CEEMDAN-based method for screening and monitoring soil moisture stress in agricultural fields

Abstract

The present invention discloses a CEEMDAN-based method for screening and monitoring soil moisture stress in farmland, characterised by the steps: preprocessing of remote sensing images, construction of NDVI long time series, CEEMDAN decomposition, calculation of statistical descriptors, screening of soil moisture stress sequences, ground data measurement, construction of soil moisture stress characteristic curves, fitting of soil moisture stress response characteristic curves and predicting the content of soil moisture stress. The invention adopts CEEMDAN decomposition, which solves the problems of noise residue and low reconstruction accuracy in the previous methods, and the high reconstruction accuracy of decomposed component data is more conducive to capturing the transient effects of soil moisture stress, and realizes the screening and extraction of soil moisture stress by combining with the ground measured data. The inverse model of soil moisture content is fitted by combining the effects of multiple indicators, and the CEEMDAN algorithm with remote sensing technology tools to achieve accurate monitoring of soil moisture in a large area of farmland.

Claims

1. A CEEMDAN-based method for screening and monitoring soil moisture stress in agricultural fields, comprising the steps as follows. Step 1: Preprocessing of remote sensing images. Radiometric calibration, atmospheric correction and geometric correction pre-processed on all N remote sensing images. Step 2: Construction of NDVI long time series. Calculation of NDVI long time series x(n),1?n?N based on pre-processed remote sensing image data. $\begin{array}{cc}x\left(n\right)=\frac{{?}_{\mathrm{NIR}\left(n\right)}-{?}_{\mathrm{RED}\left(n\right)}}{{?}_{\mathrm{NIR}\left(n\right)}+{?}_{\mathrm{RED}\left(n\right)}}& \left(\mathrm{formula}1\right)\end{array}$ In formula 1?.sub.NIR(n) is the reflectance of the nth remote sensing image in the near infrared band and ?.sub.RED(n) is the reflectance of the nth remote sensing image in the red band to construct a long time series of NDVI for crops in natural farmland ecosystems. Step 3: CEEMDAN decomposition. The NDVI long time series is decomposed based on CEEMDAN algorithm. The decomposition process is as follows. First add adaptive white noise Z.sup.(m)(n) to the NDVI long time series x(n) where m denotes the number of times the noise is added, 1?m?50, to obtain the first signal to be decomposed:
x.sub.1.sup.(m)(n)=x(n)+?.sub.mZ.sup.(m)(n) ?.sub.m is the standard deviation of the mth addition of white noise, the first IMF component obtained from the CEEMDAN decomposition is as follows. ${\mathrm{IMF}}_1=\frac{1}{50}\underset{m=1}{\overset{50}{.Math.}}{\mathrm{IMF}}_1^{{}^{}\left(m\right)}$ IMF.sub.1.sup.(m) denotes the IMF component obtained by EMD decomposition of the first signal x.sub.1.sup.(m)(n)to be decomposed. The first residual term is S.sub.1(n)=x(n)?IMF.sub.1. The rth to-be-decomposed signal x.sub.r.sup.(m)(n)=S.sub.r?1(n)+?.sub.mZ.sup.(m)(n) is obtained by superimposing white noise on the (r?1)th residual term, and the rth to-be-decomposed signal x.sub.r.sup.(m)(n) is decomposed again by EMD to obtain the IMF component IMF.sub.r.sup.(m). The rth IMF component is then as follows. ${\mathrm{IMF}}_r=\frac{1}{50}\underset{m=1}{\overset{50}{.Math.}}{\mathrm{IMF}}_r^{{}^{}\left(m\right)}$ The rth residual term is S.sub.r(n)=S.sub.r?1(n)?IMF.sub.r, r=1, . . . , R. Step 4: Calculation of statistical descriptors. Calculating statistical descriptors for the first to Rth IMF component IMF.sub.r,r=1, . . . , R, the statistical descriptors including period of fluctuation (P.sub.r), mean (M.sub.r), variance (V.sub.r), variance contribution margin (C.sub.r) and pearson correlation coefficient (PS.sub.r), where K.sub.r is the number of extreme value points of the rth eigenmodal component IMF.sub.r, r=1, . . . , R. $\begin{array}{cc}{P}_r=\frac{N}{K_r}& \left(\mathrm{formula}2\right)\end{array}$ $\begin{array}{cc}{M}_r=\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\mathrm{IMF}}_r\left(n\right)& \left(\mathrm{formula}3\right)\end{array}$ $\begin{array}{cc}{V}_r=\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\left({\mathrm{IMF}}_r\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\mathrm{IMF}}_r\left(n\right)\right)}^2& \left(\mathrm{formula}4\right)\end{array}$ $\begin{array}{cc}{C}_r=\frac{\underset{n=1}{\overset{N}{.Math.