CN1924611A - Land deterioration (desert) evaluation parameter remote control inversion and supervision technique method - Google Patents

Abstract

This invention discloses one earth degradation remote evaluation parameter feedback and monitor technique method, which comprises the following steps: selecting and evaluating the remote parameters for remote data earth degradation and establishing earth degradation monitor evaluation index system; rehearsing national degradation remote parameters; degradation formula evaluation and realization; using plant starting affair analysis plant and non-plant change; using NDVI various to analyze dynamic situations.

Description

Land deterioration (desertification) evaluation parameter remote inverting and monitoring technology method

Technical field

The present invention relates to the application technology of message area, relate in particular to and utilize the obtained information of remote sensing technology that the application technology of decision information is provided for land deterioration (desertification) evaluation and monitoring and resource environment Sustainable Development Planning.

Background technology

The parameter that needs to obtain multiple face of land natural conditions in land deterioration research comprises: soil covering, surface temperature, soil moisture and vegetation parameter etc.These parameters can be obtained with on-site inspection by research station observations such as conventional meteorology, the hydrology and farmlands and obtain.But the data of just putting by actual observation is obtained can't guarantee though utilize some polymerizations the observed result of point can be expanded to its precision of zone, can not represent the spatial distribution characteristic in these big zones of parameter.

The developing into to address this problem of remote sensing technology provides important means, and the result who a little goes up observation is generalized to becomes possibility on the face.According to of the reaction of different surface parameter, can obtain specific face of land parameter with different mathematical models and algorithm to the remote sensing wave band.

Although utilize remote sensing technology to do the research of a large amount of land deterioration evaluatings in the world, adopt the parameter of remote-sensing inversion to have only 1-2.Development along with multispectral romote sensing technology, utilize remote sensing technology inverting land deterioration evaluating as far as possible, and specific aim China physical feature, set up the model that is suitable for, realize whole nation land deterioration (desertification) evaluating inverting on a large scale, the monitoring of carrying out national land deterioration is fast had important scientific meaning and practical value.

Summary of the invention

Therefore, the object of the present invention is to provide a kind of application process of remote sensing technology, can select suitable spectrum parameter, set up data model, remotely-sensed data is handled, to extract the information of land deterioration and desertification by the data that remote sensing technology is obtained.

For achieving the above object, the present invention proposes a kind of land deterioration and desertification remote-sensing monitoring method, comprising: the foundation of desertification remote sensing monitoring comprehensive index system; The inverting of desertification remote sensing monitoring parameter; Determining of desertification degree remote sensing evaluation index; Desertification remote sensing classification; Desertification dynamic change monitoring.

Description of drawings

Fig. 1 is the classification chart of different desertification watch index combinations;

Fig. 2 is a China calendar year 2001 growth season MSAVI aggregate-value distribution plan;

Fig. 3 is China's calendar year 2001 growth season average land surface temperature distribution plan;

Fig. 4 is a NDVI-Ts feature space reduced graph;

Fig. 5 is a China calendar year 2001 growth season average T VDI distribution plan;

Fig. 6 is an average albedo distribution plan of China's growth season calendar year 2001;

Fig. 7 is a maximum vegetation coverage distribution plan of China's calendar year 2001;

Fig. 8 is nineteen ninety-five Chinese soil desertification present situation figure;

Fig. 9 is calendar year 2001 Chinese soil desertification present situation figure;

Figure 10 is DESERTIFICATION IN CHINA soil dynamic change figure;

Figure 11 is the desert boundary graph of the Central Asia and Chinese arid and semi-arid district different year;

Figure 12 is the non-vegetation distribution plan of Central Asia vegetation;

Figure 13 is the 1982-2000 Central Asia and Chinese arid and semi-arid district vegetation rudiment event distribution frequency plot;

Figure 14 is the cov distribution plan of the Central Asia and Chinese arid and semi-arid district different year.

Embodiment

In preferred case study on implementation of the present invention, at first set up desertification remote sensing monitoring comprehensive index system, comprise the selection of land deterioration (desertification) remote sensing parameter and determining of evaluation and desertification level index system.

The selection and the evaluation of 1) land deterioration (desertification) remote sensing parameter

Because land deterioration occurs in arid, semiarid and inferior moistening arid biogeographic zone basically, these regional areas are vast, and natural conditions are abominable, have inconvenient traffic, and therefore, just utilize remote sensing technology to carry out research work to land deterioration as far back as the eighties in the world.But since J.Tucker set up NOAA AVHRR vegetation index (NDVI) desertification in Africa was studied the eighties, utilized remote sensing technology that the research of desertification is also rested on one, two indexs such as adopting NDVI basically in the world and estimated.

Development along with multispectral romote sensing technology, some desert characteristic parameter can utilize the remotely-sensed data inverting to come out, and the index that the present invention chooses has: albedo (Albedo), top temperature (LST) and the soil moisture (TVDI) of the vegetation coverage (FVC) of reflection desertification soil natural quality and ecologic regime and modified soil adjustment vegetation index (MSAVI) and reflection desertification soil physical attribute is totally 5 indexs.

For guaranteeing the monitoring accuracy and the irreplaceability of used remote sensing monitoring index, the monitoring accuracy of desertification watch index is estimated.With the Horqin sand ground is subjects, in 5 indexs selecting, use identical training sample, be the single index that obtains of 1 kilometer NOAA-AVHRR data inversion and the combination of a plurality of indexs with the resolution of nineteen ninety-five respectively, carry out the check of Horqin sand ground desertification watch precision, choose the combination of only index or index.

