A method for detecting farmland plot irrigation by integrating multi-source remote sensing data

Abstract

The present invention discloses a method for detecting farmland plot irrigation by integrating multi-source remote sensing data, which includes obtaining time-series remote sensing data and auxiliary data and performing preprocessing; based on the above data, obtaining the normalized difference moisture index, land surface temperature and precipitation at the plot scale; constructing a scatter plot of NDMI-LST feature space with the normalized difference moisture index at the plot scale as the abscissa and the land surface temperature as the ordinate, obtaining the upper and lower boundary lines of the NDMI-LST feature space, and calculating the temperature vegetation drought index; constructing a time-series curve of the temperature vegetation drought index at the plot scale, setting a rule set for plot irrigation detection, establishing a plot irrigation detection algorithm with an adaptive dynamic threshold, obtaining the plot irrigation cycle, and obtaining the irrigation information of the analysis and statistical area. Therefore, by using the above method, it is possible to quickly and accurately extract farmland plot irrigation events and cycle information, providing a reference for decision-making such as water resource management and agricultural production activities.

Description

Technical Field

[0001] The present invention relates to the field of agricultural remote sensing monitoring, and in particular to a method for detecting farmland plot irrigation by fusing multi-source remote sensing data. Background Art

[0002] With the rapid development of spatial information technology, remote sensing and geospatial mapping technologies, which have the advantages of wide coverage, high information timeliness, and low operation cost, have become the main means for obtaining regional agricultural irrigation information.

[0003] Since irrigation can cause changes in soil moisture or crop moisture, and can also cause changes in crop growth; therefore, by using multi-temporal remote sensing observations to capture the spectral feature differences of soil with different water contents and crops in different growth states before and after irrigation, the process of soil moisture change or crop growth change can be quantified, and the extraction of farmland irrigation information can be realized. Currently, the change detection model based on vegetation index has become the main method for extracting farmland irrigation information. For example, Tian Xin et al. constructed an irrigation area extraction model based on the difference between VSWI and TVDI, and realized the extraction of the irrigation area in the Shenwu Irrigation Area of Inner Mongolia. Du Enyu et al. calculated the TVDI and MPDI indices using Landsat images, and constructed an irrigation area extraction model based on the difference between the two indices before and after irrigation. Yang Yongmin et al. quantified the change in the Sentinel-1 microwave backscattering coefficient caused by irrigation events, and proposed an irrigation area extraction method based on time series difference and local threshold method. However, most of these studies focus on single irrigation information with known irrigation time (such as irrigation distribution and area, water consumption, etc.), and lack in-depth exploration of multiple irrigation information (such as irrigation times, irrigation time, irrigation cycle, etc.) during the crop growth season.

[0004] Compared with the above simple extraction of agricultural irrigation scope and area, accurately identifying agricultural irrigation events and cycle information using remote sensing technology has higher application value and is more challenging; and with the development of remote sensing technology, fusing multi-source remote sensing satellites can obtain more-dimensional and longer-time series observations of the crop-soil system, providing the possibility for extracting multi-dimensional irrigation information of farmland during the crop growth season. Zhai Yongguang et al. jointly used Sentinel-1\2\3 satellite data to construct a time series curve of the Vegetation Temperature Condition Index (VTCI), and combined with peak-valley detection and matching algorithms to successfully extract the irrigation time, frequency, and cycle of the irrigation area. Chen et al. fused MODIS time series, Landsat images, and auxiliary data, and constructed an irrigation event detection algorithm based on threshold segmentation on the basis of analyzing the response of crop greenness index to irrigation, and extracted irrigation frequency and time information. These studies have greatly improved the fineness of farmland irrigation remote sensing monitoring.