}}{\left({\mathrm{IMF}}_r\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\mathrm{IMF}}_r\left(n\right)\right)}^2}{\underset{n=1}{\overset{N}{.Math.}}{\left(x\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}x\left(n\right)\right)}^2}& \left(\mathrm{formula}5\right)\end{array}$ $\begin{array}{cc}{\mathrm{PS}}_r=\frac{\underset{n=1}{\overset{N}{.Math.}}\left({\mathrm{IMF}}_r\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\mathrm{IMF}}_r\left(n\right)\right)\left(x\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}x\left(n\right)\right)}{\sqrt{\underset{n=1}{\overset{N}{.Math.}}{\left({\mathrm{IMF}}_r\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\mathrm{IMF}}_r\left(n\right)\right)}^2\underset{n=1}{\overset{N}{.Math.}}{\left(x\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}x\left(n\right)\right)}^2}}& \left(\mathrm{formula}6\right)\end{array}$ Step 5: Screening of soil moisture stress sequences. Soil moisture stress sequences are identified by combining the statistical descriptors described above with the mechanistic characteristics of soil moisture stress. IMF components that meet the characteristics of short-period soil moisture stress and the identification conditions are summed and synthesized into soil moisture stress sequences. Step 6: Ground data measurement. Chlorophyll content of crop leaves is determined using a hand-held chlorophyll meter by weighing the fresh biomass of the crop plant (FB), using the drying method at a predetermined killing temperature for a predetermined time, turning the temperature to a predetermined drying temperature and drying to a constant weight, then weighing its dry biomass (DB) and calculating the plant water content (PWC). $\begin{array}{cc}PWC=\frac{\mathrm{FB}-\mathrm{DB}}{\mathrm{FB}}?100\%& \left(\mathrm{formula}7\right)\end{array}$ Weigh the wet weight of the soil sample (FW), dry it to a constant weight using the drying method at a preset drying temperature, then weigh its dry weight (DW) and calculate the soil moisture content (SMC). $\begin{array}{cc}SMC=\frac{\mathrm{FW}-\mathrm{DW}}{\mathrm{DW}}?100\%& \left(\mathrm{formula}8\right)\end{array}$ Step 7: Construction of soil moisture stress characteristic curves. Take the two points n and n+1 of the soil moisture stress sequence including the measured time t and the corresponding values y(n) and y(n+1), and fit the corresponding value of the measured time t as the corresponding soil moisture stress content Y(t) at the measured time point.
Y(t)=(t?n)(y(n+1)?y(n))+y(n) (formula 9) Step 8: Fitting of soil moisture stress response characteristic curves. Soil moisture stress content is used as the independent variable and chlorophyll content is used as the dependent variable to construct the chlorophyll response index. Soil moisture stress is used as the independent variable and plant water content is used as the dependent variable to construct the wheat moisture content response index; the function of soil moisture stress and chlorophyll content and the function of soil moisture stress and plant water content are fitted. Step 9: Predicting the content of soil moisture stress. Constructe the model to predict the degree of soil moisture stress by inverting the soil moisture content between three indicators, soil moisture stress, chlorophyll response to soil moisture stress and wheat moisture content response to soil moisture stress.