The used concrete data of each index of desertification watch are respectively maximum vegetation coverage of year, growth season (the 4-10 month) accumulated value of MSAVI, the albedo mean value in growth season, the land surface temperature mean value in growth season, soil moisture (TVDI) mean value in growth season in area, Horqin.

Use identical training sample data in the classification, obtain the numerical range of each desertification remote sensing monitoring index, use the classification of decision tree classification device at different desertification intensity grades.But use the desertification watch index of different numbers and index combination to classify respectively.The evaluation of classification results is that index is carried out with overall nicety of grading.Overall nicety of grading is meant that the correct pixel data of classification count the number percent of gained divided by total pixel, and total pixel is meant all pixels in the training sample, comprises the pixel that classification is correct.The concrete combination and the nicety of grading of used desertification remote sensing monitoring index see Table 1.

The overall nicety of grading table of the desertification degree monitoring of the different monitoring index combinations of table 1

An index	MSAVI	Albedo	LST		FVC	TVDI
							Nicety of grading	68.91％	56.24％	43.56％		66.84％	30.57％
Two indexs	MSAVI+Albedo	MSAVI+LST	MSAVI+TVDI		FVC+Albedo	FVC+LST
							Nicety of grading	86.59％	84.63％	77.65％		85.31％	82.49％
Three indexs	MSAVI+Albedo +LST	MSAVI+TVDI +Albedo	MSAVI+LST +TVDI		Albedo+TVDI +FVC	Albedo+LST +FVC
							Nicety of grading	90.23％	89.18％	86.84％		88.52％	87.67％
Four indexs	MSAVI+Albedo+LST+TVDI			FVC+Albedo+LST+TVDI
							Nicety of grading	93.83％			92.05％
Five indexs	MSAVI+Albedo+LST+TVDI+FVC
							Nicety of grading	95.21％

As can be seen from Table 1, use a desertification watch index to carry out the branch time-like separately, the nicety of grading of MSAVI is the highest, reach 68.91% (Fig. 1), next is vegetation coverage and albedo, nicety of grading is respectively 66.84% and 56.24%, and that nicety of grading is minimum is TVDI, and precision only is 30.57%.When using two monitoring indexes to divide time-like, the assembled classification precision of MSAVI+Albedo is the highest, reaches 86.59% (Fig. 1), and the precision when only using an index is significantly improved.When using three index classifications, the assembled classification precision of MSAVI+Albedo+LST is the highest, is 90.23% (Fig. 1).And use four classification indicators to divide time-like, though the nicety of grading height (Fig. 1) that the MSAVI+Albedo+LST+TVDI combination is made up than FVC+Albedo+LST+TVDI, both are more or less the same.When five monitoring indexes all participate in the branch time-like, nicety of grading reaches 95.21%, precision the highest (Fig. 1) in used combination.As seen, in desertification watch, index of every increase, the desertification information of reflection increases thereupon, and the precision of desertification watch is also improved greatly.From the result that uses an index classification also as can be seen, the nicety of grading of MSAVI is the highest, illustrates that the desertification information that MSAVI reflects is maximum, to the contribution maximum of desertification remote sensing monitoring.

2) desertification level index system determines

In the invention process case, utilize remote sensing technology to determine the evaluation index of desertification degree, its step is as follows:

Used data be nineteen ninety-five the China concilliation panel that prevents and controls desertification organize Ministry of Forestry investigation planning and design institute, the China forest-science academy, Ministry of Forestry northwest investigation planning and design institute, the Chinese Academy of Sciences combines and examines committee, units such as Beijing Forestry University are with national desert, Gobi desert and sandy land generaI investigation are main foundation, 1: 100 ten thousand DESERTIFICATION IN CHINA soil distribution status figure of establishment, and 5 desertification remote sensing monitoring indexs that go out by nineteen ninety-five NOAA-AVHRR data inversion, i.e. nineteen ninety-five growth season MSAVI accumulated value, the growth season average albedo, growth season average land surface temperature, year maximum vegetation coverage and growth season average T VDI value.

Definite method of index system is, at first with the desertification degree sample figure that chooses, it is the Chinese soil desertification figure digitizing of the nineteen ninety-five of desertification watch center, Chinese forest-science academy drafting, the grid map of each desertification remote sensing monitoring index that digitized sample figure and NOAA-AVHRR are finally inversed by carries out the GIS overlay analysis then, add up each desired value of different desertification degree types, according to the numerical value of the most of pixels of adding up of different desertification types, determine NOAA-AVHRR desertification watch index system.According to the index system after determining, use the NOAA-AVHRR data, just can monitor the China of nineteen ninety-five and the desertification situation of Central Asia.

NOAA-AVHRR desertification watch index system sees Table 2,3,4 and 5.

