CN110427995B - Bayesian soil moisture estimation method based on multi-source remote sensing data - Google Patents

Bayesian soil moisture estimation method based on multi-source remote sensing data Download PDF

Info

Publication number
CN110427995B
CN110427995B CN201910671852.6A CN201910671852A CN110427995B CN 110427995 B CN110427995 B CN 110427995B CN 201910671852 A CN201910671852 A CN 201910671852A CN 110427995 B CN110427995 B CN 110427995B
Authority
CN
China
Prior art keywords
data
soil moisture
value
probability
obtaining
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910671852.6A
Other languages
Chinese (zh)
Other versions
CN110427995A (en
Inventor
王春梅
顾行发
谢秋霞
韩乐然
余涛
孟庆岩
占玉林
杨健
李娟�
魏香琴
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Aerospace Information Research Institute of CAS
Original Assignee
Aerospace Information Research Institute of CAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Aerospace Information Research Institute of CAS filed Critical Aerospace Information Research Institute of CAS
Priority to CN201910671852.6A priority Critical patent/CN110427995B/en
Publication of CN110427995A publication Critical patent/CN110427995A/en
Application granted granted Critical
Publication of CN110427995B publication Critical patent/CN110427995B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/24—Earth materials
    • G01N33/246—Earth materials for water content
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G06F18/24155—Bayesian classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/251—Fusion techniques of input or preprocessed data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G06V10/758—Involving statistics of pixels or of feature values, e.g. histogram matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Chemical & Material Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Food Science & Technology (AREA)
  • Pathology (AREA)
  • Immunology (AREA)
  • Biochemistry (AREA)
  • Analytical Chemistry (AREA)
  • Medicinal Chemistry (AREA)
  • Remote Sensing (AREA)
  • Geology (AREA)
  • General Life Sciences & Earth Sciences (AREA)
  • Environmental & Geological Engineering (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Multimedia (AREA)
  • Image Processing (AREA)
  • Testing Or Calibration Of Command Recording Devices (AREA)

Abstract

The invention relates to a Bayesian soil moisture estimation method based on multisource remote sensing data. According to the method, 12 kinds of multi-source data are fused for the first time to calculate weighted probability soft data aiming at how to obtain high-precision soil moisture weighted probability soft data, wherein the weighted probability soft data comprises downscaling FY3-B soil moisture products, obtaining albedo (A) by using MODIS products, vegetation index NDVI (V) and surface temperature LST (T); acquiring elevation data using an ASTER product; and calculating the elevation data obtained by ASTER products to obtain gradient, slope direction, plane curvature, section curvature, surface roughness, humidity index and fluctuation amplitude data; and weighted probability soft data is obtained by adopting two weight determining methods of multivariate correlation analysis and principal component analysis. The invention also analyzes the precision analysis of different soft data quantity, proposes to ensure sufficient soft data quantity, and plays an important role in obtaining the soil moisture spatial distribution with higher precision.