2. The method for screening and monitoring soil moisture stress in agricultural fields based on CEEMDAN according to claim 1, characterized in that the conditions for screening soil moisture stress sequences in step 5 are that the fluctuation period P.sub.r is less than 7, the mean value M.sub.r is the smallest, and the variance V.sub.r, variance contribution C.sub.r and pearson correlation coefficient PS.sub.r are the largest.

3. The method for screening and monitoring soil moisture stress in agricultural fields based on CEEMDAN according to claim 1, characterized in that screening out soil moisture stress sequences in step 5 synthesizes the first and second IMF components IMF.sub.1 and IMF.sub.2 cumulatively into a soil moisture stress sequence. $\begin{array}{cc}y\left(n\right)=\underset{n=1}{\overset{N}{.Math.}}\left({\mathrm{IMF}}_1\left(n\right)+{\mathrm{IMF}}_2\left(n\right)\right)& \left(\mathrm{formula}10\right)\end{array}$

4. The method for screening and monitoring soil moisture stress in agricultural fields based on CEEMDAN according to claim 1, characterized in that the quadratic curve is used to fit chlorophyll content as a function of soil moisture stress and a composite curve is used to fit wheat moisture content as a function of soil moisture stress in step 8.

5. The method for screening and monitoring soil moisture stress in agricultural fields based on CEEMDAN according to claim 4, characterized in that the chlorophyll content as a function of the amount of soil moisture stress content in step 8 described is y=58.241?19.917x?393.742x.sup.2 and the plant water content as a function of the amount of soil moisture stress content is y=68.121+0.404.sup.x.

6. The method for screening and monitoring soil moisture stress in agricultural fields based on CEEMDAN according to claim 2, characterized in that the model between soil moisture stress, chlorophyll response to soil moisture stress and wheat moisture response to soil moisture stress and soil moisture content in step 9 is as follows.
SMC=20.58?0.02x.sub.1+10.sup.?2x.sub.2?1.76x.sub.3?2?10.sup.?3x.sub.1x.sub.2?0.065x.sub.1x.sub.3?0.045x.sub.2x.sub.3?1.83?10.sup.?4x.sub.1.sup.2+7.91?10.sup.?5x.sub.2.sup.2+3.99x.sub.3.sup.2 x.sub.1 represents chlorophyll response to soil moisture stress, x.sub.2 represents wheat moisture response to soil moisture stress and x.sub.3 represents soil moisture stress content.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The invention is described in further detail below in conjunction with the attached drawings and specific embodiments.

[0035] FIG. 1 is the flow chart of the invention.

[0036] FIG. 2 is a NDVI long time series diagram in example 2 of the invention.

DETAILED DESCRIPTION

Example 1:

[0037] A CEEMDAN-based method for screening and monitoring soil moisture stress in agricultural fields, comprising the steps as follows.

[0038] Step 1: Preprocessing of remote sensing images. Radiometric calibration, atmospheric correction and geometric correction pre-processed on all N remote sensing images.

[0039] Step 2: Construction of NDVI long time series. Calculation of NDVI long time series x(n),1?n?N based on pre-processed remote sensing image data.

[00008]$\begin{array}{cc}x\left(n\right)=\frac{{?}_{\mathrm{NIR}\left(n\right)}-{?}_{\mathrm{RED}\left(n\right)}}{{?}_{\mathrm{NIR}\left(n\right)}+{?}_{\mathrm{RED}\left(n\right)}}& \left(\mathrm{formula}1\right)\end{array}$

[0040] In formula 1, ?.sub.NIR(n) is the reflectance of the nth remote sensing image in the near infrared band and ?.sub.RED(n) is the reflectance of the nth remote sensing image in the red band to construct a long time series of NDVI for crops in natural farmland ecosystems.