[0005] Although existing research has achieved good results in the remote sensing monitoring of farmland irrigation, it still faces new challenges: (1) There have been many studies on the spatial distribution and probability mapping of irrigation using multi-source remote sensing data, but the discussion on irrigation detection, that is, irrigation events and cycles, is relatively lacking. (2) In the existing irrigation event detection algorithms based on threshold segmentation, the same threshold is set for the entire study area, lacking a dynamic threshold algorithm that adapts to changes in plot space and crop growth periods. (3) Existing studies either take the entire study area as an irrigation unit, making it difficult to express the differences in irrigation between different farmland plots; or take a single pixel as an irrigation unit, which is vulnerable to salt-and-pepper noise interference, limits the accuracy of remote sensing irrigation detection, and also causes the problem that remote sensing irrigation monitoring is difficult to accurately correspond to actual agricultural farm management units.

[0006] In summary, there are many problems in the current irrigation detection methods in aspects such as determining remote sensing irrigation detection units, multi-source data fusion strategies, calculating model construction, and algorithm application, lacking a dynamic threshold method that adapts to changes in plot space and crop growth periods. Therefore, there is an urgent need for a method for identifying farmland irrigation events and extracting cycles at the plot scale that integrates multi-source remote sensing data to support agricultural remote sensing monitoring and agricultural water resource management. Summary of the Invention

[0007] The purpose of the present invention is to provide a method for detecting farmland plot irrigation that integrates multi-source remote sensing data, which can achieve...

[0008] To achieve the above purpose, the present invention provides a method for detecting farmland plot irrigation that integrates multi-source remote sensing data, including the following steps:

[0009] S1. Obtain time-series remote sensing data and auxiliary data, and perform preprocessing to obtain a raster dataset with the same geospatial reference and spatio-temporal resolution and pixel spatio-temporal alignment;

[0010] S2. Based on the obtained time-series remote sensing data and auxiliary data, obtain the normalized difference water index, land surface temperature, and precipitation at the plot scale, and construct an irrigation remote sensing monitoring dataset at the plot scale;

[0011] S3. Construct a scatter plot of NDMI-LST feature space with the normalized difference water index at the plot scale as the abscissa and the land surface temperature as the ordinate, extract and linearly fit the upper and lower boundary lines of the NDMI-LST feature space, and calculate the temperature vegetation drought index of the plot;

[0012] S4. Construct a time-series curve of TVDI at the plot scale, set a plot irrigation detection rule set, establish an adaptive dynamic threshold plot irrigation detection algorithm, and obtain the plot irrigation cycle;

[0013] S5. Analyze and statistically calculate the irrigation information of the area based on the plot irrigation cycle.

[0014] Preferably, in step S3, the upper and lower boundary lines of the NDMI-LST feature space specifically include:

[0015] Divide the abscissa of the Normalized Difference Moisture Index (NDMI) into intervals, and count the mean, standard deviation, and confidence interval of the Land Surface Temperature (LST) within the divided intervals. Find the maximum and minimum values of the LST within the confidence interval, and record the corresponding scatter points as the upper and lower boundary points of the moisture of the crops in the divided intervals. Fit the upper and lower boundary points of the moisture to obtain the upper and lower boundary lines of the NDMI-LST space.

[0016] Preferably, the upper and lower boundary lines of the NDMI-LST space are specifically expressed as:

[0017] f u = a u × NDMI + b u

[0018] f b = a b × NDMI + b b

[0019] In the formula, f u is the upper and lower boundary lines of the NDMI-LST space, a u , b u are the slope and intercept of the upper boundary line obtained by fitting, f b is the upper and lower boundary lines of the NDMI-LST space, a b , b b are the slope and intercept of the lower boundary line obtained by fitting.

[0020] Preferably, in step S3, calculate the Temperature Vegetation Drought Index (TVDI) as follows:

[0021] Ts = f u (NDMI) - LST

[0022] MI = f u (NDMI) - f b (NDMI)

[0023] TVDI = Ts / MI

[0024] In the formula, Ts is the distance between the plot coordinates (NDMI, LST) and the upper boundary line in the vertical axis direction, and MI is the distance between the upper and lower boundaries in the vertical axis direction.

[0025] Preferably, in step S4, the set of plot irrigation detection rules includes:

[0026] The maximum number of irrigations during the entire crop growing season;

[0027] The minimum threshold for the increase in TVDI before and after irrigation;

[0028] The minimum time interval between two adjacent irrigations;

[0029] The minimum precipitation that causes the increase in TVDI to exceed the minimum threshold.