Description

Bayesian soil moisture estimation method based on multi-source remote sensing data
Technical Field
The invention relates to a soil moisture estimation method based on multi-source data, in particular to a Bayesian soil moisture estimation method for fusing multi-source remote sensing data such as downscaled soil moisture product data and topography and the like.
Background
The lack of Soil Moisture (SM) data with high spatial resolution has become one of the major bottlenecks in improving accuracy of the watershed scale ecological hydrologic model. The Bayesian Maximum Entropy (BME) algorithm is an estimation algorithm for modeling large-scale spatial heterogeneity, and can integrate multiple types of data with different accuracies and qualities. Theoretically, integrating multiple types of data related to SM space into SM space estimation using BME algorithms can improve SM accuracy.
Soil Moisture (SM) is not only a key parameter of hydrologic models, climate predictive models, drought monitoring models and crop yield estimation models, but also an important factor of global climate change and surface data assimilation. The traditional in-situ SM measurement can accurately measure SM at a single point, and cannot meet the requirement of large-scale dynamic SM monitoring. With the development and perfection of satellite remote sensing technology, many SM monitoring methods based on visible/near infrared, thermal infrared and active/passive microwave satellite data have been developed, which also make SM monitoring possible. In particular, passive and active microwave satellite data are only minimally affected by overcast and rainy weather, whereas passive microwave data in particular provide a high-sensitivity method of observing SM on a large scale. Therefore, the remote sensing microwave data becomes the main input of land SM remote sensing products, and has wide application prospect in global SM monitoring. However, most passive microwave remote sensing products in most areas are not as accurate as expected (RMSE> 0.06cm 3 / cm 3 ). The practical value of passive microwave SM products is greatly limited due to their low accuracy. Furthermore, most passive microwave SM search algorithms are optimized for uniform surfaces. Due to remote sensing devices and imaging sourcesThe spatial resolution of passive microwave satellites is very low, in the range of tens of kilometers. This results in internal heterogeneity of the microwave pixels, which complicates verification of SM products. Therefore, development of SM products having high spatial resolution is urgently required.
Disclosure of Invention
The BME algorithm considers uncertainty of data and can effectively fuse various data according to the principle of maximum information entropy. The BME defines accurately measured data as Hard Data (HD) and data with uncertainty as Soft Data (SD). SD is converted to probability SD by using probability distribution functions (e.g., histogram and gaussian distribution functions), and the probability SD of SM is calibrated using HD to obtain a post-probability distribution function. Finally, SM is estimated by using a probability post distribution function. The invention comprises the following specific contents:
a Bayesian soil moisture estimation method based on multi-source remote sensing data comprises the following steps:
step 1, obtaining soil moisture soft data and performing data processing
Acquiring soil moisture data of a designated area and a designated date by using FY3-B soil; performing downscaling treatment on the FY3-B soil moisture data, and improving the spatial resolution of the FY3-B soil moisture data with the resolution of 25km to the soil moisture data with the resolution of 1km by using a regression model;
（1）
wherein the method comprises the steps ofA 25 Is the average value of 25km for albedo (a),V 25 is the average value of 25km of the vegetation index NDVI (V),T 25 is the average value of 25km of the surface temperature LST (T),mandnrespectively the first of FY3-B pixelsiLine and thjThe number of 1km pixels in a column;A ij is FY3-BiLine and thjThe albedo value of 1Km pixels in the column,V ij is FY3-BiLine and thjNDVI values of 1Km pixels in the column,T ij Is FY3-B iLine and thjSurface temperature LST values of 1Km pixels in the column;
the linear regression model was: sm=a 1 +a 2 A+a 3 V+a 4 T+a 5 AV+a 6 AT+a 7 VT （2）
Wherein SM is soil moisture; a, a 1 、a 2 、a 3 、a 4 、a 5 、a 6 And a 7 Is a regression coefficient; a is albedo, V is vegetation index, and T is surface temperature;
first, A calculated by equation (1) 25 、V 25 And T 25 Carrying out formula (2) according to FY3-B soil moisture data SM corresponding to pixel scale 25km resolution 25 Calculating regression coefficients in the model by regression analysis;
secondly, taking the albedo (A) of 1km resolution of the MODIS product, the vegetation index NDVI (V) and the surface temperature LST (T) into a model formula (2), and correspondingly calculating to obtain a soil moisture value SM of 1km resolution 1 ；
Step 2, obtaining environment factor soft data
Obtaining albedo (a), vegetation index NDVI (V) and surface temperature LST (T) using a MODIS product; acquiring elevation data using an ASTER product; and calculating the elevation data obtained by ASTER products to obtain gradient, slope direction, plane curvature, section curvature, surface roughness, humidity index and fluctuation amplitude data;
step 3, obtaining downscaled FY3-B soil moisture data SM by using the step 1 1 Step 2, obtaining environment factor soft data, and obtaining probability soft data by adopting a histogram probability distribution form;
step 4, determining the weight value of the environment factor soft data in the step 2 by using a multivariate correlation analysis method and a principal component analysis method, and obtaining weighted probability soft data;
first, calculate the firstjQuantitative probability between individual environmental factors and soil moisture data, denoted P SD （EF j ）：
（3）
Wherein Ef represents the environmental factor data obtained in step 2;nrepresenting SMnEach interval;krepresent SM's firstkIn each of the intervals of time,ktake the value from 1 ton； mRepresenting EfmEach interval;is SM in the first interval of EfkThe number of sampling points in each interval; />Is SM. ThkThe total number of sampling points in each interval; ef (Ef) m Is the firstmAn environmental factor for each interval; in each interval, the probability distribution data is a constant value, and the sizes of all the intervals are the same;
second, calculate the firstjWeight value w of individual environmental factor data j The method comprises the following steps:
（4）
wherein,ifor the number of environmental factors in step 2,itake values from 1 to 11, rj is the firstjCorrelation coefficients between the individual environmental factors and the FY3B soil moisture data; ef (Ef) j Is the firstjEnvironmental factors; NCC (Ef) j ) Is the firstjNormalized correlation coefficient of each environmental factor, the sum of NCC of all environmental factors is equal to 1, wherein NCC (Ef j ) The environmental factors with the values larger than or equal to 0.18 are taken as dominant environmental factors, the corresponding weight values of the environmental factors are reserved, NCC (Ef j ) An environmental factor weight value of less than 0.18 is 0;
again, the weighted probability of SM is calculated and denoted as W PSD ，
（5）
Step 5, obtaining soil moisture hard data through actual measurement
Collecting soil samples with the depth of 0-5cm in the same area and on the same date in the step 1, and measuring the moisture content of the soil by using an aluminum box baking method; all parameters were measured three times and the average was used for subsequent analysis and recorded as；
Step 6, weighting probability soft data of soil moisture in the step 4 by adopting a Bayesian maximum entropy methodAnd the measured data in step 5 +.>Fusion was performed as hard data, soil moisture data was estimated, and the result was recorded as +.>；
Step 7, evaluating soil moisture prediction using Root Mean Square Error (RMSE) and correlation coefficient (R)Accuracy of the results, wherein
(6)
(7)
Wherein,nis the total number of times of actually measuring the soil moisture in the step 5,is the first step in step 5iSecondary actual measurement of soil waterScore of->Is in step 5nAverage value of soil moisture measured for the second time,/->Is the first step in step 6isub-Bayes maximum entropy soil moisture, ++>In step 6nMean value of sub-bayesian maximum entropy soil moisture.
Further, the number of the weighted probability soft data is set to be 4 types, namely 500,450,400 and 350, and soil moisture estimation results under the four types of different weighted probability soft data are obtained.
The method adopts hard data and soil moisture weighted probability soft data to estimate soil moisture. According to the method, actually measured soil moisture data are used as hard data, data such as topography and the like are selected for soft data, and influences of topography and the like on soil moisture are considered. The invention has the advantage of integrating multi-source remote sensing data into a BME algorithm to estimate SM with higher resolution and higher accuracy.
Drawings
FIG. 1 is a schematic diagram of a Bayesian soil moisture estimation method based on multi-source remote sensing data;
FIG. 2 study area overview;
FIG. 3 spatial prediction of SM content for three stages of corn growth based on MCA and PCA methods;
FIG. 4 is a comparison of estimated SM and field measurements SM for a BME algorithm and downscaled FY3-B SM product.
Detailed Description
The aim of this study is to estimate SM with higher resolution and higher accuracy by integrating multisource telemetry data into the BME algorithm, the process is schematically shown in fig. 1, and the specific embodiment is as follows:
1. hard data preparation
The study area (FIG. 2) was located in Hebei Hemiyao, hebei province, and in northwest China (38℃3'00"N,115℃27'54" E). This is a typical agricultural area with only a few land cover (bare soil, corn, orchards and grasslands) and a uniform soil texture. The area belongs to a typical continental monsoon climate, has high air temperature in summer and abundant rainfall, and is cold and dry in winter. Therefore, SM content plays an important role in agricultural production, especially in summer. To measure SM content, field trials were performed in summer. The sample points are shown in fig. 2. Corn, cotton and peanut are the primary crops in these sites. The main soil types are loams, including approximately 44.6% silt, 17.5% clay and 37.9% sand.