The NOAA desertification watch index system of the inferior moistening arid biogeographic zone of table 2

The desertification degree	MSAVI	Vegetation coverage	Albedo	The land surface temperature	TVDI
						Non-desertification	＞3.0	＞0.8	＜180	＜29	＜0.39
Slightly	2.7-3.0	0.7-08	180-220	29-32	0.39-0.51
						Moderate	2.2-2.7	0.53-0.7	220-250	32-35	0.51-0.60
Severe	1.9-2.2	0.4-0.53	250-280	35-38	0.60-0.70
						Utmost point severe	＜1.9	＜0.40	＞280	＞38	＞0.70

The NOAA desertification watch index system of table 3 semiarid region

The desertification degree	MSAVI	Vegetation coverage	Albedo	The land surface temperature	TVDI
						Non-desertification	＞2.2	＞0.6	＜200	＜30	＜0.35
Slightly	1.9-2.2	0.3-0.6	200-220	30-36	0.35-0.45
						Moderate	1.2-1.9	0.23-0.3	220-260	36-40	0.45-0.56
Severe	0.8-1.2	0.15-0.23	260-300	40-46	0.56-0.65
						Utmost point severe	＜0.8	＜0.15	＞300	＞46	＞0.65

The NOAA desertification watch index system of table 4 arid biogeographic zone

The desertification degree	MSAVI	Vegetation coverage	Albedo	The land surface temperature	TVDI
						Non-desertification	＞1.2	＞0.54	＜210	＜37	＜0.5
Slightly	0.8-1.2	0.15-0.54	210-230	37-41	05-0.6
						Moderate	0.4-0.8	0.08-0.15	230-250	41-49	0.6-0.68
Severe	0.1-0.4	0.03-0.08	250-300	49-53	0.68-0.78
						Utmost point severe	＜0.1	＜0.03	＞300	＞53	＞0.78

The NOAA desertification watch index system in the high and cold district of table 5

The desertification degree	MSAVI	Vegetation coverage	Albedo	The land surface temperature	TVDI
						Non-desertification	＞1.3	＞0.24	＜230	＜30	＜0.49
Slightly	1.1-1.3	0.12-0.24	230-255	30-33	0.49-0.56
						Moderate	0.7-1.1	0.09-0.12	255-290	33-37	0.56-0.65
Severe	0.4-0.7	0.04-0.09	290-320	37-40	0.65-0.71
						Utmost point severe	＜0.4	＜0.04	＞320	＞45	＞0.71

Calendar year 2001 the present invention has used the MODIS data to carry out the desertification remote sensing monitoring.But calendar year 2001 does not have suitable sample data determining in order to MODIS monitoring index system, and because the image feature of MODIS and NOAA-AVHRR data exists than big-difference, and each desertification watch index is in computation process, the method therefor difference, therefore using the desertification watch index system identical with NOAA-AVHRR, also is irrational.In invention, MODIS and NOAA-AVHRR are carried out regretional analysis in the same index of the same space position, as can be seen, the MODIS data of the NOAA-AVHRR of nineteen ninety-five and calendar year 2001 have stronger correlationship (table 6).Therefore, utilize this correlationship can calculate the pairing desertification index system of MODIS (table 7,8,9 and 10) according to the desertification watch index system that NOAA-AVHRR sets up.According to the desertification watch index system of being set up, utilize the MODIS data of calendar year 2001, promptly can carry out the desertification earth monitor of China and Central Asia.

The correlationship of table 6NOAA and MODIS desertification watch index

Monitoring index	Inferior moistening arid biogeographic zone		Semiarid region		Arid biogeographic zone		High and cold district
									Relational expression	Related coefficient	Relational expression	Related coefficient	Relational expression	Related coefficient	Relational expression	Related coefficient
	MSAVI	Y＝0.46*X+1.06	0.94	Y＝0.66*X-0.12	0.943	Y＝0.83*X-0.003	0.96	Y＝0.59*X+0.02									0.93
FVC									Y＝0.77*X+0.06	0.927	Y＝0.84*X-0.004	0.92	Y＝0.24*X+1.02	0.96	Y＝0.92*X+0.05	0.85
	Albedo	Y＝1.03*X-19.01	0.93	Y＝1.03*X-23.89	0.91	Y＝0.51*X+83.78	0.87	Y＝0.81*X+2.63									0.87
LST									Y＝0.35*X+16.19	0.77	Y＝0.54*X+9.52	0.83	Y＝0.32*X+17.87	0.78	Y＝0.91*X-12.91	0.85
	TVDI	Y＝0.65*X+0.48	0.78	Y＝0.77*X+0.25	0.82	Y＝0.56*X+0.23	0.8	Y＝1.19*X-0.15									0.93

The MODIS desertification watch index system of the inferior moistening arid biogeographic zone of table 7

The desertification degree	MSAVI	Vegetation coverage	Albedo	The land surface temperature	TVDI
						Non-desertification	＞1.9	＞0.70	＜170	＜29	＜0.43
Slightly	1.65-1.9	0.55-0.70	170-200	29-32	0.43-0.56
						Moderate	1.45-1.65	0.45-0.55	200-220	32-34	0.56-0.67
Severe	1.25-1.45	0.35-0.45	220-240	34-36	0.67-0.76
						Utmost point severe	＜1.25	＜0.35	＞240	＞36	＞0.76

The MODIS desertification watch index system of table 8 semiarid region

The desertification degree	MSAVI	Vegetation coverage	Albedo	The land surface temperature	TVDI
						Non-desertification	＞1.50	＞0.6	＜180	＜29	＜0.54
Slightly	1.20-1.50	0.43-0.6	180-205	29-32	0.54-0.60
						Moderate	0.90-1.20	0.31-0.43	205-230	32-34	0.60-0.69
Severe	0.60-0.90	0.1-0.31	230-250	34-37	0.69-0.72
						Utmost point severe	＜0.60	＜0.1	＞250	＞37	＞0.72

The MODIS desertification watch index system of table 3-9 arid biogeographic zone

The desertification degree	MSAVI	Vegetation coverage	Albedo	The land surface temperature	TVDI
						Non-desertification	＞1.10	＞0.40	＜180	＜30	＜0.53
Slightly	0.80-1.10	0.20-0.40	180-200	30-35	0.53-0.58
						Moderate	0.50-0.80	0.20-0.32	200-250	33-35	0.58-0.61
Severe	0.40-0.50	0.08-0.20	250-265	35-37	0.61-0.68
						Utmost point severe	＜0.40	＜0.08	＞265	＞37	＞0.68