[0041] Step 3: CEEMDAN decomposition. The NDVI long time series is decomposed based on CEEMDAN algorithm. The decomposition process is as follows.

[0042] First add adaptive white noise Z.sup.(m)(n) to the NDVI long time series x(n) where m indicates the number of times the noise is added, generally take 10 to 50, this example take 50, then the NDVI long time series signal after the mth adaptive white noise is added for the first time can be expressed as follows.

x.sub.1.sup.(m)(n)=x(n)+?.sub.mZ.sup.(m)(n)

[0043] ?.sub.m is the standard deviation of the mth addition of white noise and m takes values from 1 to 50, the first IMF component obtained from the CEEMDAN decomposition is as follows.

[00009]${\mathrm{IMF}}_1=\frac{1}{50}\underset{m=1}{\overset{50}{.Math.}}{\mathrm{IMF}}_1^{{}^{}\left(m\right)}$

[0044] IMF.sub.1.sup.(m) denotes the IMF component obtained by EMD decomposition of the first signal x.sub.1.sup.(m)(n) to be decomposed. The first residual term is S.sub.1(n)=x(n)?IMF.sub.1. The new signa x.sub.2.sup.(m)(n)=S.sub.1(n)+?.sub.mZ.sup.(m)(n) to be decomposed is obtained by superimposing white noise on the first residual term, and the new signal x.sub.2.sup.(m)(n) to be decomposed is again subjected to EMD decomposition to obtain the IMF component, the second component IMF.sub.2.sup.(m) obtained by CEEMDAN decomposition is as follows.

[00010]${\mathrm{IMF}}_2=\frac{1}{50}\underset{m=1}{\overset{50}{.Math.}}{\mathrm{IMF}}_2^{{}^{}\left(m\right)}$

[0045] The second residual term is S.sub.2(n)=S.sub.1(n)?IMF.sub.2. Repeat the above process to obtain the first to Rth IMF components IMF.sub.r,r=1, . . . , R, with the final residual term S.sub.R=S.sub.R?1(n)?IMF.sub.R.

[0046] Step 4: Calculation of statistical descriptors. Calculating statistical descriptors for the first to Rth IMF component IMF.sub.r,r?1, . . . , R. The statistical descriptors including period of fluctuation (P.sub.r), mean (M.sub.r), variance (V.sub.r), variance contribution margin (C.sub.r) and pearson correlation coefficient (PS.sub.r), where K.sub.r is the number of extreme value points of the rth eigenmodal component IMF.sub.r, r=1, . . . , R.

[00011]$\begin{array}{cc}{P}_r=\frac{N}{K_r}& \left(\mathrm{formula}2\right)\end{array}$$\begin{array}{cc}{M}_r=\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\mathrm{IMF}}_r\left(n\right)& \left(\mathrm{formula}3\right)\end{array}$$\begin{array}{cc}{V}_r=\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\left({\mathrm{IMF}}_r\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\mathrm{IMF}}_r\left(n\right)\right)}^2& \left(\mathrm{formula}4\right)\end{array}$$\begin{array}{cc}{C}_r=\frac{\underset{n=1}{\overset{N}{.Math.}}{\left({\mathrm{IMF}}_r\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\mathrm{IMF}}_r\left(n\right)\right)}^2}{\underset{n=1}{\overset{N}{.Math.}}{\left(x\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}x\left(n\right)\right)}^2}& \left(\mathrm{formula}5\right)\end{array}$$\begin{array}{cc}{\mathrm{PS}}_r=\frac{\underset{n=1}{\overset{N}{.Math.}}\left({\mathrm{IMF}}_r\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\mathrm{IMF}}_r\left(n\right)\right)\left(x\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}x\left(n\right)\right)}{\sqrt{\underset{n=1}{\overset{N}{.Math.}}{\left({\mathrm{IMF}}_r\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}{\mathrm{IMF}}_r\left(n\right)\right)}^2\underset{n=1}{\overset{N}{.Math.}}{\left(x\left(n\right)-\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}x\left(n\right)\right)}^2}}& \left(\mathrm{formula}6\right)\end{array}$