In summer, corn represents approximately 50% of the total area studied. In this study, corn is classified into spring corn and summer corn. Spring corn is sown in the last ten days of 4 months or 5 months and harvested in the last 10 days of 8 months. Summer corn is typically sown in the late 5 months or in the early 6 months and harvested in the 10 months. Thus, experiments were performed from 6 months to 9 months. This period includes sowing of corn growth, heading, flowering and maturity.
Experiments were performed in 2014 on day 27 of 6 months (sowing stage, 178 days), day 14 of 8 months (heading stage of corn, 226 days) and day 7 of 9 months (flowering stage, 249 days). We have chosen an area of about 25km x 25km as the key sample area. The sample area consisted of 23 sample points, and parameters for each site included SM, soil permittivity, vegetation water content and Leaf Area Index (LAI). Soil samples were collected at a depth of 0-5cm to measure SM content using an aluminum box. All parameters were measured three times and the average was used for subsequent analysis. In this study, in situ actual measurement SM data was split into two parts. A portion of in situ actual measurement SM data was used as HD for BME analysis. The other is used to verify the estimated SM from the BME algorithm. Soil dielectric constants were measured in the laboratory using E5071C Vector Network Analyzer (Keysight Technologies inc., santa Rosa, CA, USA). The relevant length of the coarse mean square height and roughness parameters was measured using a needle plate containing 101 needles and having a total length exceeding 1 m. Vegetation moisture content (VWC) was measured using a sampling and drying method. LAI was measured five times using a LAI-2200C plant canopy analyzer (LI-COR Biosciences, lincoln, NE, USA). Ground data, including SM content, soil permittivity and vegetation water content, are shown in table 1. )
TABLE 1 ground data summary
DO N SMC (g/cm 3 ) VWC (kg/m 2 ) H (cm) LAI
Day 178 Bare soil/grassland/orchard 64 5.62–48.1 0 2.3–20.1 1.578 0.4–3.4 /
Day 226 Corn field/orchard 60 9.1–40.9 0.48–5.13 3.7–25.0 1.33 0.2–3.0 2.89
Day 249 Corn field/orchard 60 8.6–39.3 1.53–3.90 3.4–17.8 1.33 0.4–2.2 3.35
( N: sampling points, Ɛ r: soil dielectric constant: average soil volume weight, VWC: vegetation water content, H: average corn height; DO: main object )
2. Multisource remote sensing soft data acquisition
Factors influencing SM retrieval are numerous, such as surface temperature, vegetation canopy moisture content and roughness. Traditionally, SM is estimated using mostly 2-3 kinds of auxiliary data, or the influence of topography and topography is not considered, or the soil moisture after downscaling is not considered. Therefore, it is difficult to evaluate SM using only one type of data. Therefore, we combine FY3-B SM, NDVI-LST-Albedo MODIS (NDVI, albedo and LST products of MODIS) and ASTER DEM satellite product data at 25km resolution to estimate the spatial distribution of SM. In this study we used three MODIS products, NDVI, albedo and LST products, with a resolution of 1km. Advanced satellite borne thermal radiation and reflected radiometers global digital elevation model (ASTER DEM) with spatial resolution up to 30 m. FY3-B (third wind cloud B satellite) SM product at 25km resolution, these assistance data were used to develop higher resolution SM data using BME algorithm.
FY3-B SM product is free (download link http:// satellite. Nsmc. Org. Cn/PortalSite/default. Aspx). The product was developed based on the SM retrieval algorithm of AMSR-E (advanced microwave scanning radiometer-earth observation system) sensors on Coriolis satellites for horizontal and vertical polarization data of 10.65 GHz data. In the AMSR-E algorithm, the Q/H model is selected as the roughness model to estimate the effect of surface roughness.
The three MODIS products are MOD11A1 (surface temperature product, LST), MCD43B3 (Albedo product) and MOD13A2 (vegetation index product: normalized vegetation index product and enhanced vegetation index, NDVI and EVI), all with a spatial resolution of 1km. The MOD11A1 LST product was retrieved at 1km resolution using the MODIS 31 st and 32 nd band split window algorithm (Yang et al 1199). The MCD43B3 Albedo product was produced from a 16 day anisotropic model, representing the average of the underlying 500m values of the MCD43A Albedo product. MOD13A2 NDVI products are retrieved from daily, atmospheric corrected bi-directional surface reflections based on product quality assurance metrics by using the specific synthetic method of MODIS, removing low quality pixels and maintaining high quality NDVI values. ASTER DEM is based on detailed observations of the NASA new generation earth observation satellite tera. This is currently one of the most complete global digital elevation Data (DEM) with a resolution of 30 meters. In this study, 30 meters of arter DEM data were resampled to 1km resolution.
Using the ASTER DEM data, 7 data such as gradient, slope direction, plane curvature, section curvature, surface roughness, humidity index, and relief amplitude were extracted (table 2). The study uses LST, albedo, NDVI, ASTER DEM and 7 types of data generated by ASTEM DEM as environmental factor data, and directly influences SM estimation. NDVI-LST-Albedo/MODIS, FY3-B SM and ASTER DEM data were used for nearly 3 epochs: three periods of fertility of corn (day 178,226 and 249) are shown, namely, day 27 of 2014, day 14 of 2014, and day 9 of 2014.
Table 2 seven topographical data (slope, slope direction, plane curvature, section curvature, surface roughness, humidity index and relief amplitude) generated from ASTER DEM data
Name of the name Description of the algorithm
Slope direction And a downhill direction with the greatest rate of change from each pixel to its neighboring pixels.
Gradient of slope The slope is the second derivative of the digital elevation model change.
Curvature of plane It is two of the surfaces perpendicular to the direction of maximum slopeAnd (5) an order derivative.
Curvature of section It is the second derivative of the surface along the direction of maximum slope.
Surface roughness (sr) ,Is the slope in radians. SR is the ratio of the earth's surface area to its projected area in a given area.
Humidity index (WI) As represents the area of confluence per unit of contour length flowing through a point on the surface. Is a gradient.
Amplitude of fluctuation ,(DEMmax)nIs thatn×nMaximum in the region; (DEMIN)nIs thatn×nMinimum in the region. In the course of this study, the test,nset to 3.
3. Bayesian soil moisture estimation and verification
Some researchers have shown that the development of synergistic methods using microwave optical/infrared products depends on the schematic relationship between SM, NDVI and LST. Many studies have used MODIS satellite image data to obtain SM information at a finer resolution (1 km) than the FY3-B original resolution (25 km).
In this study, MODIS albedo (A), NDVI (V) and LST (T) were matched to 25km FY3-B SM spatial resolution. SM was obtained at a spatial resolution of 1km using a linear regression model. FY3-B SM, MODIS albedo, NDVI and LST were used for linear regression analysis, which correlated the intensities of the microwave and optical/IR parameters to increase the FY3-B spatial resolution of the coarse resolution. The MODIS albedo, resolution of NDVI and LST at 25km is calculated as follows:
(1)
wherein A is 25 ，V 25 And T 25 The albedo, average value of 25km for NDVI and LST,mandnrespectively the first of FY3-B pixelsiLine and thjNumber of 1km pixels in a column.
The linear regression model was used as follows:
(2)
wherein SM is 25km SM, a1, a2, a3, a4, a5, a6 and a7 are regression coefficients; regression equations for the study area were estimated using 4-10 month time series clear sky data. Then, an estimated regression model was applied, including regression coefficients, estimated 25km FY3-B,1km albedo, 1km SM for NDVI and LST. While each regression coefficient (i.e., a1, a2, a3, a4, a5, a6, and a 7) remains unchanged for each FY3-B pixel on a given date, each value varies between different dates because all variables vary over time and the regression relationship changes based on these variables. However, this approach ignores the importance of SM ground data. Therefore, it is necessary to combine the ground data to obtain a higher precision digital map of the SM.
In order to carry out probability statistics on each environmental factor data, the same discretization method is adopted to model each environmental factor dataThe circumference is also divided intomIntervals. And combining the discrete SM and the environmental factor data, and deducing the probability SD of the SM by calculating the quantitative probability between the discrete environmental factor and the SM data in each interval. Between the obtained SM and the environmental factorm×nThe quantitative probability relationship is the probability SD (i.e., P SD (Ef))。
（3）
Where Ef represents one of the environmental factors, namely NDVI, LST, albedo, DEM, slope direction, planar curvature, cross-sectional curvature, surface roughness, humidity index and relief amplitude in this study.nRepresenting SMnEach interval;krepresent SM's firstkIntervals of 1 ton； mRepresenting EfmEach interval; l represents the first interval of SM from 1 tom；Is SM in the first interval of EfkNumber of samples in each interval. />Is SM. ThkThe total number of sampling points in each interval; ef (Ef) m Is the firstmEnvironmental factors of individual intervals. In each section, the probability distribution data is a constant value, and the sizes of all sections are the same. This probability distribution form is called histogram probability distribution form.
Since different environmental factors have different effects on SM content, weighting values for different environmental factors should be specified. The weighting values are normalized using equation 10. Assuming that the number of environmental factors is i, the weighted probability SD of SM (i.e., W) is obtained using the following equation PSD ）。