The MODIS desertification watch index system in the high and cold district of table 3-10

The desertification degree	MSAVI	Vegetation coverage	Albedo	The land surface temperature	TVDI
						Non-desertification	＞0.80	＞0.45	＜180	＜21	＜0.4
Slightly	0.70-0.80	0.34-0.45	180-200	21-25	0.4-0.55
						Moderate	0.50-0.70	0.21-0.34	200-220	25-29	0.55-0.67
Severe	0.30-0.50	0.07-0.21	220-260	29-33	0.67
						Utmost point severe	＜0.30	＜0.07	＞260	＞33	＞0.71

Preferably, the present invention has set up five kinds of land deteriorations (desertification) evaluation parameter remote inversion algorithm model.

1) modified soil is adjusted vegetation index (MSAVI)

The present invention selects for use MSAVI as the desertification watch index, is because MSAVI can eliminate the influence of Soil Background as far as possible, does not also need artificially to determine parameter, and is more convenient objective.

MSAVI (Modified Soil-Adjusted Vegetation Index) regulates vegetation index for modified soil, and the computing formula of this index is:

$\mathrm{MSAVI}=(2\mathrm{NIR}+1-\sqrt{{(2\mathrm{NIR}+1)}^2-8(\mathrm{NIR}-R)})/2---\left(1\right)$

SAVI＝[(NIR-R)/(NIR+R+L)]×(L+1) (2)

R is meant visible light wave range, and NIR is meant near-infrared band.

Obtained Chinese calendar year 2001 growth season MSAVI aggregate-value distribution plan (Fig. 2) as calculated.

2) top temperature

Use the land surface temperature in algorithm computation per ten days of China of Becker and Li (1990).Concrete steps are:

● calculate normalized differential vegetation index NDVI

NDVI＝(Ch2-Ch1)/(Ch2+Ch1) (3)

Wherein, Ch1 is a visible channel, and Ch2 is the near infrared passage.The value of NDVI is between-1 to+1 scope.

● calculate face of land emissivity ε

Josef et al. (1997) is in existing work (Griend and Owe, 1993; Salisbury, 1994) on the basis, calculate the ε of AVHRR ₄With Δ ε, equation is:

ε ₄＝0.9897+0.029ln(NDVI) (4)

Δε＝ε ₄-ε ₅＝0.01019+0.01344ln(NDVI) (5)

The present invention uses formula (4) and (5) to calculate the emissivity ε of subrane ₄And ε ₅

● calculate land surface temperature Ts

Originally deliver the land surface temperature of per ten days that the division window algorithm that adopts Becker and Li (1990) calculates China and Central Asia:

T _s＝1.274+(T ₄+T ₅)/2{1+[0.15616(1-ε)/ε]-0.482(Δε/ε ²)} (6)

+(T ₄-T ₅)/2{6.26+[3.98(1-ε)/ε]+38.33(Δε/ε ²)}

Wherein, ε=(ε ₄+ ε ₅)/2, Δ ε=(ε ₄-ε ₅).

After calculating the land surface temperature in per ten days, be averaged in per three ten days and try to achieve the growth temperature of season (the 4-10 month) every month, then to growth season 7 totally months land surface temperature ask on average, obtain the grow average land surface temperature value in season of China, see Fig. 3.

3) soil moisture and damage caused by a drought

Utilize the NDVI-TS feature space of simplifying to propose the water stress index according to (2002) such as Sandholt, it is temperature vegetation damage caused by a drought index, at the feature space of simplifying, the limit (Ts-min) of will wetting is treated to the straight line parallel with the NDVI axle, and non-irrigated limit (Ts-max) and NDVI are linear.The NDVI-Ts feature space of this simplification is seen (Fig. 4).

The simplification of NDVI-Ts feature space is treated to triangle, China is divided into 3 agroclimatic regions has carried out the damage caused by a drought monitoring respectively.According to the per 8 days TVDI value that calculates, the mean value in the long season of seeking survival (the 4-10 month) obtains the soil moisture index of the average T VDI value in Chinese growth season as desertification watch respectively.

$\mathrm{TVDI}=\frac{[T_s-(a_1+b_1*\mathrm{NDVI})]}{[(a_2+b_2*\mathrm{NDVI})-(a_1+b_1*\mathrm{NDVI})]}---\left(7\right)$

In the formula, a ₁, b ₁, a ₂, and b ₂It is respectively the coefficient of non-irrigated limit and wet limit fit equation.

From the TVDI distribution plan (5) of China as can be seen, the TVDI value of Desert Area is bigger, and the TVDI value of Qinghai-Tibet Platean is relatively low.The TVDI value is big more, illustrates that the soil water content in this district is few more, otherwise, illustrate that soil water content is bigger.

4) calculating of albedo

Adopt the albedo of the Chinese calendar year 2001 of Saunders (1990) method calculating.After at first calculating per 10 days synthetic albedos, the albedo mean value of getting per three ten days obtains a moon albedo, then to growth season (the 4-10 month) 7 totally months albedo ask on average, obtain China's average albedo distribution plan of season (Fig. 6) of growing respectively.

