CN102162850A - Wheat yield remote sensing monitoring and forecasting method based on model - Google Patents

Abstract

The invention provides a method for remote sensing and estimating the yield of crops. The method comprises the following steps of: building a simple wheat yield forecasting model based on the instantaneity and wide regional coverage obtained by remote sensing information in combination with wheat yield forming process and weather environment relationship; coupling the remote sensing information with the estimated yield model through a modular design method, namely, replacing the parameter variable corresponding to the wheat estimation yield model by using the LAI inversed by remote sensing images in the heading stage and the biomass so as to estimate the single point wheat yield, wherein the yield estimation precision is more than 90%; and furthermore, remote sensing, sorting, monitoring and forecasting the area wheat yield by using a point (sample yield) and surface (remote sensing area) size conversion method so as to obtain area wheat yield remote sensing, sorting, forecasting specific topic graphs. The method is featured with directness, specificity, good time effectiveness and excellent practicability for the agricultural personnel to obtain the area wheat layout information or guide the production management.

Description

Wheat yield remote sensing monitoring forecasting procedure based on model

1, affiliated technical field

The inventive method relates to a kind of method of crop Remote Sensing Yield Estimation, especially can dynamically predict wheat yield in the regional extent, and can make the wheat yield classified Monitoring forecasting procedure of output classification forecast thematic map.

2, background technology

Traditional agricultural output assessment method mainly contains agronomy forecasting procedure, statistical fluctuation method, meteorological statistics method etc.Because these the yield by estimation methods need mass data statistics, manual research etc., have the problem that speed is slow, workload is big, cost is high, mostly only be fit to interior among a small circle agricultural output assessment.Since the seventies in last century, remote sensing technology has worldwide obtained developing rapidly and widespread use, and it provides a kind of new scientific method for the crop yield estimation.About wheat yield remote sensing estimation, most of research methods are on the basis of the spectral information of analyzing image data and wheat growing way or output formation relation, are undertaken by setting up regression model.Though these class methods are simple to operate, empirical strong, versatility is relatively poor.Because utilize remote sensing image can obtain the instantaneous growth information of a certain growth phase of wheat, but only very large deviation can appear during the prediction of the growth information by this stage maturity stage output, because climatic environment condition in prediction period (temperature, illumination, soil nitrogen level, soil water regime etc.) is changing constantly, to the very big influence that is formed with of wheat ultimate capacity.

3, summary of the invention

There are empirical strong, the relatively poor characteristics of versatility at traditional Remote Sensing Yield Estimation method, this method is coupled the quantification analogue technique of remote-sensing inversion technology and wheat yield forming process, on the basis that makes up the wheat yield forecast model, utilize remote-sensing inversion information to correct recovery prediction model running track, by " point " (sampling point recovery prediction value) and " face " (remote sensing image) formal transformation, reach purpose again to regional wheat yield monitoring and prediction.The wheat Remote Sensing Yield Estimation method that the present invention set up, mechanism is strong, versatility is stronger, can be between different year, the remote sensing classified Monitoring forecast of wheat yield provides technology or method support in the zones of different.Main summary of the invention and technical matters are described below:

1. the wheat yield forecast model is described

Wheat yield (Yield) can by utilize plant overground part dry matter when ripe (Above-ground BiomassWeight, ABW) with harvest index (Harvest Index, product HI) obtains, its algorithm is as follows:

Yield＝ABW _i×HI (1)

In the formula (1), ABW _i(unit is kgha to plant overground part dry matter during for maturation ^-1), i is for the fate (d) when being seeded into maturation, this moment, i equaled breeding time (d).

During wheat growth, plant overground part dry matter can obtain by following formula:

ABW _i＝ABW _i-l+ΔABW _i (2)

In the formula (2), ABW _iAnd ABW _I-1Be respectively the plant overground part dry matter (kgha of i days and i-1 days ^-1).ABW ₁(the overground part dry matter of emerging first day) is defined as half of sowing weight (kgha).Δ ABW _iBe the daily gain (kgha of i days plant overground part dries ^-1D ^-1), its algorithm is as follows:

ΔABW _i＝ΔDABW _i-RG _i-RM _i (3)

In the formula (3), Δ DABW _i, RG _iAnd RM _iBe respectively the photosynthesis assimilation amount (kgha of i days plant ^-1D ^-1), growth respiration consumption amount (kgha ^-1D ^-1) and keep respiration consumption amount (kgha ^-1D ^-1).RG _iAnd RM _iAlgorithm as follows:

RG _i＝ΔDABW _i×Rg (4)

RM _i＝ABW _i×Rm×Q ₁₀ ^(T-25)/10 (5)

In the formula (4), the Rg continuous call of making a living is inhaled coefficient, value 0.32.In the formula (5), Rm is for keeping coefficienting respiration, value 0.015.Q ₁₀Be respiratory temperature coefficient, value 2.T be daily mean temperature (℃).