[0030] Therefore, by adopting the above-mentioned method for detecting farmland plot irrigation that integrates multi-source remote sensing data, the following technical effects are achieved:

[0031] (1) By using NDMI, a stronger negative correlation between the NDMI and LST of the crop-soil system is obtained, accurately expressing the degree of water stress in the crop-soil system;

[0032] (2) Taking the farmland plot as the irrigation analysis unit, the interference of pixel salt-and-pepper noise is reduced, and the accuracy of irrigation detection is improved;

[0033] (3) Based on the TVDI time series curve of the plot, an irrigation detection algorithm with an adaptive dynamic threshold is established to more accurately extract the irrigation events and irrigation cycle information of the plot.

[0034] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Description of the Drawings

[0035] Figure 1 is a flowchart of an embodiment of a method for detecting farmland plot irrigation that integrates multi-source remote sensing data;

[0036] Figure 2 is a flowchart of the preprocessing of multi-source remote sensing data in an embodiment of a method for detecting farmland plot irrigation that integrates multi-source remote sensing data;

[0037] Figure 3 is a flowchart of TVDI calculation in an embodiment of a method for detecting farmland plot irrigation that integrates multi-source remote sensing data;

[0038] Figure 4 is a flowchart of irrigation cycle detection based on the TVDI time series curve in an embodiment of a method for detecting farmland plot irrigation that integrates multi-source remote sensing data;

[0039] Figure 5 is a map of farmland plot irrigation information in an embodiment of a method for detecting farmland plot irrigation that integrates multi-source remote sensing data in Ningxia County, Figure 5 in which (a) is a spatial distribution map of the number of irrigations, Figure 5 in which (b) is a monthly statistical chart of the irrigation area. Detailed Embodiment

[0040] The present invention can be more specifically explained by the following embodiments. The purpose of disclosing the present invention is to protect all changes and improvements within the scope of the present invention. The present invention is not limited to the following embodiments.

[0041] As Figure 1 shown, a method for detecting farmland plot irrigation by integrating multi-source remote sensing data provided by the present invention includes the following steps:

[0042] S1. Obtain time series remote sensing data (including Sentinel-2 images, MOD11A1 land surface temperature data, precipitation data, etc.) and auxiliary data (including farmland plot vector data, crop planting and irrigation statistics data, etc.), and perform preprocessing. Based on the Sentinel-2 image, generate a raster dataset with the same geospatial reference, spatio-temporal resolution, and pixel spatio-temporal alignment, as Figure 2 shown.

[0043] S2. Calculate the Normalized Difference Moisture Index (NDMI) based on the Sentinel-2 image; extract the Land Surface Temperature (LST) based on the MOD11A1 land surface temperature data; extract the surface precipitation based on the precipitation data; calculate the pixel band feature values, land surface temperature LST, and precipitation at the plot scale based on the farmland plot vector data, and construct an irrigation remote sensing monitoring dataset at the plot scale.

[0044] In particular, the time series of optical images involved in this embodiment is not limited to Sentinel-2 images, and is also applicable to other multi-temporal optical images (such as NASA Landsat images).

[0045] S3. Construct a scatter plot of the NDMI-LST feature space with the Normalized Difference Moisture Index (NDMI) at the plot scale as the abscissa and the Land Surface Temperature (LST) as the ordinate, extract and linearly fit the upper and lower boundary lines (i.e., the dry edge and the wet edge) of the NDMI-LST feature space, and calculate the Temperature Vegetation Dryness Index (TVDI) of the plot, as Figure 3 shown.

