CN104462809A - Grassland productivity estimation method based on remote sensing and GIS (geographic information system) - Google Patents

Abstract

The invention relates to a grassland productivity estimation method based on remote sensing and a GIS (geographic information system). The method solves the defects that the traditional method is time-consuming and labor-consuming when estimating grassland productivity, and difficultly realizes large-range and long-term continuous acquisition. The method comprises the steps of: Step I, preprocessing grassland raw data, Step II, selecting an NDVI (normalized difference vegetation index), an RVI (ratio vegetation index), an MSAVI (modified soil adjusted vegetation index) and an EVI (enhanced vegetation index) to construct various grassland productivity estimation models based on the vegetation indices, Step III, selecting the optimal grassland productivity estimation model, Step IV, making grid operation by using the optimal grassland productivity estimation model and the vegetation indices to obtain a grassland productivity spatial distribution chart, estimating total grassland productivity according to total grassland area, and Step V, establishing a grassland productivity prediction model by taking the vegetation indices as independent variables and the grassland productivity in a future certain period as a dependent variable. The method is applied to the field of ecology and the remote sensing.

Description

A kind of Productivity evaluation method combined based on remote sensing and GIS

Technical field

The present invention relates to the Productivity evaluation method combined based on remote sensing and GIS.

Background technology

In recent years, the ecologic environment instability caused due to human social development obviously increases, thus caused a series of ecological environment problem, as pursuing economic benefit and excess reclamation and herding the Grassland Quality serious degradation caused, the problem such as capacity for raising livestock on the grasslands sharply declines, soil erosion is increasingly serious, bio-diversity is impaired.Therefore, rapid evaluation and future developing trend Accurate Prediction are carried out to grassland situation, of far-reaching significance to instructing region grassland science reasonably to utilize.

Remote sensing, the developing rapidly of GIS technology, for the various ecological environment problem of research earth surface provides good means.Especially, to the quick interpretation within the scope of large area space, compared with traditional microcosmic case study, there is the outstanding advantage saving human and material resources, financial resources and efficiently and accurately.Therefore, utilize macroscopic view and real-time remotely-sensed data, in conjunction with the Spatial Data Analysis means of GIS, explore region Grassland Quality assessment and predicting means, the improvement for region environment is significant.

Summary of the invention

The present invention be to solve that classic method estimation Productivity is consuming time, deficiency that effort and being difficult to realizes on a large scale, long-time continuous obtains, and provide a kind of Productivity evaluation method combined based on remote sensing and GIS.

The Productivity evaluation method combined based on remote sensing and GIS realizes according to the following steps:

Step one, meadow raw data pre-service:

By unified for meadow raw data under the same coordinate system and projection; Wherein, described in be projected as Albers projection, described the same coordinate system adopts east longitude 105 °, and described meadow raw data comprises Productivity data, vegetation index data, meadow spatial distribution data and land use data;

Step 2, choose normalized differential vegetation index NDVI, ratio vegetation index RVI, modified form soil adjustment vegetation index MSAVI and enhancement mode meta file EVI, build the polymorphic type Productivity appraising model based on many vegetation indexs;

Step 3, utilize the Productivity data same period of fieldwork to carry out precision test to the polymorphic type Productivity appraising model based on many vegetation indexs, choose optimum Productivity appraising model;

Step 4: under ArcGIS software, utilizes optimum Productivity appraising model and vegetation index, carries out grid computing, obtain Productivity spatial distribution map, and according to the meadow total area, estimation obtains Productivity total amount;

Step 5: with vegetation index in early stage, correlation analysis and regretional analysis are carried out to the Productivity data of fieldwork, optimum vegetation index and optimal fitting equation are determined in contrast, setting up with vegetation index is independent variable, with the Productivity forecasting model that following regular period Productivity is dependent variable, realize the prediction of Productivity.