The daylight of plant closes assimilation quantity Δ DABW in the formula (4) _iAlgorithm, be expressed as following formula:

${\mathrm{\ΔDABW}}_i=\frac{B}{K\×A}\×\mathrm{Ln}\left(\frac{1+D}{1+D\×\mathrm{Exp}(-K\×{\mathrm{LAI}}_i)}\right)\×\mathrm{DL}\×\mathrm{\δ}\×\min (\mathrm{NF},\mathrm{WF})---\left(6\right)$

D＝A×0.47×(1-α)×Q/DL

In the formula (6), K is colony's extinction coefficient.LAI is a leaf area index, and D is an intermediate variable, and α is the wheat population reflectivity, value 8%.Q is total solar radiation every day amount (MJm ^-2).B, A are experiment coefficient, difference value 5 and 20.δ is CH ₂O and CO ₂Between conversion coefficient, value 0.68.NF, WF are respectively the nitrogen factor of influence and the moisture effects factor.DL is day long (h), can calculate by following formula to obtain:

(7)

β＝23.5×Sin[360×(n+284)/365]

In the formula (7), For geographic latitude (°), β is a solar declination.N be julain day (n=1,2,3 ..., 365).

2. the coupled mode of remote-sensing inversion information and wheat yield forecast model

Leaf area index in the wheat yield forecast model (LAI) and biomass (ABW) are that output forms very crucial population quality indexes, utilize remote sensing technology to catch easily.Utilize remote sensing vegetation index inverting LAI and ABW desired value, coupling wheat yield model can reach the purpose to the wheat yield estimation.Remote-sensing inversion information and yield model coupling adopt the modular design method that meets the COM of Microsoft standard, and its coupled mode structure sees also accompanying drawing 1.

The wheat yield estimation model encapsulates according to the COM standard of the Microsoft form with DLL, designs as follows:

Assembly name: WheatRS.dll

Interface name: IWheatInoutput

Interface function: WheatInoutputfunction (VARIANT FAR*Meto, VARIANT FAR*Interface, VARIANT FAR*RS, VARIANT FAR*Output) wherein, WheatInoutputfunction is a function name, and VARIANT FAR*Meto, VARIANT FAR*Interface, VARIANT FAR*RS and VARIANT FAR*Output are respectively weather data, interface input data, remotely-sensed data information and yield result output variant.

3. wheat yield remote sensing classified Monitoring forecasting procedure

It is the indispensable prerequisite or the important foundation of Remote Sensing Yield Estimation that the wheat planting area extracts.At first, utilizing the ISODATA method to carry out the wheat planting area extracts.Then, utilize the ground GPS reference mark to extract the NDVI value of each sampling point of remote sensing image, institute is extracted NDVI value input leaf area index (LAI) and biomass inverse model, just can obtain the LAI and the biomass predicted data of each sampling point.At last, the LAI and the biomass data of each sampling point are input in the wheat yield forecast model, just can obtain the recovery prediction data message of each sampling point.Adopt the linear transformation method,, form regional production information figure based on remote sensing image with each sampling point yield data information and image NDVI coupling.Carry out classification according to regional wheat breed output performance situation, carry out output statistics analysis and output thematic map at last and make (referring to accompanying drawing 2).

4, beneficial effect

Winter wheat yields is carried out the forecast of remote sensing classified Monitoring, help agricultural management department and in time obtain winter wheat zone production information, be convenient to it and formulate effective cultivation management measure, reach the purpose of volume increase.

2008-2009, utilize the inventive method, the winter wheat yields on ground such as Taixing City, Jiangsu Province, Jiangyan City, Rugao City, Hai'an city, Dafeng City, Xing Hua city, Yizheng City is carried out monitoring and prediction, the yield by estimation precision reaches more than 90%.Further, adopt the method for " point " (model prediction) and " face " (zone shows) linear transformation, made wheat yield remote sensing monitoring classification prog chart (referring to accompanying drawing 3), have good characteristics directly perceived, concrete, ageing, the basic agriculture technician is obtained regional wheat layout information or instructs production management to have better practicability, use area and reach more than 300 ten thousand mu.

5, embodiment

1. materials and methods

1.1 material

2008 in Taixing City, Jiangsu Province (116 ° 18 '～121 ° 57 of east longitude ', 30 ° 45 '～35 ° 20 of north latitude ') choose 20 experiment sampling points, pass by period at satellite, utilize differential GPS fix a point investigation and sampling.Each is pressed the triangle type and selects 3 districts (interval 5m), and the plant sample is got by 0.5m * 0.5m area by every district, and final data is by the average acquisition in 3 districts.The content of investigation comprises: the geographic position of sampling point, field farming feelings overview, leaf area index, and information such as biomass.Kind is for raising wheat No. 14 and raising wheat No. 13.Leaf area index adopts the specific leaf weight method to measure.Biomass is measured, and the 20min that completes under 105 ℃ earlier 75 ℃ of oven dry down, takes by weighing oven-dry weight subsequently.Determination of yield, each sampling point all utilize 50m * 50cm sample frame, according to field piece diagonal line 5 points (four angle points are 10m at interval) sampling, and each some 1m ², get 5m altogether ²Seed, oven dry takes by weighing weight then.