5) vegetation coverage

In MODIS (NOAA-AVHRR) the normalized differential vegetation index NDVI model running, the information of a grid is made up of by the weighted mean of area exposed soil and vegetation.The information φ of same MODIS (NOAA-AVHRR) normalized differential vegetation index NDVI each pixel that sensor measured via satellite just can be expressed as the information contributed by vegetation region and the weighted sum of the information contributed by the exposed soil zone.Be the weighted mean that the NDVI value of each pixel in the image can be regarded NDVI with the NDVI of no vegetation cover part of vegetation cover part as, the i.e. vegetation coverage σ of pixel for this reason of weight of the NDVI of vegetation cover part is wherein arranged _v, and the weight of not having a NDVI of vegetation cover part is (1-σ _v).

Φ＝Φ _vσ _v+(1-σ _v)Φ _s (8)

In the formula (8), subscript v and s represent to have fully the value in vegetation-covered area and exposed soil district respectively.

Formula (8) is applied directly among the NDVI, and the simplest expression formula (9) that can obtain vegetation coverage is:

${\mathrm{\σ}}_v=\frac{\mathrm{NDVI}-{\mathrm{NDVI}}_s}{{\mathrm{NDVI}}_v-{\mathrm{NDVI}}_s}---\left(9\right)$

Because visible light R and near-infrared band NIR are ratio for the contribution of energy stream, therefore formula (9) and NDVI algorithm are combined, obtain formula (10):

${\mathrm{\σ}}_v=\frac{\mathrm{NDVI}-{\mathrm{NDVI}}_s}{{\mathrm{NDVI}}_v-{\mathrm{NDVI}}_s+(1-a)(\mathrm{NDVI}-{\mathrm{NDVI}}_v)}---\left(10\right)$

In the formula (10), a=(R+NIR) v/ (R+NIR) s.When a=1, then formula (10) becomes linear formula (9).

Another nonlinear formula (Choudhury et al.1994; Baret et al.1995; Wittich 1997) be

${\mathrm{\σ}}_v=1-{\left(\frac{{\mathrm{NDVI}}_v-\mathrm{NDVI}}{{\mathrm{NDVI}}_v-{\mathrm{NDVI}}_s}\right)}^b---\left(11\right)$

When the vegetation coverage of estimation large scale, because therefore the shortage of ground data uses the method based on linear relationship to be more suitable for than other comparatively complicated method.The present invention estimates that the formula of Chinese vegetation coverage is as follows:

${\mathrm{\σ}}_v=\frac{N_{p,\max }-N_s}{N_{c,v}-N_s}---\left(12\right)$

In the formula, N _{P, max}Be meant the year maximum NDVI value of this pixel; N _{C, v}The vegetation coverage that is meant every class soil cover type is 100% o'clock corresponding pixel NDVI value; N _sBe meant the NDVI minimum value of every class soil cover type.

Because inevitably exist The noise in the image, it may produce low or too high NDVI value, only can obtain wrong result certainly if calculate vegetation coverage with such value.Take place for fear of such mistake, at definite N _{C, v}And N _sThe time, should determine a degree of confidence in principle, make the NDVI probability distribution of each pixel set of every kind of soil cover type in the image, calculate the maximal value N in the fiducial interval _{C, v}With minimum value N _sN _sShould not change in time, for the bare area surface of most of types, in theory should be near zero, general value 0.05.N for other soil cover type of IGBP _s, have littler NDVI value in winter, and in summer, because cloud pollutes and the influence of atmosphere, N _sValue have bigger uncertainty.Therefore, each IGBP soil cover type is determined a N respectively _sValue is unpractiaca, and the present invention is when the estimation vegetation coverage, the N of all soil cover types _sAll value is 0.05.

People such as Zeng estimated global vegetation coverage in 1992.

The present invention utilizes the MODIS data respectively different soils cover type to be done the cumulative frequency diagram of NDVI according to the degree of confidence that Zeng determines, determines the N of different soils cover type then _{C, v}(table 11).NDVI maximal value image and the land cover classification figure of calendar year 2001 of calendar year 2001 are synthesized to a file, and each pixel to file covers classification according to soil under it then, finds and show corresponding N in (11) _{C, v}Value is the NDVI value of pixel, N _{C, v}And N _sSubstitution formula (12) is calculated the vegetation coverage of the Central Asia and Chinese calendar year 2001.Fig. 7 is China's maximum vegetation coverage distribution plan of calendar year 2001.As can be seen from Figure 7, by the north-westward southeast, vegetation coverage presents gradually the trend that increases, and is the boundary line with the staggered band of agriculture and animal husbandry, to the west of vegetation coverage be significantly less than to the east of the zone.

The present invention mainly utilizes the decision tree classification device that the desertification and the dynamic change of China are studied, and detailed step is discussed below:

In order to compare, the present invention has adopted unsupervised classification respectively, three kinds of sorters of maximum likelihood method in the supervised classification (MLC) and decision tree, with the Horqin sand ground is example, use five desertification remote sensing monitoring indexs of the MODIS data inversion of calendar year 2001, choosing under the identical desertification degree training sample situation, carrying out the classification of desertification degree, realizing the desertification remote sensing monitoring of China and Central Asia so that choose the best sorter of nicety of grading.

After classification is finished, 500 sampling points of picked at random, the desertification watch figure in conjunction with based on Horqin sand ground end of the nineties of TM image carries out precision evaluation to three kinds of classification results, and nicety of grading the results are shown in Table (12).

As can be seen from Table 12, in the overall nicety of grading of three kinds of sorters, the result is the highest for decision tree classification, and nicety of grading is 94.63%; Next is the maximum likelihood classification method, and nicety of grading is 88.59%; The nicety of grading minimum be non-supervision law, precision is 80.88%.The decision tree classification method exceeds 6.04% and 13.75% respectively than the nicety of grading of maximum likelihood and unsupervised classification.