Invention effect:

The present invention utilizes statistical regression methods, in conjunction with remotely-sensed data and Productivity measured data, relation between the tender Productivity of pine and differ ent vegetation index is discussed and analyzes, and then determine optimum vegetation index and the optimization model of Productivity estimation, and apply this model space inversion is carried out to Productivity.Carry out quantitative evaluation by reserved Productivity measured data to estimation result, analyze simultaneously with other people estimation Comparative result, find, the method utilizing vegetation index to realize Productivity estimation of the present invention's proposition is practicable.Simultaneously, prove that yield-power is the closest with NDVI relation in early stage by experiment, contrast all kinds of model accuracy, choose power index model based on NDVI as 16 days Optimal predictor models, and it is verified, the forecast precision obtaining this model is 74%, also demonstrates and utilizes following Productivity of the vegetation index in early stage prediction to possess scientific basis.More than utilize remotely-sensed data to carry out the method for Productivity shortcut estimation and forecast, can be applied in the middle of the practical application such as Pasture management and planning of science activities completely.

Accompanying drawing explanation

Fig. 1 is loose tender Productivity sampling sampling point distribution plan in an embodiment;

Fig. 2 is loose browse ground Productivity modelling precision test in an embodiment;

Fig. 3 is loose tender Productivity estimation result space distribution plan in an embodiment.

Embodiment

Embodiment one: the Productivity evaluation method combined based on remote sensing and GIS of present embodiment realizes according to the following steps:

Step one, meadow raw data pre-service:

By unified for meadow raw data under the same coordinate system and projection; Wherein, described in be projected as Albers projection, described the same coordinate system adopts east longitude 105 °, and described meadow raw data comprises Productivity data, vegetation index data, meadow spatial distribution data and land use data;

Step 2, choose normalized differential vegetation index NDVI, ratio vegetation index RVI, modified form soil adjustment vegetation index MSAVI and enhancement mode meta file EVI, build the polymorphic type Productivity appraising model based on many vegetation indexs;

Step 3, utilize the Productivity data same period of fieldwork to carry out precision test to the polymorphic type Productivity appraising model based on many vegetation indexs, choose optimum Productivity appraising model;

Step 4: under ArcGIS software, utilizes optimum Productivity appraising model and vegetation index, carries out grid computing, obtain Productivity spatial distribution map, and according to the meadow total area, estimation obtains Productivity total amount;

Step 5: with vegetation index in early stage, correlation analysis and regretional analysis are carried out to the Productivity data of fieldwork, optimum vegetation index and optimal fitting equation are determined in contrast, setting up with vegetation index is independent variable, with the Productivity forecasting model that following regular period Productivity is dependent variable, realize the prediction of Productivity.

Present embodiment effect:

Present embodiment utilizes statistical regression methods, in conjunction with remotely-sensed data and Productivity measured data, relation between the tender Productivity of pine and differ ent vegetation index is discussed and analyzes, and then determine optimum vegetation index and the optimization model of Productivity estimation, and apply this model space inversion is carried out to Productivity.Carry out quantitative evaluation by reserved Productivity measured data to estimation result, analyze simultaneously with other people estimation Comparative result, find, the method utilizing vegetation index to realize Productivity estimation of the present invention's proposition is practicable.Simultaneously, prove that yield-power is the closest with NDVI relation in early stage by experiment, contrast all kinds of model accuracy, choose power index model based on NDVI as 16 days Optimal predictor models, and it is verified, the forecast precision obtaining this model is 74%, also demonstrates and utilizes following Productivity of the vegetation index in early stage prediction to possess scientific basis.More than utilize remotely-sensed data to carry out the method for Productivity shortcut estimation and forecast, can be applied in the middle of the practical application such as Pasture management and planning of science activities completely.

Embodiment two: present embodiment and embodiment one unlike: in described step one, the concrete procurement process of Productivity data is:

The meadow ground biomass in Productivity maximum period is collected in field, records sampling point position with GPS.The inside, sample ground of each 10m × 10m arranges 3 or 5 subquadrats as repetition, and flush with ground to collect in each subquadrat green portion on the ground, dries to weight for 65 DEG C, the mean value of all subquadrat biomasss in various kinds ground and various kinds ground ground biomass.According to the ratio (root/shoot ratio) of meadow underground biomass and ground biomass, obtain meadow total biomass; By the conversion coefficient of biomass and carbon, obtain Productivity data.

