(j\right)}---\left(3\right)$

Wherein a_ijRepresenting weight coefficient, i, j represent the dimension of independent variable, secondary or cubic regression equation respectively and can obtain Preferably result.Due to tri-variablees of SSM, NDVI and T all it is known that i.e. can get weight coefficient a by inverse_ij.Finally, since The parameters such as SSM, NDVI and the T in formula (3) are it is known that i.e. can get coefficient a by inverse_ij, and then this coefficient available is by former NDVI and the LST data of the 1Km resolution begun substitute into formula (3), it is achieved the NO emissions reduction of soil moisture processes, and show that calendar year 2001 is extremely Resolution in 2013 is the soil moisture data of 1Km.

(3) step () is divided the Part II CCI data product obtained and be converted to dividing in 1979 to 2000 Resolution is the soil moisture data of 1Km:

Basic ideas are: by the resolution that step (two) obtains be 1Km NO emissions reduction soil moisture data (calendar year 2001- Data in 2013) substitute into public affairs with corresponding original 25Km CCI soil water data (data in-2013 years calendar year 2001s) point Formula (4), utilizes statistical relationship method between soil moisture different scale, obtains matching conversion coefficient a_x,yAnd b_x,yValue.

Between different scale, the relation of soil moisture is

${\mathrm{SSM}}_{x,y,t}^F=a_{x,y}\·{\mathrm{SSM}}_{x,y,t}^C+b_{x,y}---\left(4\right)$

Wherein,WithIt is expressed as t (x, y) high-resolution and corresponding low resolution thereof on position Soil moisture in rate pixel；a_x,yAnd b_x,yRepresent conversion coefficient.

May cause pixel edge that discontinuous " mosaic " effect occurs for eliminating spatial scaling, use geography to weight back Return (Geographically Weighted Regression, GWR) method first by the CCI soil water mark of 25Km resolution According to being resampled to 1Km.Spatially have gradual change feature due to soil moisture, GWR method is based on First Law of Geography, fully Consider the dependency between soil moisture in the range of different spaces, by certain pixel soil moisture by himself and adjacent picture elements Value weighted average obtains, power 1 weight coefficient use the method for falling distance function determine, thus avoid soil moisture spatially dash forward Cash as.

On this basis, utilize formula (4) by other time (Land Surface Parameters loss period, before calendar year 2001) CCI data Be converted to the data that spatial resolution is 1Km, that is to say and step () is divided the Part II CCI data (resolution obtained For 25Km) substitute in formula (4)The matching conversion coefficient a that will obtain in step (1)_x,yAnd b_x,yValue generation respectively Enter a in formula (4)_x,yAnd b_x,y, obtain the soil moisture that resolution is 1Km in Land Surface Parameters (MODIS data) loss period Data.

Here the ultimate principle of institute's foundation is: it is special that soil moisture has time stability (Temporal stability) Levy, i.e. the long timing variations of local soil moisture and region populations trend has good concordance.Briefly, by soil on a large scale The general impacts of earth moisture, in certain zonule, space soil moisture time change generally with on a large scale in soil water code insurance Hold consistent.This spatial scaling being soil moisture is laid a good foundation, according to this feature, and the relation of soil moisture between different scale Formula (4) has been given by.

In addition it should be pointed out that: in the present invention MODIS data can obtain the period be calendar year 2001 to 2013, when CCI is long The sequence remote sensing soil moisture data retrievable period is 1979 to 2013, and this is to be determined by this area current techniques present situation , but in the case of art technology levels to the corresponding data that can obtain beyond the above-mentioned two period, this The method of bright offer also is able to solve technical problem to be solved proposed by the invention.

Two, the method that said method the data obtained is carried out calibration and Checkout

The present invention finally evaluates the effectiveness of NO emissions reduction result in terms of precision and spatial accuracy two.In accuracy test Aspect, this method, according to the time stability feature (Temporal stability) of soil moisture, uses long sequential ground to see The soil moisture data of measuring point, by regression analysis and two kinds of method calibration remote sensing soil moistures of linear stretch.Regression analysis side Method describes the linear relationship between ground observation point and remote sensing soil moisture data, and expression formula is:

SM_cal=a × SM_RS+b (5)

In formula, SM_cal、SM_RSRepresent calibration and original remote sensing soil moisture data respectively；A and b is undetermined parameter, general root According to ground observation point and the corresponding remote sensing soil moisture data of long sequential, use method of least square to estimate, make two groups of numbers According to error minimum.