For more scientifically estimating the performance of three kinds of sorters, the present invention adopts the Kappa coefficient to carry out the comparison of nicety of grading simultaneously.The Kappa statistics has more statistical discrimination power on the nicety of grading of estimating different sorters.From last column of table 12 as can be seen, the Kappa coefficient comparative result of three kinds of sorters is identical with the nicety of grading comparative result, and what promptly nicety of grading was the highest is the decision tree classification device.Therefore, according to The above results, the present invention will adopt the decision tree classification device to carry out the classification of desertification degree.

Table 11 is used for the Nc of each soil cover type of MODIS vegetation coverage monitoring, v

IGBP land cover		％	N c，v
				1	Evergreen coniferous forest	75％	0.89
2	Evergreen broadleaf forest	75％	0.90
				3	The fallen leaves coniferous forest	75％	0.86
4	Deciduous broad-leaved forest	75％	0.88
				5	Mixed forest	75％	0.89
6	Close filling	90％	0.88
				7	Dredge and irritate	90％	0.62
8	Wooden rare tree prairie	75％	0.85
				9	Rare tree prairie	75％	0.56
10	The meadow	75％	0.48
				11	Permanent wetland	75％	0.78
12	The farmland	75％	0.84
				13	Urban land	90％	0.88
14	The farmland natural vegetation mixes	75％	0.81
				15	Ice and snow
16	Bare area	90％	0.60
				17	Water body

The desertification degree nicety of grading table of the different sorters of table 12

	Unsupervised classification	Maximum likelihood classification	Decision tree classification
				Non-desertification	81.03％	87.64％	96.15％
Slightly	80.67％	89.75％	91.47％
				Moderate	75.01％	71.63％	89.30％
Severe	69.41％	85.72％	93.69％
				Utmost point severe	85.56％	90.63％	98.03％
Overall nicety of grading	80.88％	88.59％	94.63％
				The Kappa coefficient	0.735	0.802	0.885

Table 13 DESERTIFICATION IN CHINA earth monitor result relatively

	The State Administration of Forestry	The present invention
			The research period	1994-1999 (5 years)	1995-2001 (6 years)
The desertification soil increases area (ten thousand km2)	5.2	5.68
			The desertification soil increases (ten thousand km2) every year	1.04	0.95
Slight desertification soil proportion	20％	18％ *
			Moderate desertification soil proportion	33％	31％ *
Severe desertification soil proportion	21％	21％ *
			Utmost point severe desertification soil proportion	26％	30％ *

^*Data were as the criterion with calendar year 2001.

On the basis of desertification remote sensing monitoring index system, the nineteen ninety-five of China and desertification distribution range and the desertification degree of calendar year 2001 are estimated respectively.

The result of DESERTIFICATION IN CHINA soil remote sensing monitoring according to the present invention, nineteen ninety-five, the desertification land area of China is 250.87 * 10 ⁴Km ², account for 25.86% of area.The concrete distribution seen figure (8).

The result of desertification soil remote sensing monitoring according to the present invention, calendar year 2001, the DESERTIFICATION IN CHINA land area that the present invention obtains is 256.55 * 10 ⁴Km ², account for 26.45% of area.The concrete distribution seen figure (9).

On the basis of China's nineteen ninety-five and calendar year 2001 desertification degree distribution figure, obtain the dynamic change figure of the desertification of China.By statistical study, can draw the variation of different desertification type land areas and the differentiation situation (Figure 10) of desertification degree to desertification dynamic change figure.

National forestry portion had once successively carried out twice national desertification soil generaI investigation and monitoring in 1994 and 1999, desertification soil mutation analysis on comparable basis, and we compare (table 13) with achievement in research of the present invention with it.

From table 13, the average annual growth rate in desertification soil is less than the monitoring result of national forestry portion among the desertification earth monitor result of the present invention, wherein main cause is the difference of both research ranges, and research range of the present invention does not comprise the island areas of inferior moistening arid biogeographic zone, and area reduces relatively.In addition, the difference in used remotely-sensed data source when carrying out desertification watch, the not equal of remotely-sensed data acquisition time also is the reason that produces difference.

In a preferred embodiment of the invention, utilize vegetation rudiment event analysis vegetation and nonvegetated area boundary line to change; And utilize the NDVI coefficient of variation (CoV) to analyze the desertification dynamic change.

The used method of the present invention comprises two aspects, is based on year boundary line, desert of border NDVI time series data (or boundary line of vegetation region and nonvegetated area) on the one hand and extracts, and is based on the remote sensing monitoring of desertification for many years of CoV index on the other hand.The former is used to extract the secular variation trend in limit, desert; The latter is used to monitor the desertification dynamic change situation outside the boundary line, desert on the basis of the former work.

1) boundary line, desert is extracted

The many of forefathers studies show that the seasonal variations of NDVI and vegetation phenological period are closely related.The budding period of vegetation is meant that vegetation is subjected to ordering about of moisture or temperature, and from the photosynthetic period that stationary state begins to enliven, in this period, the NDVI value increases suddenly.In the arid and semi-arid area, most vegetation all is seasonal, begins growth from vegetation, and NDVI value increases gradually, reaches maximum up to vegetation growth NDVI value when the most luxuriant, and NDVI reduces gradually then.The present invention extracts the boundary line, desert according to the budding period that whether can monitor vegetation in the zone.