Select the Landsat/TM image data for use, substar resolution is 30m.In Taixing City is on May 2nd, 2008 time of passing by.The same day is fine, and is cloudless, and the satellite image quality is better, just as flowering stage of wheat.Image processing when utilizing ERDAS software that image is carried out geometry correction, is carried out geometric accurate correction in conjunction with the GPS reference mark of actual measurement, guarantees that correction error is less than 1 pixel.Atmosphere radiation is proofreaied and correct and reflectivity conversion is to utilize the original DN value of actual measurement reflectivity data with the satellite image of correspondence of ground calibration body, adopts the conversion of experience linear approach to obtain.

1.2 recovery prediction and the forecast of output classified Monitoring

It is the indispensable prerequisite or the important foundation of Remote Sensing Yield Estimation that the wheat planting area extracts.At first, utilize the ISODATA method to carry out the wheat planting area and extract, and the verification of stack GPS sampling point.Then, in ERDAS software, utilize 20 GPS reference mark to extract the NDVI value of each sampling point of remote sensing image, institute is extracted NDVI value input leaf area index (LAI) and biomass inverse model, just can obtain the LAI and the biomass predicted data of each sampling point.At last, the LAI and the biomass data of each sampling point are input in the wheat yield estimation model, just can obtain the recovery prediction data message of each sampling point.Adopt the linear transformation method,, form regional production information figure based on remote sensing image with each sampling point yield data information and image NDVI coupling.Carry out classification according to the local wheat breed output performance situation in Taixing, utilize generalized information system software to carry out output statistics analysis and output thematic map making (referring to accompanying drawing 2).

2. interpretation of result

Through wheat area remote Sensing Interpretation, and the adjustment of data of stack test sampling point, obtaining Taixing City's winter wheat cultivated area in 2008 is 44678.71ha, and this city actual winter wheat cultivated area in 2008 is 43333.33ha (this area data is provided by local agricultural sector), degree of accuracy is 95%, and the result is comparatively reliable.

On the basis of analyzing " NDVI → growing way index (LAI and biomass) → Yield Estimation Model → output " relation, extract sampling point NDVI numerical value with the sampling point Vector Message earlier, utilize sampling point NDVI inverting (reckoning) LAI and the upperground part biomass data again, in conjunction with Yield Estimation Model (model parameter sees Table 1), calculate the sampling point yield data.

Table 1 wheat breed parameter information

Winter wheat is in and comes into bloom May, and the NDVI data of this moment have correlativity preferably with the formation of final output.Based on the yield values of sampling point, through linear transformation, the output that can obtain whole imagery zone distributes.Stack sampling point wheat yield data (LAI and the forecast production information that comprise the sampling point actual measurement) are revised, and obtain this regional wheat yield classified Monitoring forecast thematic map (seeing accompanying drawing 4).

According to wheat yield remote sensing classification figure, utilize the GIS software analysis and export the area distributions situation of each output grade winter wheat, list in table 2.As can be seen, high-yield field (output 〉=6750kgha ^-1) and than high-yield field (output 〉=6000kgha ^-1, and＜6750kgha ^-1) proportion is bigger, is respectively 10942.60ha and 21967.53ha, accounts for 73.7% of the total area; Middle field (output 〉=the 5250kgha that produces ^-1, and＜6000kgha ^-1) area is 7732.53ha, accounts for 17.3% of the total area; Output is lower than 5250kgha ^-1Field piece area quite a few, area is 4036.05ha, account for 9.0% of the total area, be mainly the periphery and the hypsography low-lying area that are distributed in large tracts of land field piece, explanation is selected also existing problems in field management and planting patterns, has the yield potentiality that can excavate, and how to strengthen the management of these field pieces, change the relatively poor situation of growing way, the output that improves low-yield land is still the very important from now on task of local agricultural sector.

Each output grade winter wheat area distribution situation of table 2 Taixing City

The inventive method is on the basis of setting up the winter wheat yields forecast model, in conjunction with the parameter of sensor information inverting, through data assimilation, bring forecast model into and obtain simulation output, obtain higher degree of fitting with the sampling point output of surveying, the yield by estimation precision can reach more than 90%.Further, the method that adopts " point " (sampling point output) and " face " (remote sensing zone) yardstick to change, made wheat yield remote sensing monitoring classification prog chart, have good characteristics directly perceived, concrete, ageing, the basic agriculture technician is obtained regional wheat layout information or instructs production management to have better practicability.

Claims (1)

1. recovery prediction model based on the wheat yield forming process; Realized the coupling of sensor information and recovery prediction model by the Componentized method for designing, utilized the LAI of remote sensing image inverting at heading stage and biomass in time to replace the corresponding parametric variable of wheat yield estimation model, and then realize estimation the single-point wheat yield; The method that adopts " point " (sampling point output) and " face " (remote sensing zone) yardstick to change is carried out regional wheat yield remote sensing classified Monitoring forecast, can form regional wheat yield remote sensing monitoring classification forecast thematic map.

Images