[0047] Step 5: Screening of soil moisture stress sequences. Soil moisture stress sequences are identified by combining the statistical descriptors described above with the mechanistic characteristics of soil moisture stress. Soil moisture stress subsequences are screened on the condition that the fluctuation period P.sub.r is less than 7, the mean value M.sub.r is the smallest, and the variance V.sub.r, variance contribution C.sub.r and pearson correlation coefficient PS.sub.r are the largest. The first and second IMF components IMF.sub.1 and IMF.sub.2 met the short-period soil moisture stress characteristics and screening conditions, and the first and second IMF components IMF.sub.1 and IMF.sub.2are summed to form the soil moisture stress sequence as follows.

[00012]$\begin{array}{cc}y\left(n\right)=\underset{n=1}{\overset{N}{.Math.}}\left({\mathrm{IMF}}_1\left(n\right)+{\mathrm{IMF}}_2\left(n\right)\right)& \left(\mathrm{formula}7\right)\end{array}$

[0048] Step 6: Ground data measurement. Chlorophyll content of crop leaves is determined using a hand-held chlorophyll meter, avoiding the leaf veins and nearby locations of the plant leaves when measuring. The intact plants are dug up using a sapper and brought back to the laboratory in a sealed bag, the roots are cleaned of impurities and dried with absorbent paper, the fresh biomass is weighed on a balance with 0.1 g accuracy and the fresh biomass (FB) is measured using the drying method at 105? C. for two hours, then the temperature is turned down to 80? C. and dried to a constant weight and the dry biomass (DB) is weighed and the plant water content (PWC) is calculated. The soil within 5 to 10 centimeter of the sample point is sealed and brought back to the laboratory. The wet weight (FW) of the soil sample is weighed with a balance of 0.1 g accuracy, the soil sample is dried to a constant weight in an oven at 105? C., then its dry weight (DW) is weighed, and finally the soil moisture content (SMC) is calculated.

[00013]$\begin{array}{cc}PWC=\frac{\mathrm{FB}-\mathrm{DB}}{\mathrm{FB}}?100\%& \left(\mathrm{formula}8\right)\end{array}$$\begin{array}{cc}SMC=\frac{\mathrm{FW}-\mathrm{DW}}{\mathrm{DW}}?100\%& \left(\mathrm{formula}9\right)\end{array}$

[0049] Step 7: Construction of soil moisture stress characteristic curves. Take the two points n and n+1 of the soil moisture stress sequence including the measured time t and the corresponding values y(n) and y(n+1), and fit the corresponding value of the measured time t as the corresponding soil moisture stress content Y(t) at the measured time point.

Y(t)=(t?n)(y(n?1)?y(n))+y(n) (formula 10)

[0050] Step 8: Fitting of soil moisture stress response characteristic curves. Soil moisture stress content is used as the independent variable and chlorophyll content is used as the dependent variable to construct the chlorophyll response index. Soil moisture stress is used as the independent variable and plant water content is used as the dependent variable to construct the wheat moisture content response index; the function of soil moisture stress and chlorophyll content and the function of soil moisture stress and plant water content are fitted.

[0051] Step 9: Predicting the content of soil moisture stress. Constructe the model to predict the degree of soil moisture stress by inverting the soil moisture content between three indicators, soil moisture stress, chlorophyll response to soil moisture stress and wheat moisture content response to soil moisture stress.