Linear stretch method then considers the practical situation of ground observation data variation feature, enters remote sensing soil moisture data Row stretch processing so that it is there is identical average and variance with ground observation data, its expression formula is:

${\mathrm{SM}}_{cal}=({\mathrm{SM}}_{RS}-{\stackrel{\‾}{SM}}_{RS})\×(s_{in}/s_{RS})+{\stackrel{\‾}{SM}}_{in}---\left(6\right)$

In formula,WithIt is respectively ground observation point and corresponding remote sensing long sequential soil moisture meansigma methods；s_inAnd s_RS The standard deviation of two kinds of data respectively.

For verify NO emissions reduction soil moisture data space feature accuracy, the present invention use TVDI index method checking with The dependency of soil moisture.This algorithm utilizes 1Km MODIS surface temperature and NDVI data, builds temperature/vegetation index empty Between, the accuracy of checking NO emissions reduction data space feature.For avoiding the discordance in different pieces of information dimension, two kinds of data are respectively It is normalized to 0-1 interval, represents the alternation degree of soil dry-wet situation.Use the coefficient of determination (Determination coefficient)R²The general status of reflection NO emissions reduction data space accuracy；Simultaneously take account of the geographical poor of the factors such as meteorology The opposite sex, selects representative section line as validation region, checks the Soil Under Conditions moisture NO emissions reduction result such as different latitude, meteorology.

Three, application example

As a example by Poyang Lake Basin, demonstrate the effect of this algorithm.Fig. 2 show 2001-2013 original CCI data with Based on Land Surface Parameters auxiliary scale-transformation method soil moisture NO emissions reduction Comparative result figure, (left figure is the original CCI of 2001-2013 Average data for many years, right figure is average data for many years after NO emissions reduction), it is found that soil moisture NO emissions reduction result and original CCI Data have stronger Space Consistency, show as basin many annuals soil moisture on a declining curve from north orientation south, NO emissions reduction Result preferably reflects the minutia of soil moisture spatial distribution.Original CCI data flow domain long-time average annual value is 0.350cm³/cm³, standard deviation is 0.035cm³/cm³, the many annuals in basin of NO emissions reduction soil moisture and standard deviation are respectively 0.346cm³/cm³And 0.008cm³/cm³, special heterogeneity more original CCI data decrease.Basin soil moisture exceeds most Now in lake region, original CCI data and NO emissions reduction result soil moisture are respectively 0.402cm³/cm³And 0.351cm³/cm³。

According to acquired spatial scaling coefficient a_x,yAnd b_x,y, use formula (4) to obtain the 1979-2013 1km soil water Divided data collection, its mean space distribution for many years is as it is shown on figure 3, basin average out to 0.350cm for many years³/cm³, standard deviation is 0.010cm³/cm³。

Utilize TVDI data, the accuracy of checking NO emissions reduction Soil Moisture Space Distribution Features.Found that two kinds of data Having stronger concordance on Watershed Scale, correlation coefficient is 0.45 (R²=0.20).Geography in view of factors such as meteorologies Diversity, the present invention selects two hatchings crossing over study area east-west direction and North and South direction as validation region, inspection respectively Test different latitude and soil moisture NO emissions reduction result (as shown in Figure 4) under meteorological condition.Fig. 5 shows fall chi on hatching direction Degree soil moisture and the distribution results of TVDI, wherein (a) part is east-west direction, and (b) part is North and South direction.It is found that Relative to original CCI data, after NO emissions reduction, soil moisture is preferably coincide the spatial distribution characteristic of soil moisture, coefficient R (R between 0.3～0.4²>0.1)。

Use linear regression method, utilize soil moisture after ground observation data calibration fall and inspection yardstick, result such as Fig. 6 Shown in.Statistical result shows, the NO emissions reduction soil moisture quite well variation tendency of ground observation data after calibration, relevant Coefficient and RMSE are respectively 0.65 (R²=0.42) and 0.044cm³/cm³, show that soil moisture NO emissions reduction obtains with calibration process Better effects.