（5）
Where W1 is the weight of the first environmental factor and Wj is the firstjPersonal environmental factorIs a weight value of (a). In this study, the two methods MCA (multivariate correlation analysis) and PCA (principal component analysis) methods were used to calculate the weight values of the different environmental factors. The normalized correlation coefficients of different environmental factors are considered as weight values.
The MCA method is a multivariate linear regression model:
correlation coefficient normalization equation for the environmental factor (R):
（4）
SM is SM content, which in this study represents downscaled FY3-B SM; r1 is the correlation coefficient of the first environmental factor; ef1 is the first environmental factor; NCC (Ef) j ) Is the firstjNormalized correlation coefficient of each environmental factor. In this study, the environmental factor data were helper data, i.e., NDVI-LST-Albedo/MODIS product data, ASTER DEM data, slope direction, plane curvature, cross-sectional curvature, surface roughness, humidity index, and relief amplitude, for a total of 11 helper data (i.e., i=11). The sum of NCCs for all environmental factors is equal to 1. The NCC value may be considered as a weight value of the environmental factor to produce the weighted probability SD.
The number of environmental factors is reduced by adopting a Principal Component Analysis (PCA) method. In this approach, the multivariate factors are reduced to a plurality of principal components, simplifying modeling and yielding more robust results. First, 11 environmental factor data affecting soil moisture recovery were processed by PCA to obtain 11 main components. Ef (Ef) i The data is not normalized environmental factor data in eq.11, but is the main component after PCA. NCC values of 11 principal components were calculated. Several principal components highly related to SM were then selected as dominant factors for the different corn growth stages. Finally, the NCC values of these dominant factors are regarded as weight values for obtaining the weighted probabilities SD.
The prior probability of SM content can be modified using HD according to a conditional probability formula of bayesian maximum entropy. The parameter z is defined as the posterior probability in predicting the position of x 0. z 0 The posterior probability of (a) is expressed as
Where h is HD point number; m-h is SD number; SM estimation was developed using space-time cognitive knowledge integrated graphical user interface (SEKS-GUI) software. This is a freely available software package that builds a spatiotemporal model and predicts and maps parameters by combining scientific software libraries, bayesian maximum entropy libraries, and related GUI files. And predicting the SM content by adopting a Bayesian maximum entropy method in combination with the environmental factor data of FY3-BSM inversion products, MODIS and ASTERDED products. The probability SD adopts a histogram probability distribution form to discretize the SM content and the environmental factors. The post PDF of SM was developed in SEKS-GUI software, i.e. the weighted probability SD of SM was calibrated with HD data (i.e. field SM measurements).
In this study, two statistical indicators were used to evaluate the accuracy of SM predictions: RMSE (root mean square error), i.e., equation 6 and R (correlation coefficient), i.e., equation 7.
(6)
(7)
Wherein,is in situ actual measurement SM; />SM predicted using BME algorithm;nis the total number of in situ actual measured SMs.
In this example, correlation coefficients of FY3-B SM downscaling and 11 auxiliary data are calculated, including gradient, slope direction, plane curvature, section curvature, surface roughness, humidity index and fluctuation amplitude product generated by NDVI-LST-Albedo/MODIS and ASTER DEM data. (Table 2).
Correlation coefficients and NCC (normalized correlation coefficients) were calculated at three stages of corn growth as shown in table 3. In addition to day 226 (heading period of corn), the strongest correlation between NDVI and SM content. NCC is highest at heading time (0.68) and lowest at sowing time (0.427). On day 226, the correlation between LST and SM content was the strongest (NCC: -0.718). Thus, the dominant factors vary depending on the growth stage. At the sowing stage (day 178), LST, NDVI, plane curvature and section curvature (absolute value of NCC. Gtoreq.0.372) are the main factors of SM retrieval. LST, NDVI, albedo and profile curvature (absolute NCC. Gtoreq.0.395) are dominant factors during heading (day 226). At the flowering time (day 249), NDVI, cross-sectional curvature, surface area and moisture index (NCC absolute value. Gtoreq.0.198) are the major factors.
TABLE 3 correlation coefficient between downscaled FY3-B SM and Ef
Ef R/178 th NCC/178 th R/226 th NCC/226 th R/249 th NCC/249 th
Albedo of / / 0.454 -1.767 0.010 0.010
LST -0.355 2.241 -0.718 2.795 -0.070 -0.073
NDVI 0.427 -2.697 0.680 -2.647 0.580 0.604
DEM -0.003 0.020 0.259 -1.008 -0.180 -0.188
Surface roughness 0.006 -0.035 0.023 -0.090 -0.010 -0.010
Amplitude of fluctuation -0.116 0.733 0.024 -0.093 -0.050 -0.052
Humidity index 0.092 -0.582 -0.269 1.046 0.190 0.198
Curvature of plane -0.383 2.418 0.210 -0.817 0.030 0.031
Gradient of slope -0.019 0.117 0.014 -0.055 0.010 0.010
Slope direction -0.180 1.134 -0.085 0.332 0.190 0.198
Curvature of section 0.372 -2.349 -0.395 1.536 0.260 0.271
（Ef: environmental factor, R: correlation coefficient, NCC: normalized correlation coefficient, day 178: corn sowing time, 226 th day: corn heading period, 249 th day: corn flowering period
Several major components highly related to SM were then selected as dominant factors for the different corn growth stages. The NCCs of these dominant factors are considered as weight values of these dominant factors as shown in table 4. From Table 4, on day 178, PC2, PC4, PC8 and PC9 were selected as the major factors (NCC. Gtoreq.0.181). On day 226, PC2, PC6, PC8 and PC9 were selected as dominant factors, NCC. Gtoreq.0.243. On day 249, PC2, PC4, PC8 and PC10 were selected as the dominant factors, and NCC was ≡ 0.593 (Table 4).
TABLE 4 reduction of FY3-B SM and after PCAEfCorrelation coefficient between
Ef R/178 th NCC/178 th R/226 th NCC/226 th R/249 th NCC249 th
PC1 -0.146 0.114 -0.137 0.327 0.019 0.059
PC2 -0.507 0.394 -0.637 1.522 -0.193 -0.593
PC3 -0.003 0.003 -0.044 0.104 0.053 0.162
PC4 -0.257 0.200 0.072 -0.171 0.281 0.862
PC5 -0.076 0.059 -0.034 0.082 -0.044 -0.136
PC6 0.056 -0.044 0.204 -0.488 0.069 0.211
PC7 -0.058 0.045 -0.029 0.070 -0.050 -0.153
PC8 -0.392 0.306 0.102 -0.243 0.199 0.609
PC9 0.233 -0.181 0.139 -0.332 0.156 0.480
PC10 -0.134 0.104 -0.087 0.209 -0.260 -0.797
PC11 / / 0.034 -0.080 0.096 0.294
（Ef: environmental factor, R: correlation coefficient, NCC: normalized correlation coefficient, day 178: corn sowing time, 226 th day: corn heading period, 249 th day: jadeRice blooming period
Typically, two types of data (HD and SD) are used in BME analysis. HD refers to certain data or facts with small errors and high accuracy, such as in-situ data, historical data with high accuracy. In this context, HD comes from SM content measured in three field experiments. HD, i.e. in situ SM data, is used to estimate posterior probability density distribution in the posterior stage of the BME algorithm in SEKS-GUI software. From a cognitive perspective, our understanding of something does not depend entirely on certain data or on certain facts, but also on uncertain information such as remote sensing data with some errors, expert opinion and empirical knowledge. These data are referred to as SD in BME analysis. Downscaled FY3-B SM data and auxiliary data with some errors can be considered SD. The auxiliary data comprise seven types of data obtained by NDVI-LST-Albedo/MODIS, ASTER DEM data and ASTER DEM data, namely gradient, slope direction, plane curvature, section curvature, surface roughness, humidity index and fluctuation amplitude. Since the number of probabilities SD may affect the probability SD, different probability SD points are set in 4 cases (table 5), case 1 is 500 soft data points, case 2 is 450 soft data points; case 3 is 400 soft data points; case 4 is 350 soft data points (table 5).
TABLE 5 HD points and probability SD points
( SD: soft data; HD: hard data; day 178: corn sowing time; day 226: corn heading period; day 249: flowering period of corn )
In this study, the MCA and PCA methods were used to obtain weight values for different environmental factors, respectively (tables 3 and 4). Thus, 2 (two methods calculate the weighing values, MCA and PCA) ×3 (three growth phases of corn) ×4 cases (table 5), a total of 24 probabilistic SD datasets were estimated in this study.
The SEKS-GUI software has a user friendly interface and is built with BMElib (BME library) written in Matlab. SM content profiles at three corn growth stages (sowing stage, day 178, heading stage, day 226, flowering stage, day 249) are shown in figure 3. In fig. 3, the first four columns represent four cases, case 1 through case 4 (see table 5). The last column represents downscaled FY3-B SM on day 178 (corn sowing period), day 226 (heading period of corn) and day 249 (flowering period of corn). In each case, 6 profiles of SM estimates generated using MCA and PCA methods at three stages of corn growth (sowing, heading and flowering) are shown.