SMC=20.58?0.02x.sub.1+10.sup.?2x.sub.2?1.76x.sub.3?2?10.sup.?3x.sub.1x.sub.2?0.065x.sub.1x.sub.3?0.045x.sub.2x.sub.3?1.83?10.sup.?4x.sub.1.sup.2+7.91?10.sup.?5x.sub.2.sup.2+3.99x.sub.3.sup.2

[0052] x.sub.1 represents chlorophyll response to soil moisture stress, x.sub.2 represents wheat moisture response to soil moisture stress and x .sub.3 represents soil moisture stress content.

[0053] This example uses a quadratic curve to fit the chlorophyll content as a function of soil moisture stress and a composite curve to fit the plant water content as a function of soil moisture stress. The relationship between chlorophyll content and soil moisture stress is y=58.241?19.917x?393.742.sup.2, while the relationship between plant water content and soil moisture stress is y=68.121+0.404.sup.x. The fitting results show that the chlorophyll response to soil moisture stress index and the wheat moisture response ti soil moisture stress index can effectively reflect the chlorophyll content and plant water content in response to soil moisture stress.

[0054] The NDVI long time series are constructed based on the remote sensing images of GF-1. The long time series are decomposed by CEEMDAN to obtain each IMF component, and the statistical descriptive indexes such as fluctuation period, mean, variance, variance contribution and pearson correlation coefficient are calculated for each component. The ground data are measured at different fertility stages of crops, and the chlorophyll content of plant leaves, plant water content and soil moisture content are measured to obtain the real values of ground indicators in natural agro-ecosystems. The IMF components are compared and analysed to extract the soil moisture stress subsequence, and then the soil moisture stress subsequences are synthesised to obtain the soil moisture stress sequence. Combining the soil moisture stress data with the ground measured data, we construct the chlorophyll response index and the wheat moisture content response index for soil moisture stress, and finally build a multi-indicator inversion model for accurate monitoring of soil moisture content.

[0055] The reconstruction accuracy of this example using CEEMDAN decomposition reached 100%, indicating that the summation of each IMF component after decomposition can obtain the nature of the original data. CEEMDAN decomposition only produces a unique IMF residual term, effectively solving the problem of transferring white noise from high to low frequencies, fully demonstrating the completeness of data reconstruction and reflecting the advantages of CEEMDAN decomposition. CEEMDAN reduces the interference of white noise on the original data and retains the detailed information of the original data. The use of CEEMDAN algorithm can better capture the transient effects of soil moisture stress belonging to short-term stress components, which is conducive to the accurate screening and extraction of soil moisture stress and the subsequent improvement of soil moisture content inversion accuracy.

[0056] When crops are stressed by soil moisture, there is a high reflectance in the visible band and a low reflectance in the near infrared band, resulting in a significant decrease in NDVI values. Soil moisture stress is a short-term stress with a short fluctuation period, the short duration of the stress means that the components characterising short-term stress are more correlated with the original data than those characterising long-term stress. The first and second IMF components IMF.sub.1 and IMF.sub.2 are identified as soil moisture stress subsequence because they had the lowest mean, the highest variance and variance contribution, and the highest pearson correlation coefficient with the original data, with a fluctuation period of less than 7 months (winter wheat growth cycle). The first to and second IMF components IMF.sub.1 and IMF.sub.2 are synthesised to characterise the effect of soil moisture stress on winter wheat during the growth cycle. Table 1 provides a statistical description of each IMF for this example.

TABLE-US-00001 TABLE 1 variance pearson correlation IMF fluctuation period mean variance contribution coefficient IMF.sub.1 1.600000 ?0.010651 0.013138 0.322034 0.606323 IMF.sub.2 4.285714 ?0.007224 0.019400 0.475524 0.693123 IMF.sub.3 8.000000 0.001574 0.006527 0.159974 0.304420 IMF.sub.4 13.333333 ?0.001395 0.001710 0.041914 0.133364 IMF.sub.5 20.000000 ?0.003233 0.000453 0.011113 0.196660 IMF.sub.6 60.000000 0.391738 0.000052 0.001282 0.228038