Other step and parameter identical with embodiment one.

Embodiment three: present embodiment and embodiment one or two unlike:

One, the MOD13Q1 vegetation index data set (download network address: https: //wist.echo.nasa.gov) developed according to Unified Algorithm from NASA (National Aeronautics and Space Administration)/MODIS (Moderate Resolution Imaging Spectroradiometer) land product group of NDVI and EVI, RVI and MSAVI utilizes the ruddiness of vegetation index data centralization and near-infrared band reflectivity data to calculate, and specifically sees formula (1), (2), (3), (4):

$\mathrm{NDVI}=\frac{{\mathrm{\ρ}}_{\mathrm{NIR}}-{\mathrm{\ρ}}_R}{{\mathrm{\ρ}}_{\mathrm{NIR}}+{\mathrm{\ρ}}_R}---\left(1\right)$

$\mathrm{EVI}=\frac{2.5({\mathrm{\ρ}}_{\mathrm{NIR}}-{\mathrm{\ρ}}_R)}{{\mathrm{\ρ}}_{\mathrm{NIR}}+6.0{\mathrm{\ρ}}_R-7.5{\mathrm{\ρ}}_B+1}---\left(2\right)$

$\mathrm{RVI}=\frac{{\mathrm{\ρ}}_{\mathrm{NIR}}}{{\mathrm{\ρ}}_R}---\left(3\right)$

$\mathrm{MSAVI}=0.5\×[2{\mathrm{\ρ}}_{\mathrm{NIR}}+1-\sqrt{{({2\mathrm{\ρ}}_{\mathrm{NIR}}+1)}^2-8({\mathrm{\ρ}}_{\mathrm{NIR}}-{\mathrm{\ρ}}_R)}]---\left(4\right)$

In formula: ρ _nIRand ρ _rbe respectively the reflectivity of red spectral band and near-infrared band, ρ _bfor the reflectivity of blue wave band, calculate four vegetation indexs;

Two, random selecting sampled point Productivity data, reject sampled point Productivity data outliers, the mean value of all types of vegetation indexs of 3 × 3 pixels around sampled point is extracted centered by sampled point, as the vegetation index data of sampled point, utilize statistic software SPSS, simple linear regression analysis and curvilinear regression analysis are carried out to vegetation index data and Productivity data, the polymorphic type Productivity appraising model based on many vegetation indexs that to set up with all types of vegetation index be independent variable.

Random selecting part Productivity sampling number certificate, when there being multiple sampled point in a pixel, gets the mean value of these sampled point yield-power; Rejecting abnormalities point, extracts the vegetation index mean value of 3 × 3 pixels (750m × 750m) around this point centered by sampled point.

Other step and parameter identical with embodiment one or two.

Embodiment four: one of present embodiment and embodiment one to three unlike: described step 3 is specially:

One, Integrated comparative is based on the polymorphic type Productivity appraising model fitting precision of many vegetation indexs, finally selects optimum vegetation index and optimum Productivity appraising model:

Two, utilize the effective yield-power data having neither part nor lot in Productivity appraising model and set up, according to formula (5), (6) listed evaluation index, precision test carried out to optimum Productivity appraising model:

$\mathrm{RMSE}=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(Y_i-Y_i^{\′})}^2}{N}}---\left(5\right)$

$\mathrm{REE}=\sqrt{\frac{1/N\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(Y_i-Y_i^{\′})}^2}{\stackrel{\‾}{Y}}}\×100---\left(6\right)$

In formula: RMSE, mean absolute error; REE, average relative error (%); Y _i, the actual measurement Productivity (g/m of sample ground i ²); Y _i', the simulated sward yield-power (g/m of sample ground i ²); actual average Productivity (g/m ²); N, sample number.