In this study, based on two methods (MCA and PCA), the spatial distribution of SM estimated using the BME algorithm was 1km resolution for calculating the weight values of the environmental factors. MCA and PCA methods are implemented using IDL (interactive data language) programs. In each pixel region (1 km×1 km), we assume that the estimated value of SM is uniform. As can be seen from fig. 3, the BME algorithm for three corn growth, sowing, heading and flowering phases of the study area estimated the spatial profile of SM. At the bottom of each SM profile there are some blocks, especially case 4, in which the estimated SM is not very good. For example, at the heading date of corn, i.e., day 249, the predicted SM using the MCA method is the large yellow patch at the bottom of the SM profile, especially case 4. From case 1 to case 4 (the number of soft data points is from 500 to 350 on day 249 of MCA), the number of soft data points is smaller and the area of the yellow patch is larger. On other days (178 or 226 days), the smaller the amount of soft data, the larger the area of the block. The reason for the block generation may be that there is a missing value of probability soft data or a value lacking auxiliary data in the block area. In each epoch (each column of fig. 3), the spatial structure of the estimated SM for the 4 cases is similar, and they have similar SM ranges of variation. On day 178, the estimated SM values for the 4 cases varied from 0.08 to 0.3 m3/m3, with a smaller number of SM pixels when the SM value was less than 0.012 m3/m3. On day 226, the estimated SM values for case 4 varied from 0.08 to 0.36 m3/m3. On day 249, the estimated SM values for the 4 cases varied very narrowly, ranging from 0.12 to 0.26 m3/m3. Clearly, a large spatial variation of the estimated SM value occurred at day 226. On days 178 and 249, the SM appeared more uniform. The reason for the homogeneous SM on days 178 and 249 may be that the vegetation is more uniform. Comparing the estimated SM based on the MCA method with the PCA method, the spatial structure of the estimated SMA-based PCA is more consistent in 4 cases. The spatial distribution characteristics between estimated SMs from MCA and PCA methods show a preliminary implementation. Therefore, quantitative verification work of estimating SM was performed in the present study.
In situ actual SM measurements are split into two parts. A portion of in situ actual measurement SM data (88 total points) was used as hard data. The other (54 points total) is used to verify the estimated SM from the BME algorithm. The relationship between in situ actual measurements SM and SM estimated from the BME algorithm at three stages of corn growth (178,226 and 249) is shown in fig. 4. The X-axis of fig. 4 is in-situ. The Y-axis of fig. 4 is the estimated SM from the BME algorithm as a result of the measurement. Although there are problems with overestimation and underestimation in fig. 4, most points are distributed at 1: near line 1.
Furthermore, we performed an absolute validation effort with the in-situ actual measurement SM by calculating RMSE and R values for each case (case 1, case 2, case 3 and case 4) for each method (MCA and PCA). RMSE and R values were calculated using in situ actual measurements at 15, 16 and 23 points, respectively, over three time periods (day 178,226 and 249). The calculated RMSE and R values are shown in table 6.
As can be seen from Table 6, using the PCA method, in case 1, the RMSE value on day 178 was the smallest, which was 0.049 m3/m3. It is less than 0.079m3/m3 RMSE value, which is calculated between the downscaled FY3-B SM and the field measurement SM. On day 178 of the PCA used in case 1, the maximum R value was 0.639. It is well above the 0.096R value, which is calculated between the downscaled FY3-B SM and the field measurement SM. At day 226, the minimum RMSE value produced in case 2 and case 3 using the MCA method was 0.062m3/m3, less than 0.091m3/m3 produced by the scaled-down FY3-B SM. The maximum value of the R value generated in case 1 using the MCA method was 0.53. On day 249, the downscaled FY3-B SM produced a minimum RMSE value of 0.051 m3/m3. During this time, the lower RMSE value of the estimated SM from the BME algorithm was 0.053 m3/m3, which is higher than the 0.051 m3/m3 produced by the downscaled FY3-B SM. In summary, we conclude that the accuracy of the estimated SM for the BME algorithm is better than the downscaled FY3-B SM with the PCA method, with minimum RMSE value and maximum R value of 0.049 m3/m3 and 0.639, respectively. Comparing the accuracy of estimated SM based on MCA and PCA methods, we found that on days 178 and 249, the RMSE value of estimated SM based on PCA method was smaller than that based on MCA, and the R value based on PCA method was higher than that based on MCA method. However, on day 226, the estimated SM based on the PCA method has a RMSE value greater than that based on the MCA method, and the R value based on the PCA method is lower than that based on the MCA method. Furthermore, on day 226, the accuracy of estimating SM was significantly lower than on days 178 and 249. The cause may be severe vegetation effects. Therefore, estimated SM based on days 178 and 249 of the PCA method is best, except for 249, compared to MCA based SM.
TABLE 6 estimation of RMSE and R values for SM using BME algorithm and reduction of downscaling of in situ SM measurements by FY3-B SM
5. Conclusion(s)
SM content was estimated at three stages of corn growth using the BME algorithm in SEKS-GUI software, including three procedures in this study, the first: HD and probability SD preparation; and a second step of: SM estimation; third procedure: verification of the estimated SM. In the first procedure, MCA and PCA are used to calculate weight values for environmental factors. The probability SD is then generated by using the downscaled FY3-B and the helper data (i.e., NDVI-LST-Albedo/MODIS, ASTER DEM, slope, aspect ratio, planar curvature, cross-sectional curvature, surface roughness, histogram probability distribution form). Humidity index and ASTER DEM produced relief amplitude products of 4 cases during three stages of corn growth. Finally, 24 probabilistic SD data sets were obtained. In the second process, SM content is estimated in SEKS-GUI software from the HD data and probability SD generated in the first process using a BME algorithm. In the third approach, the estimated SM is evaluated for in situ SM measurements (RMSE and R, equations 13-14).
Based on the MCA and PCA methods, the main factors depend on the growth stage of corn in tables 3 and 4. During corn sowing, LST, NDVI, planar curvature and cross-sectional curvature predominate. During the heading stage, LST, NDVI, albedo and profile curvature predominate. NDVI and profile curvature dominate during anthesis. NDVI is one of the main factors of all growth phases and should be regarded as an important parameter for SM spatial prediction. Second, PCA is an effective spatial dimension reduction method that can reduce data from 11 to only 4 principal components. Based on the SM estimation results using the BME algorithm (fig. 3 and table 6), estimated SM based on days 178 and 249 of PCA was optimal. On day 226, MCA-based estimated SM was superior to SM based on PCA method. Furthermore, we conclude that SM estimated using the BME algorithm (e.g., case 1 based on PCA, rmse=0.049 m3/m3, r=0.639) is consistent with field measurement SM, rather than downscaling FY3-B SM (rmse=0.079 m3/m3, r=0.096).
In this study, we applied three different procedures to estimate and evaluate SM using the BME algorithm. In a first step we have prepared to use the BME algorithm in SEKS-GUI software to estimate HD and probability SD of SM. In situ SM measurement data was used as HD data. Probability SD is developed from downscaled FY3-BSM, NDVI-LST-Albedo/MODIS products, ASTER DEM products, and gradient, slope direction, planar curvature, section curvature, surface roughness, humidity index, and heave amplitude data based on MCA (multivariate correlation analysis) and PCA (principal component analysis), respectively. Wherein the grade, slope direction, plane curvature, section curvature, surface roughness, humidity index and relief amplitude data are from the ASTER DEM product. MCA and PCA were used to obtain weight values for different environmental factors (i.e., NDVI, LST, albedo, BME, slope direction, plane curvature, cross-sectional curvature, surface roughness, humidity index, and heave amplitude), and the study employed a histogram probability distribution as a distribution form of the probability standard deviation. This distribution form assumes that the probability distribution data for each interval is constant and that the size of all intervals is the same. The study used two methods (MCA and PCA) to test three growth periods of 4 maize, 24 probability standard deviation data sets were obtained. In the second procedure, the HD and the probability SD are used as input data for SEKS-GUI software to estimate the space SM. On this basis, a map of 16 SM estimates was obtained using BME. In the third procedure, the SM estimated using the BME algorithm is verified by field SM measurements using equations 13-14.
In this study, since the accuracy of scaling down FY3-BSM using regression models is not high, the downscaled FY3-BSM with errors uses only SD to generate probability SD with assistance data (i.e., NDVI-LST-Albedo/MODIS), ASTER DEM and grade, aspect ratio, planar curvature, cross-sectional curvature, surface roughness, humidity index, and relief amplitude product generated by ASTER DEM. The histogram probability distribution form is regarded as a form of probability SD distribution in which probability distribution data is assumed to be a constant value in each bin and the sizes of all bins are the same. In the generation of the weighted probability SD, we use both the Multivariate Correlation Analysis (MCA) method and the Principal Component Analysis (PCA) method to determine the weights between the helper data. In wear, since the probability distribution is affected by the number of sample points, we use four cases 4 (case 1:500 soft data points, case 2:450 soft data points, case 3:400 soft data points, case 4:350 soft data points). HD as deterministic value is used to correct the probability SD of SM to obtain a probability post distribution of SM, i.e. to obtain a higher accuracy probability of SM. In this study, in situ actual measurement SM was considered HD, which was used to calibrate probability soft data to obtain a probability post distribution of SM. In procedure 2, by using the BME algorithm (space-time cognitive knowledge synthesis graphical user interface) in the SEKS-GUI, the SM was estimated to be in 3 corn growth phases in 4 cases based on the probability SD of the two weight determination methods of HD (MCA and PCA) and SM. Software, which is a freely available software package, is currently a combination of the scientific software library BMElib (bayesian maximum entropy library) and the GUI (graphical user interface).
The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention, and any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of the present invention.