Lloyd and Reed etc. think the unexpected growth of NDVI be vegetation in embryo.Vegetation distributes rarely in the desert, therefore is difficult to measure budding period (Loiyd, 1990 of vegetation with the NDVI data of AVHRR; Reed et al., 1994).Whether F.Yu etc. (2004) think can the boundary line of monitoring the desert of vegetation rudiment incident according to an area.

Whether the present invention's use has vegetation rudiment incident that this index takes place is judged the boundary line, desert.For determining the unexpected growth of NDVI, promptly determine the budding period of vegetation, the NDVI seasonal variations of annual each pixel is monitored.Area then should meet following several condition if there is the rudiment of vegetation to take place: 1. NDVI must increase in one period continuously, and this section period is the time of one and a half months at least, and therefore NDVI is to increase continuously in 5 ten days at least in the present invention; 2. during this period of time, NDVI can reach maximal value; 3. the NDVI value is greater than 0.05; 4. above several conditions must occur in the rational time period, promptly occurred in the normal growth phase of vegetation: between April to August.Under the EASI of PCI image processing software programmed environment, realize the extraction in boundary line, desert.The boundary line, desert of indication of the present invention promptly is the boundary line of vegetation region and no vegetation region.

2) based on the remote sensing monitoring of desertification for many years of CoV method

Surface vegetation is the important indicator that desertification is estimated, and NDVI is used to monitor the most classical vegetation index (QiJ, 2000) that vegetation changes.NDVI is using remote sensing images to carry out being used widely in vegetation study and the research of plant phenology, and it is the best indicator of plant growth state and vegetation space distribution density, is linear dependence with the vegetation distribution density.The coefficient of variation of NDVI (CoV) is a more stable method in monitoring vegetation growth cycle.CoV generally is used for the variation between the different samples of comparison, and these samples are time series datas, so CoV represents is the NDVI situation of change of each pixel in a period of time.CoV can be used for monitoring dynamic vegetation in a period of time, and has been used to estimate the variation of vegetation and extraction (Tucker, 1991 in ecosystem boundary line; L.Milich, and E.Weiss, 1997).CoV is the general performance of multidate NDVI data, and the NDVI of phase changes more stable during therefore than list.

In the different time period (as week, the moon, year), the NDVI sequential value of each pixel can generate a width of cloth CoV image.In invention, annual CoV image is to calculate according to the moon NDVI maximal value of each pixel.At first the standard deviation (σ) of 12 months NDVI of each pixel in the computed image uses standard deviation divided by 12 months NDVI mean value (μ) of this pixel then, promptly obtains the year CoV value of this pixel.Concrete grammar is as the formula (13):

cov _ij＝σ _ij/μ _ij (13)

$\mathrm{\μ}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{p}_i/n---\left(14\right)$

$\mathrm{\σ}=\sqrt{\frac{1}{n-1}[\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{p_i}^2-\frac{{\left(\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{p}_i\right)}^2}{n}}---\left(15\right)$

In the formula (13), i and j are respectively the ranks numbers of each pixel, and in formula (14) and the formula (15), pi represents the pixel value, and n is the time series number.The CoV value is in the same yardstick, therefore more helps the comparison between different pixels.CoV can be used for measuring the growth cycle of vegetation, so the variation of CoV can be monitored the variation in vegetation growth cycle in the zone.The zone that has vegetation to cover in the arid and semi-arid district, the decline of CoV value is relevant with the reduction of the minimizing of rainfall or biomass.

People such as E.Weiss once used the CoV of NDVI and the CoV gradient (being CoV secular variation trend fitting slope of a curve) to estimate Saudi grassland situation of change, the result proves that the method can successfully be used for arid and semi-arid district (E.Weiss, S.E.Marsh, and E.S.Pfirman, 2001).

The present invention uses the CoV method to monitor the desertification situation in arid and semi-arid district outside the boundary line, desert on the basis of extracting the boundary line, desert.At first, should determine the best equation of CoV value linear fit result for many years to each pixel, the slope of regression line that obtains CoV then is the CoV gradient; The CoV gradient has reflected each pixel overall variation trend for many years.The specific algorithm of the CoV gradient adopts the minimum power OLS estimation technique, as the formula (16):

$b=\frac{\mathrm{n\Σ}{x}_i{y}_i-\mathrm{\Σ}{x}_i\mathrm{\Σ}{y}_i}{\mathrm{n\Σ}{x}_i^2-{\left(\mathrm{\Σ}{x}_i\right)}^2}---\left(16\right)$

In the formula (16), x is the time, and x is 1982,1983 among the present invention ..., 2000; Y is the CoV value of each pixel in every year, and i is 1,2 ... n, n are a year sum, and n is 18 among the present invention.

If the variation tendency of CoV descends in the research period, then deducibility goes out that an amplitude of variation reduces in year of NDVI in this district, promptly should distinguish the vegetation state variation, is in the desertification process.Otherwise the variation tendency of CoV rises, and can think that then the vegetation increase or the growth conditions in this district improves.

Utilize the NDVI data in annual 36 ten days, under the EASI of PCI programmed environment, realize the annual vegetation region and the extraction in nonvegetated area boundary line.Figure 11 is border, the desert distribution plan of enumerating in 6 years in the Central Asia and Chinese arid and semi-arid district, and they are respectively 1982,1986,1990,1993,1997 and 2000.

At first extract the Central Asia and Chinese arid and semi-arid district from 1982-2000 (except 1994) the vegetation distributed areas in totally 18 years, by the GIS stack, obtain the Central Asia and Chinese arid and semi-arid district vegetation distribution plan (Figure 12) then.Nonvegetated area territory among the figure is meant in the period of nineteen eighty-two to 2000 year (except 1994) 18 zone that vegetation never distributes.