[0046] Among them, the upper and lower boundary lines of the NDMI-LST feature space are specifically as follows:

[0047] First, divide the abscissa of the Normalized Difference Moisture Index (NDMI) into intervals at intervals of 0.01;

[0048] Secondly, extract the scatter points in each divided interval Q, and calculate the mean, standard deviation, and confidence interval of the Land Surface Temperature (LST) thereof (in this embodiment, a 95% confidence interval is taken, and the purpose is to remove outliers caused by factors such as noise);

[0049] Then, find the maximum and minimum values of LST within the confidence interval, and record the corresponding scatter points as the upper and lower boundary points of crop moisture in the divided interval Q respectively;

[0050] Finally, extract the upper boundary points of all divided intervals to form the upper boundary line of the scatter plot, and use the least squares method for linear fitting to generate the upper boundary line in the NDMI-LST space:

[0051] f u = a u × NDMI + b u

[0052] In the formula, f u is the upper and lower boundary line of the NDMI-LST space, and a u , b u are the slope and intercept of the upper boundary line obtained by fitting.

[0053] Similarly, obtain the lower boundary line of the NDMI-LST space:

[0054] f b = a b × NDMI + b b

[0055] In the formula, f b is the upper and lower boundary line of the NDMI-LST space, and a b , b b are the slope and intercept of the lower boundary line obtained by fitting.

[0056] In the NDMI-LST scatter space, with the upper and lower boundary lines as constraints, calculate the distance Ts between the coordinates (NDMI, LST) of each farmland plot and the upper boundary line in the vertical axis direction, and the distance MI between the upper and lower boundaries in the vertical axis direction, so as to obtain the Temperature Vegetation Dryness Index (TVDI) of the farmland plot:

[0057] Ts = f u (NDMI) - LST

[0058] MI = f u (NDMI) - f b (NDMI)

[0059] TVDI = Ts / MI

[0060] S4. As Figure 4 shown, construct the time series curve of TVDI at the plot scale, set the plot irrigation detection rule set based on the change of TVDI, and establish the plot irrigation detection algorithm with an adaptive dynamic threshold (Adaptive Threshold for Irrigation Detection Algorithm, ATID).

[0061] Among them, the rules for setting irrigation detection include:

[0062] (1) The maximum number of irrigations M during the entire crop growth season;

[0063] (2) The minimum threshold T for the increase in TVDI before and after irrigation;

[0064] (3) The minimum time interval D days between two adjacent irrigations;

[0065] (4) The minimum precipitation m that causes the TVDI to increase beyond T.

[0066] The parameters such as M, T, D, and m involved in the irrigation detection rules need to be set in combination with the regional environment and conditions.

[0067] In this embodiment, Zhongning County, Ningxia is selected to discuss the setting of irrigation detection rule parameters. The summer crops in the Yellow River Irrigation Area of Zhongning County are mainly corn, and the growth period is about 150 days (mid-late April to late September). It is recommended to irrigate corn 11 times during the growth season. The threshold T for the increase in TVDI caused by irrigation is taken as the difference between the average values of TVDI at all peak and trough points. Combining the recommended irrigation periods, the minimum time interval between two adjacent irrigations is set to 10 days. Combining the total water requirement of corn in Zhongning County, which is 1476.6 mm, and the 150-day growth period, the average daily water requirement for corn irrigation is 9.84 mm; this data provides a benchmark for judging whether the daily precipitation is effective precipitation. When the daily precipitation exceeds this value, it can be considered that precipitation (not irrigation) causes the TVDI to rise, that is, false irrigation detection caused by precipitation is excluded. It should be noted that: in application, the above rules can be added or modified according to the actual situation of the region to improve the applicability and reliability of irrigation detection.

[0068] The plot irrigation detection takes the TVDI event sequence curve as the input; first, search the set V of curve peak points along the time axis of the TVDI curve f and the set V of trough points g ; then, starting from any peak point, search forward along the TVDI time axis for the trough points that meet the rules, and form the trough-peak combination (V g , V f) Finally, select the top M combinations with the largest TVDI interpolation from the trough-peak combinations as the irrigation periods of the farmland plots, and use the trough points of the combinations as the irrigation times.