Claims (1)

1. A Bayesian soil moisture estimation method based on multi-source remote sensing data is characterized in that: the method comprises the following steps:
step 1, obtaining soil moisture soft data and performing data processing
Acquiring soil moisture data of a designated area and a designated date by using an FY3-B soil moisture monitoring product; performing downscaling treatment on the FY3-B soil moisture data, and improving the spatial resolution of the FY3-B soil moisture data with the resolution of 25km to the soil moisture data with the resolution of 1km by using a regression model;
wherein A is 25 For an average value of 25km for albedo (A), V 25 Is the average value of 25km, T, of vegetation index NDVI (V) 25 For a 25km average of the surface temperature LST (T), m and n are the number of 1km pixels in the ith row and jth column, respectively, in FY3-B pixels; a is that ij Is the albedo value of 1Km pixel in the ith row and jth column in FY3-B, V ij Is the NDVI value, T, of 1Km pixel in the ith row and jth column of FY3-B ij Is the surface temperature LST value of 1Km pixel in the ith row and jth column in FY 3-B;
the linear regression model was: sm=a 1 +a 2 A+a 3 V+a 4 T+a 5 AV+a 6 AT+a 7 VT (2) wherein SM is soil moisture; a, a 1 、a 2 、a 3 、a 4 、a 5 、a 6 And a 7 Is a regression coefficient; a is albedo, V is vegetation index, and T is surface temperature;
first, A calculated by equation (1) 25 、V 25 And T 25 Carrying out formula (2) according to FY3-B soil moisture data SM corresponding to pixel scale 25km resolution 25 Calculating regression coefficients in the model by regression analysis;
secondly, taking the albedo (A) of 1km resolution of the MODIS product, the vegetation index NDVI (V) and the surface temperature LST (T) into a model formula (2), and correspondingly calculating to obtain a soil moisture value SM of 1km resolution 1 ；
Step 2, obtaining environment factor soft data
Obtaining albedo (a), vegetation index NDVI (V) and surface temperature LST (T) using a MODIS product; acquiring elevation data using an ASTER product; and calculating the elevation data obtained by ASTER products to obtain gradient, slope direction, plane curvature, section curvature, surface roughness, humidity index and fluctuation amplitude data;
step 3, obtaining downscaled FY3-B soil moisture data SM by using the step 1 1 Step 2, obtaining environment factor soft data, and obtaining probability soft data by adopting a histogram probability distribution form;
step 4, determining the weight value of the environment factor soft data in the step 2 by using a multivariate correlation analysis method and a principal component analysis method, and obtaining weighted probability soft data;
first, calculate the quantitative probability between the jth environmental factor and the soil moisture data, record as P SD (EF j )：
Wherein Ef represents the environmental factor data obtained in step 2; n represents n intervals of SM; k represents a kth interval of SM, and the value of k is from 1 to n; m represents m intervals of Ef; count (l) k Is the number of samples in the kth interval of SM in the ith interval of Ef; count (Count) k Is the total number of sampling points in the kth interval of SM; ef (Ef) m Is the environmental factor of the mth interval; in each interval, the probability distribution data is a constant value, and the sizes of all the intervals are the same;
next, a weight value w of the jth environmental factor data is calculated j The method comprises the following steps:
wherein i is the number of environmental factors in the step 2, i takes values from 1 to 11, and Rj is the correlation coefficient between the j-th environmental factor and FY3B soil moisture data; ef (Ef) j Is the jth environmental factor; NCC (Ef) j ) Is the normalized correlation coefficient of the jth environmental factor, the sum of the NCCs of all environmental factors being equal to 1, where NCC (Ef j ) The environmental factors with the values larger than or equal to 0.18 are taken as dominant environmental factors, the corresponding weight values of the environmental factors are reserved, NCC (Ef j ) An environmental factor weight value of less than 0.18 is 0;
again, the weighted probability of SM is calculated and denoted as W PSD ，
WP SD ＝W 1 *P SD (Ef 1 )+...+W j *P SD (Ef j )+....+W i *P SD (Ef i ) (5)
Step 5, obtaining soil moisture hard data through actual measurement
Collecting soil samples with the depth of 0-5cm in the same area and on the same date in the step 1, and measuring the moisture content of the soil by using an aluminum box baking method; all parameters were measured three times and the average was used for subsequent analysis, recorded as SM in-situ ；
Step 6, weighting probability soft data WP of soil moisture in the step 4 by adopting a Bayesian maximum entropy method SD Actual measurement data SM in step 5 in-situ Fusion is carried out as hard data, soil moisture data is estimated, and the result is recorded as SM prediction ；
Step 7, evaluating soil moisture prediction SM using Root Mean Square Error (RMSE) and correlation coefficient (R) prediction Accuracy of the results, wherein
Wherein n is the total number of times of actually measuring soil moisture in step 5, SM i in-situ Is the soil moisture value measured in the ith time in the step 5,is the average value of the soil moisture measured n times in the step 5, SM prediction Is the (i) th Bayes maximum entropy soil moisture in step 6,/H>In the step 6, the average value of the maximum entropy soil moisture of the Bayes is obtained for n times;
and (3) setting the number of weighted probability soft data in the step (6) to be 4 types, namely 500,450,400 and 350, respectively, and obtaining soil moisture estimation results under the four types of different weighted probability soft data.
CN201910671852.6A 2019-07-24 2019-07-24 Bayesian soil moisture estimation method based on multi-source remote sensing data Active CN110427995B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910671852.6A CN110427995B (en) 2019-07-24 2019-07-24 Bayesian soil moisture estimation method based on multi-source remote sensing data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910671852.6A CN110427995B (en) 2019-07-24 2019-07-24 Bayesian soil moisture estimation method based on multi-source remote sensing data