Figure 13 is vegetation distributed areas another result by GIS stack in 18 years, promptly during the Central Asia and Chinese arid and semi-arid district nineteen eighty-two to 2000 year, and the occurrence frequency distribution plan of vegetation rudiment incident.As can be seen from the figure, the boundary line, desert in arid and semi-arid district has very strong variability, and never monitors the distribution of vegetation in the central area in desert.Through desert steppe, carry out the transition to typical steppe belt from the desert then, in 18 years, the number of times that vegetation was taken place carries out the transition to 18 times from 2 times.

A possible explanation to the polytrope in this boundary line, regional desert is the cause that the variability of the annual rainfall in arid and semi-arid district is bigger.When annual rainfall was abundant, vegetation growth was luxuriant, and the boundary line, desert moves to the desert center position; And when annual rainfall was low, the boundary line, desert moved to the grassland direction.Therefore, desert steppe shows as the phenology feature on typical grassland in the moistening time, then shows as the feature in desert in the time of arid.

Figure 14 promptly is the CoV gradient distribution figure in the Central Asia and western part of China arid and semi-arid district.Among the figure, green area represents that the gradient of CoV is almost nil, i.e. stable the or unconspicuous zone of secular variation of vegetation state; Yellow represent the CoV gradient be positive zone to red, and the expression vegetation growth is tending towards improving for many years, and, be to represent that the amplitude of variation of the CoV gradient is to increase gradually by yellow to the variation of redness, promptly the degree that improves of vegetation growth is more obvious; Blueness is that the CoV gradient is negative zone, and expression vegetation growth situation degenerates, and this district is in the desertification, and is tending towards serious by light blue to navy blue transition explanation desertification degree.Why the CoV gradient is divided into 7 grades, is for more clearly illustrating the power of desertification degree, not having physical significance more specifically.The CoV gradient is positive zone among Figure 14, and China mainly is distributed in the Xinjiang Tianshan area, southern Tibet and east Inner Mongolia area, and the result of study cardinal principle matches with the shelter-forest distribution of China.The area that desertification takes place then mainly is distributed in the edge and the staggered band of agriculture and animal husbandry of desert and sand ground.The Central Asia CoV gradient is for negative, and the zone that desertification takes place mainly is distributed in desert edge, around the lakeside, and zone such as the staggered band of agriculture and animal husbandry.

Claims (6)

1. a land deterioration (desertification) evaluation parameter remote inverting and monitoring technology method comprise:

Obtain remotely-sensed data and carry out pre-service;

To pretreated remotely-sensed data, by setting up mathematical model, inverting land deterioration (desertification) evaluating;

Utilize the parameter of remote-sensing inversion to set up the desertification watch index system;

Utilize desert monitoring index system obtain 1995 and calendar year 2001 the DESERTIFICATION IN CHINA change information;

Utilize vegetation rudiment event analysis vegetation and nonvegetated area boundary line to change;

Utilize the NDVI coefficient of variation (Cov) to analyze the Central Asia and Chinese arid, semiarid region desertification dynamic change.

2. method according to claim 1, it is characterized in that the described land deterioration evaluating of setting up the mathematical model inverting comprises: vegetation index (MSAVI), vegetation coverage (FVC), albedo (ALBEDO), land temperature (Ts), soil moisture (TVDI).

3. method according to claim 1 is characterized in that, the step that described desertification remote sensing monitoring index system is set up comprises:

Desertification remote sensing monitoring index selection principle;

Determining of desertification remote sensing monitoring overall target;

Proposition is suitable for using remote sensing and carries out the parameter index that the large scale desertification is estimated and monitored, and by monitoring parameter combination carrying out desertification watch precision evaluation, the result show adopt modified soil regulate vegetation index, vegetation coverage, albedo, land surface temperature and soil moisture index to be combined in the desertification remote sensing monitoring nicety of grading the highest, can be used as the monitoring overall target;

Based on determining of the desertification level index system of remotely-sensed data;

Choose each desertification index overlay analysis of desertification degree figure and remote sensing; And, study area is divided into inferior moistening arid biogeographic zone, semiarid region, four zones of arid biogeographic zone and high and cold district according to desert climate type difference, set up desertification degree remote sensing monitoring index system respectively at each zone.

4. the method for setting up according to claim 1, it is characterized in that, described China and Central Asia's desertification change information obtain, by utilizing the desertification remote sensing monitoring index system of setting up based on Chinese soil desertification sample, corresponding four subregion nineteen ninety-fives of China and calendar year 2001 desertification situation have been carried out monitoring and evaluation.

5. method according to claim 1, it is characterized in that, it is by utilizing 10 days generated time sequence datas of NOAA AVHRR NDVI collection of 1982-2000 8km resolution that described vegetation and nonvegetated area boundary line change, according to whether vegetation rudiment incident took place in the zone, determine the annual Central Asia and China's arid, semiarid region vegetation and non-vegetation boundary line, and 18 years vegetation boundary lines are carried out the GIS stack, analyze boundary line situation of change for many years.

6. according to the described method of claim 1, it is characterized in that, the described Central Asia and China's arid, semiarid region desertification dynamic rule are to utilize 10 days generated time sequence datas of 1982-2000 NOAA-AVHRR, calculate the coefficient of variation of annual NDVI respectively, try to achieve the NDVI coefficient of variation gradient in 18 years then with least square method.By analytical variance coefficient slope change, realize the evaluation of the Central Asia and China's arid, semiarid region desertification situation for many years.

Images