Selected optimum Productivity appraising model evaluation precision is 78%, satisfies the demands, and may be used for Productivity estimation.Carrying out correlation analysis discovery to the Productivity data of fieldwork with vegetation index in early stage, there is significant correlationship in the two, demonstrates the theoretical foundation of carrying out Productivity forecast with vegetation index in early stage.Because vegetation index used is 16 days composite values, be therefore independent variable with vegetation index, with the Productivity of the 16th day for dependent variable grassland establishment yield-power forecasting model.Contrast all kinds of model accuracy and correlation analysis result, finally determine optimum Productivity forecasting model

(y=863.51NDVI ^0.89), utilize reserved Productivity measured data to verify this model, precision is 76%, illustrates that it is feasible for utilizing vegetation index to carry out Productivity forecast.

Other step and parameter identical with one of embodiment one to three.

Embodiment:

Technical solution of the present invention is specifically implemented to be described in conjunction with following instance, and the Productivity shortcut estimation and forecast method utilizing remote sensing and GIS to combine realizes space inversion and the prediction of Grassland in Songnen Plain yield-power.

Concrete operation step is as follows:

Step one, meadow raw data pre-service: the data related to comprise loose tender Productivity field sampling data (see Fig. 1); By unified for meadow raw data under the same coordinate system and projection; Wherein, described meadow raw data comprises Productivity data, vegetation index data, meadow spatial distribution data and land use data;

Described Productivity measured data gathers acquisition on the spot in yield-power maximum period, 142 sample ground are set altogether, each plot size is 10m × 10m, choose the subquadrat of 3 representational 1m × 1m therein, sampling point position is recorded with GPS, flush with ground collects the green portion of ground in each sample prescription, take back 65 DEG C, laboratory to dry to weight, the mean value of all subquadrat biomasss in various kinds ground and various kinds ground ground biomass, according to the ratio of meadow ground biomass and underground biomass, obtain Grassland Biomass; According to the ratio of meadow ground biomass and underground biomass, obtain Grassland Biomass; By the conversion coefficient of biomass and carbon, obtain Productivity data; The unit area Productivity on the mean value of 3 subquadrats and each sample ground; The 1:100 ten thousand Vegetation of China atlas that meadow spatial distribution data is worked out from Vegetation of China figure editorial board of the Chinese Academy of Sciences, on the basis that scanning, digitizing and attribute add, modify as a reference with the Northeast's land use data that Resources and environmental sciences data center of the Chinese Academy of Sciences provides, and shear, extract and obtain song-Nen plain all types of meadows spatial distribution data;

Described vegetation index data set is the MOD13Q1 that NASA/MODIS land product group is developed according to Unified Algorithm, and temporal resolution is 16 days, and spatial resolution is 250m; Utilize that MODIS professional treatment software MODIS ReprojectionTools (MRT) projects to the data downloaded, format conversion processing, be set to unified Albers projection; Choose normalization vegetation NDVI, ratio vegetation index RVI, modified form soil adjustment vegetation index MSAVI and enhancement mode meta file EVI tetra-vegetation indexs;

First normal latitude: 25 ° of N, second normal latitude: 47 ° of N, central meridian: 105 ° of E, geodetic level: WGS84, time is corresponding with the sampling time, the first phase product started for 2009 the 209th day, the time is on August 12nd, 1 2009 on July 28th, 2009, spatially covers the West of Songnen Plain shown in study area; 1: the 100 ten thousand Vegetation of China atlas that pine browse ground distributed data is worked out from Vegetation of China figure editorial board of the Chinese Academy of Sciences, on the basis that scanning, digitizing and attribute add, modify as a reference with the Northeast's land use data that Resources and environmental sciences data center of the Chinese Academy of Sciences provides, and shear, extract obtain loose browse ground distributed data;

Above data are all unified under the same coordinate system and projection, and what adopt is projected as Albers projection, and the central meridian adopting the whole nation unified, east longitude 105 °;

Step 2, choose all types of vegetation index and build based on the polymorphic type Productivity appraising model of many vegetation indexs:

According to formula (1) ~ (4), calculate four vegetation index data Layers; Random selecting part Productivity sampling number certificate, extracts all types of vegetation index mean values of 3 × 3 pixels (750m × 750m) around this point, as the vegetation index value of this point with sampled point data grid technology; Utilize statistic software SPSS, correlation analysis (see table 1) is carried out to vegetation index and Productivity, confirm the rationality of selected vegetation index; Simple linear regression analysis and curvilinear regression analysis are carried out to all types of vegetation index and Productivity data, set up the Productivity appraising model (see table 2) being independent variable with each vegetation index, contrast each modeling precision, select optimum simulation vegetation index NDVI, determine the optimum analogy model (y=e of Productivity ^{8.05 – 1.09/NDVI}(R=0.81)); Utilize the effective yield-power data having neither part nor lot in model and set up, according to formula (5), (6) listed evaluation index, carry out precision test (see Fig. 2) to this optimization model, simulation precision is up to 78%, prove that this model accuracy is reliable, analog result science is credible; Based on the optimization model after evaluation, utilize optimum vegetation index NDVI, inverting obtains loose tender Productivity and space distribution (see Fig. 3) thereof.

Table 1 Productivity and vegetation index correlationship are analyzed

Note: sample number is 70; * represents pole significant correlation (P<0.01)

Table 2 is based on the Productivity appraising model of vegetation index

Note: * * represents and extremely significantly checks (P<0.01) by 0.01 level

Step 3, Grassland in Songnen Plain yield-power are forecast

Satellite vegetation index data contain natural meadow production information, and the biomass in early stage is the basis of late growing stage, reflects following growth tendency, therefore usable satellite vegetation index forecast meadow future production.Productivity forecast is carried out to utilize 16 days vegetation index data at this.16 days Forecast Mode refer to the current satellite vegetation index forecast Productivity of following 16th day, therefore with the vegetation index of synthesis in 16 days for independent variable, with the Productivity of corresponding 16th day for dependent variable sets up linear prediction model y=863.51NDVI ^0.89;

The MOD13Q1 vegetation index data set that the vegetation index data utilized are developed according to Unified Algorithm for NASA/MODIS land product group, the time is 28 days ~ August 12 July in 2009, and disposal route is the same.Simple linear regression analysis and curvilinear regression analysis are carried out to all types of vegetation index and Productivity data, 16 days Productivity forecasting models (see table 3) that to set up with each vegetation index be independent variable.According to table 1 correlation analysis result, yield-power is the closest with NDVI relation in early stage, contrast all kinds of model accuracy, choose power index model based on NDVI as 16 days Optimal predictor models, and it is verified, the forecast precision obtaining this model is 74%, illustrates that the Productivity utilizing vegetation index to predict following 16 days is feasible.

Table 3 Grassland in Songnen Plain yield-power forecasting model

Note: * * represents and extremely significantly checks (P<0.01) by 0.01 level

Claims (4)

1., based on the Productivity evaluation method that remote sensing and GIS combines, it is characterized in that the Productivity evaluation method combined based on remote sensing and GIS realizes according to the following steps:

Step one, meadow raw data pre-service:

By unified for meadow raw data under the same coordinate system and projection; Wherein, described in be projected as Albers projection, described the same coordinate system adopts east longitude 105 °, and described meadow raw data comprises Productivity data, vegetation index data, meadow spatial distribution data and land use data;

Step 2, choose normalized differential vegetation index NDVI, ratio vegetation index RVI, modified form soil adjustment vegetation index MSAVI and enhancement mode meta file EVI, build the polymorphic type Productivity appraising model based on many vegetation indexs;

Step 3, utilize the Productivity data same period of fieldwork to carry out precision test to the polymorphic type Productivity appraising model based on many vegetation indexs, choose optimum Productivity appraising model;

Step 4: under ArcGIS software, utilizes optimum Productivity appraising model and vegetation index, carries out grid computing, obtain Productivity spatial distribution map, and according to the meadow total area, estimation obtains Productivity total amount;

Step 5: with vegetation index in early stage, correlation analysis and regretional analysis are carried out to the Productivity data of fieldwork, optimum vegetation index and optimal fitting equation are determined in contrast, setting up with vegetation index is independent variable, with the Productivity forecasting model that following regular period Productivity is dependent variable, realize the prediction of Productivity.