Publications (2)

Publication Number Publication Date
CN110427995A CN110427995A (en) 2019-11-08
CN110427995B true CN110427995B (en) 2023-11-17

Family

ID=68412139

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910671852.6A Active CN110427995B (en) 2019-07-24 2019-07-24 Bayesian soil moisture estimation method based on multi-source remote sensing data

Country Status (1)

Country Link
CN (1) CN110427995B (en)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113033262B (en) * 2019-12-25 2022-08-05 中移(成都)信息通信科技有限公司 Model training method and crop yield estimation method
CN111089827B (en) * 2020-01-07 2021-04-06 华南农业大学 Measuring system and measuring method for water penetration of citrus garden soil
CN111651411B (en) * 2020-04-21 2023-03-10 成都信息工程大学 Complex terrain remote sensing soil moisture product downscaling method
WO2021226977A1 (en) * 2020-05-15 2021-11-18 安徽中科智能感知产业技术研究院有限责任公司 Method and platform for dynamically monitoring typical ground features in mining on the basis of multi-source remote sensing data fusion and deep neural network
CN111861222B (en) * 2020-07-22 2023-11-14 中国水利水电科学研究院 Method for obtaining roughness of cultivated land and grassland facing regional scale wind erosion
CN112560570B (en) * 2020-09-29 2023-07-25 中国科学院大学 High-resolution earth surface temperature estimation method for collaborative downscaling and data fusion
CN112528555B (en) * 2020-11-23 2023-02-21 中国科学院空天信息创新研究院 Soil moisture map drawing method and device
CN113419046B (en) * 2021-06-17 2022-02-18 北京大学 Improved soil humidity product bivariate fusion method
CN113763409B (en) * 2021-08-27 2023-05-02 南通大学 Gaussian filtering-based multi-threshold segmentation method for high-dimensional brain nuclear magnetic image
CN114461983B (en) * 2022-02-09 2023-03-21 中国科学院、水利部成都山地灾害与环境研究所 Satellite precipitation product space downscaling method based on water balance principle
CN115204479A (en) * 2022-07-01 2022-10-18 南京信息工程大学 Crop survival area prediction method and system based on multi-source remote sensing data
CN117235607B (en) * 2023-11-13 2024-02-13 天津市农业发展服务中心 Soil moisture content real-time monitoring system and monitoring method