Finally illustrating, above example is only in order to illustrate technical scheme and unrestricted, although with reference to relatively The present invention has been described in detail by good embodiment, it will be understood by those within the art that, can be to the skill of the present invention Art scheme is modified or equivalent, and without deviating from objective and the scope of technical solution of the present invention, it all should be contained at this In the middle of the right of invention.

Claims (2)

1. one kind long sequential remote sensing soil moisture NO emissions reduction method, it is characterised in that it comprises the following steps:

(1) divide according to acquisition situation sequential long to the CCI remote sensing soil moisture data product of MODIS data:

Long for CCI sequential remote sensing soil moisture data product is divided into Part I CCI data product and Part II CCI data Product；Described Part I CCI data product is that MODIS data can obtain the long sequential remote sensing soil water mark of the CCI in the period According to product；Described Part II CCI data product is the CCI long sequential remote sensing soil water mark in the MODIS shortage of data period According to product；

(2) shown that calendar year 2001 is the soil moisture data of 1Km to intrinsic resolution in 2013 by MODIS data:

(1) reception calendar year 2001 is the MODIS data of 1Km to intrinsic resolution in 2013, and described MODIS data include: vegetation index Data NDVI and surface temperature data T；

(2) data NDVI described in vegetation index and surface temperature data T are normalized, eliminate the difference in number of levels The opposite sex:

T^*=(T-T_min)/(T_max-T_min) (1)

NDVI^*=(NDVI-NDVI_min)/(NDVI_max-NDVI_min) (2)

Wherein, T_maxFor the maximum of surface temperature data T, T_minMinima for surface temperature data T；NDVI_maxFor normalization The maximum of vegetation index data NDVI, NDVI_minFor the minima of normalized differential vegetation index data NDVI, T^*After normalization Surface temperature data, NDVI^*For the vegetation index data after normalization；

(3) to surface temperature data T after normalization^*With vegetation index data NDVI after normalization^*Carry out being resampled to differentiate Rate is 25Km；

Based on surface temperature data T after the normalization in the calendar year 2001 to 2013 that resampling obtains^*With the vegetation after normalization Exponent data NDVI^*Set up the statistical relationship of first stage CCI soil moisture data:

$SSM=\underset{i=0}{\overset{n}{\Σ}}\underset{j=0}{\overset{n}{\Σ}}{a}_{ij}\·{\mathrm{NDVI}}^{*\left(i\right)}\·T^{*\left(j\right)}---\left(3\right)$

Wherein a_ijRepresenting weight coefficient, i, j represent the dimension of independent variable respectively, and SSM is the soil water in calendar year 2001 to 2013 Divided data；

(4) it is the surface temperature of 1Km by vegetation index data NDVI that resolution is 1Km and the resolution that obtain in step (1) Data T data substitute in formula (1), (2) and (3), it is achieved the NO emissions reduction of soil moisture processes, in drawing calendar year 2001 to 2013 The soil moisture data that resolution is 1Km；

(3) data obtained with step (two) are as reference data, obtain soil moisture spatial scaling parameter, and then by step (1) divide the Part II CCI data product that obtains and be converted to the soil moisture data that resolution is 1Km:

(1) matching conversion coefficient is determined:

Between different scale, the relation of soil moisture is

${\mathrm{SSM}}_{x,y,t}^F=a_{x,y}\·{\mathrm{SSM}}_{x,y,t}^C+b_{x,y}---\left(4\right)$

Wherein,WithIt is expressed as t (x, y) high-resolution and corresponding low resolution picture thereof on position Soil moisture in unit；a_x,yAnd b_x,yRepresent matching conversion coefficient；

The soil moisture data that resolution is 1Km in the calendar year 2001 to 2013 obtain step (two) substitute in formula (4)Step (one) divides the Part I CCI data obtained substitute in formula (4)Show that matching is changed Coefficient a_x,yAnd b_x,yValue；

(2) step () is divided the Part II CCI data obtained to substitute in formula (4)Step (1) will be obtained The matching conversion coefficient a arrived_x,yAnd b_x,yValue substitute into a in formula (4) respectively_x,yAnd b_x,y, obtain in 1979 to 2000 Resolution is the soil moisture data of 1Km.

One the most according to claim 1 long sequential remote sensing soil moisture NO emissions reduction method, it is characterised in that described It is that calendar year 2001 was to 2013 that MODIS data can obtain the period.