2. a kind of Productivity evaluation method combined based on remote sensing and GIS according to claim 1, is characterized in that in described step one, the concrete procurement process of Productivity data is:

The meadow ground biomass in Productivity maximum period is collected in field, records sampling point position with GPS.The inside, sample ground of each 10m × 10m arranges 3 or 5 subquadrats as repetition, and flush with ground to collect in each subquadrat green portion on the ground, dries to weight for 65 DEG C, the mean value of all subquadrat biomasss in various kinds ground and various kinds ground ground biomass.According to the ratio (root/shoot ratio) of meadow underground biomass and ground biomass, obtain meadow total biomass; By the conversion coefficient of biomass and carbon, obtain Productivity data.

3. a kind of Productivity evaluation method combined based on remote sensing and GIS according to claim 1 and 2, is characterized in that described step 2 is specially:

One, the MOD13Q1 vegetation index data set developed according to Unified Algorithm from NASA land product group of NDVI and EVI, RVI and MSAVI utilizes the ruddiness of vegetation index data centralization and near-infrared band reflectivity data to calculate, and specifically sees formula (1), (2), (3), (4):

$\mathrm{NDVI}=\frac{{\mathrm{\&rho;}}_{\mathrm{NIR}}-{\mathrm{\&rho;}}_R}{{\mathrm{\&rho;}}_{\mathrm{NIR}}+{\mathrm{\&rho;}}_R}---\left(1\right)$

$\mathrm{EVI}=\frac{2.5({\mathrm{\&rho;}}_{\mathrm{NIR}}-{\mathrm{\&rho;}}_R)}{{\mathrm{\&rho;}}_{\mathrm{NIR}}+6.0{\mathrm{\&rho;}}_R-7.5{\mathrm{\&rho;}}_B+1}---\left(2\right)$

$\mathrm{RVI}=\frac{{\mathrm{\&rho;}}_{\mathrm{NIR}}}{{\mathrm{\&rho;}}_R}---\left(3\right)$

$\mathrm{MSAVI}=0.5\&times;[{2\mathrm{\&rho;}}_{\mathrm{NIR}}+1-\sqrt{{({2\mathrm{\&rho;}}_{\mathrm{NIR}}+1)}^2-8({\mathrm{\&rho;}}_{\mathrm{NIR}}-{\mathrm{\&rho;}}_R)}]---\left(4\right)$

In formula: ρ _nIRand ρ _rbe respectively the reflectivity of red spectral band and near-infrared band, ρ _bfor the reflectivity of blue wave band, calculate four vegetation indexs;

Two, random selecting sampled point Productivity data, reject sampled point Productivity data outliers, the mean value of all types of vegetation indexs of 3 × 3 pixels around sampled point is extracted centered by sampled point, as the vegetation index data of sampled point, utilize statistic software SPSS, simple linear regression analysis and curvilinear regression analysis are carried out to vegetation index data and Productivity data, the polymorphic type Productivity appraising model based on many vegetation indexs that to set up with all types of vegetation index be independent variable.

4. a kind of Productivity evaluation method combined based on remote sensing and GIS according to claim 3, is characterized in that described step 3 is specially:

One, Integrated comparative is based on the polymorphic type Productivity appraising model fitting precision of many vegetation indexs, finally selects optimum vegetation index and optimum Productivity appraising model:

Two, utilize the effective yield-power data having neither part nor lot in Productivity appraising model and set up, according to formula (5), (6) listed evaluation index, precision test carried out to optimum Productivity appraising model:

$\mathrm{RMSE}=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(Y_i-Y_i^{\&prime;})}^2}{N}}---\left(5\right)$

$\mathrm{REE}=\frac{\sqrt{1/N\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(Y_i-Y_i^{\&prime;})}^2}}{\stackrel{\&OverBar;}{Y}}\&times;100---\left(6\right)$

In formula: RMSE, mean absolute error; REE, average relative error (%); Y _i, the actual measurement Productivity (g/m of sample ground i ²); Y _i', the simulated sward yield-power (g/m of sample ground i ²); actual average Productivity (g/m ²); N, sample number.