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105550423A (en) * 2015-12-09 2016-05-04 浙江大学 CMORPH satellite precipitation data downscaling method based on Fuzzy-OLS (Ordinary Least Squares)
CN106019408A (en) * 2016-05-10 2016-10-12 浙江大学 Multi-source-remote-sensing-data-based high-resolution-ratio satellite remote-sensing estimation method
CN106446444A (en) * 2016-10-14 2017-02-22 中国科学院遥感与数字地球研究所 Soil moisture spatial predication research based on Bayes maximum entropy and priori knowledge
CN106483147A (en) * 2016-10-14 2017-03-08 中国科学院遥感与数字地球研究所 The long-term sequence passive microwave soil moisture accuracy improvements research worked in coordination with based on MODIS and measured data
CN108169161A (en) * 2017-12-12 2018-06-15 武汉大学 A kind of corn planting regional soil humidity appraisal procedure based on modified MODIS indexes
CN108268735A (en) * 2018-01-29 2018-07-10 浙江大学 Soil moisture NO emissions reduction method based on multi-source remote sensing satellite fused data
CN109002604A (en) * 2018-07-12 2018-12-14 山东省农业科学院科技信息研究所 A kind of soil moisture content prediction technique based on Bayes's maximum entropy
CN109460789A (en) * 2018-11-07 2019-03-12 中国农业科学院农田灌溉研究所 A kind of soil moisture fusion method based on Bayes's maximum entropy

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015195988A1 (en) * 2014-06-18 2015-12-23 Texas Tech University System Portable apparatus for soil chemical characterization

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105550423A (en) * 2015-12-09 2016-05-04 浙江大学 CMORPH satellite precipitation data downscaling method based on Fuzzy-OLS (Ordinary Least Squares)
CN106019408A (en) * 2016-05-10 2016-10-12 浙江大学 Multi-source-remote-sensing-data-based high-resolution-ratio satellite remote-sensing estimation method
CN106446444A (en) * 2016-10-14 2017-02-22 中国科学院遥感与数字地球研究所 Soil moisture spatial predication research based on Bayes maximum entropy and priori knowledge
CN106483147A (en) * 2016-10-14 2017-03-08 中国科学院遥感与数字地球研究所 The long-term sequence passive microwave soil moisture accuracy improvements research worked in coordination with based on MODIS and measured data
CN108169161A (en) * 2017-12-12 2018-06-15 武汉大学 A kind of corn planting regional soil humidity appraisal procedure based on modified MODIS indexes
CN108268735A (en) * 2018-01-29 2018-07-10 浙江大学 Soil moisture NO emissions reduction method based on multi-source remote sensing satellite fused data
CN109002604A (en) * 2018-07-12 2018-12-14 山东省农业科学院科技信息研究所 A kind of soil moisture content prediction technique based on Bayes's maximum entropy
CN109460789A (en) * 2018-11-07 2019-03-12 中国农业科学院农田灌溉研究所 A kind of soil moisture fusion method based on Bayes's maximum entropy

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
被动微波土壤水分产品真实性检验研究进展;王春梅 等;《浙江农业学报》;20190531;第31卷(第05期);第846-854页 *

Also Published As

Publication number Publication date
CN110427995A (en) 2019-11-08

Similar Documents

Publication Publication Date Title
CN110427995B (en) Bayesian soil moisture estimation method based on multi-source remote sensing data
Huang et al. Assimilating a synthetic Kalman filter leaf area index series into the WOFOST model to improve regional winter wheat yield estimation
AU2016244067B2 (en) Forecasting national crop yield during the growing season
Betbeder et al. Assimilation of LAI and dry biomass data from optical and SAR images into an agro-meteorological model to estimate soybean yield
Ramarohetra et al. How satellite rainfall estimate errors may impact rainfed cereal yield simulation in West Africa
Montzka et al. Modelling the water balance of a mesoscale catchment basin using remotely sensed land cover data
EP3719722A1 (en) Forecasting national crop yield during the growing season
Naz et al. Improving soil moisture and runoff simulations at 3 km over Europe using land surface data assimilation
Choudhary et al. Random Forest for rice yield mapping and prediction using Sentinel-2 data with Google Earth Engine
Gomez et al. Analysis of the uncertainties affecting predictions of clay contents from VNIR/SWIR hyperspectral data
Yin et al. Estimation of grassland height based on the random forest algorithm and remote sensing in the Tibetan Plateau
Zhang et al. Winter wheat identification by integrating spectral and temporal information derived from multi-resolution remote sensing data
Chen et al. Improving the practicability of remote sensing data-assimilation-based crop yield estimations over a large area using a spatial assimilation algorithm and ensemble assimilation strategies
Fitzgerald et al. Directed sampling using remote sensing with a response surface sampling design for site-specific agriculture
Dimov et al. Sugarcane yield estimation through remote sensing time series and phenology metrics
Liu et al. Winter wheat yield estimation based on assimilated Sentinel-2 images with the CERES-Wheat model
Hu et al. Assessment of SMAP and SMOS soil moisture products using triple collocation method over Inner Mongolia
Hong et al. 15 Global Precipitation Estimation and Applications
Cheng et al. ChinaCropSM1 km: a fine 1 km daily soil moisture dataset for dryland wheat and maize across China during 1993–2018
Faouzi et al. Sensitivity analysis of CN using SCS-CN approach, rain gauges and TRMM satellite data assessment into HEC-HMS hydrological model in the upper basin of Oum Er Rbia, Morocco
Zhang et al. Estimation of rice yield from a C-band radar remote sensing image by integrating a physical scattering model and an optimization algorithm
Zhang et al. Enhanced Feature Extraction From Assimilated VTCI and LAI With a Particle Filter for Wheat Yield Estimation Using Cross-Wavelet Transform
CN114611699A (en) Soil moisture downscaling method and device, electronic equipment and storage medium
Leng et al. Enhanced surface soil moisture retrieval at high spatial resolution from the integration of satellite observations and soil pedotransfer functions
Sui et al. Disentangling error structures of precipitation datasets using decision trees

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
TA01 Transfer of patent application right

Effective date of registration: 20230717

Address after: 100190 No. 19 West North Fourth Ring Road, Haidian District, Beijing

Applicant after: Aerospace Information Research Institute,Chinese Academy of Sciences

Address before: No. a 20, Datun Road, Chaoyang District, Beijing 100101

Applicant before: Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences

TA01 Transfer of patent application right
GR01 Patent grant
GR01 Patent grant