[0069] S5. Based on the extraction results of the irrigation periods of the farmland plots, statistically analyze the regional irrigation information from multiple dimensions. From the aspect of irrigation area, statistically analyze the monthly (or by crop phenological period) irrigation area of the region, and produce a thematic map of the monthly irrigation spatial distribution; from the aspect of irrigation frequency, statistically analyze the monthly (or by crop phenological period) irrigation frequency of the region or a certain plot, and produce a spatial distribution map of the monthly irrigation frequency, as Figure 5 shown.

[0070] From the monthly statistics of the irrigation area, as Figure 5 shown in (b) of [reference], the irrigation of the farmland plots reaches the peak from March to May, starts to decline in June, rises again in July, reaches the lowest point in August, rises again in September, and then gradually decreases from October to the end of the year. The irrigation area is relatively high in spring and summer, while relatively low in autumn and winter, which is closely related to the crop growth cycle and climate conditions. The high value of the irrigation area generally appears in the vigorous growth period of spring and summer crops, while the low value of the irrigation area appears in the slow growth or dormant stage of autumn and winter crops; the irrigation area is the smallest in August, mainly because the precipitation in that month is sufficient to meet the crop growth. The low irrigation area at the end of the year and the beginning of the year indicates that the water demand is small in the non-crop growth season.

[0071] From the spatial distribution of the irrigation frequency, as Figure 5 shown in (a) of [reference], April to May is the peak irrigation period in Zhongning County. Basically, each plot has 2 to 3 irrigations. Among them, the plots with two irrigations are concentrated on both sides of the Yellow River in the middle of the study area, and the plots with three irrigations are concentrated in the area south of the Yellow River and on both sides of the Yellow River in the northeast; most plots have only 1 irrigation from June to August, and the total number of irrigated plots is relatively small compared with other months; the number of irrigated plots increases from September to October, still mainly with 1 irrigation, and some plots have 2 irrigations. These results are consistent with the judgment and understanding of the spatio-temporal distribution of agricultural irrigation in Zhongning County.

[0072] Therefore, by adopting the above-mentioned method for detecting farmland plot irrigation that integrates multi-source remote sensing data, the present invention can relatively quickly and accurately extract the irrigation events and cycle information of farmland plots, and generate an agricultural irrigation information map of the region, providing a reference for supporting decision-making such as water resource management and agricultural production activities.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions of the present invention or make equivalent replacements, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for detecting farmland plot irrigation by integrating multi-source remote sensing data, characterized in that, It includes the following steps: S1. Obtain time series remote sensing data and auxiliary data, and perform preprocessing to obtain a raster dataset with the same geospatial reference and spatio-temporal resolution, and with pixel spatio-temporal alignment; S2. Based on the obtained time series remote sensing data and auxiliary data, obtain the normalized difference moisture index, land surface temperature and precipitation at the plot scale, and construct an irrigation remote sensing monitoring dataset at the plot scale; S3. Construct a scatter plot of the NDMI-LST feature space with the normalized difference moisture index at the plot scale as the abscissa and the land surface temperature as the ordinate, extract and linearly fit the upper and lower boundary lines of the NDMI-LST feature space, and calculate the temperature-vegetation drought index of the plot; S4. Construct a time series curve of the TVDI at the plot scale, set a plot irrigation detection rule set, establish an adaptive dynamic threshold plot irrigation detection algorithm, and obtain the plot irrigation cycle; Among them, the plot irrigation detection rule set includes: The maximum number of irrigations during the entire crop growing season; The minimum threshold for the increase in TVDI before and after irrigation; The minimum time interval between two adjacent irrigations; The minimum precipitation that causes the TVDI to increase beyond the minimum threshold; S5. Based on the plot irrigation cycle, analyze and statistically calculate the irrigation information of the region.

2. The method for detecting farmland plot irrigation by integrating multi-source remote sensing data according to claim 1, wherein, In step S3, the upper and lower boundary lines of the NDMI-LST feature space specifically include: Divide the interval of the normalized difference moisture index NDMI of the abscissa, and statistically calculate the mean, standard deviation and confidence interval of the land surface temperature LST within the divided interval. Find the maximum and minimum values of LST within the confidence interval, and record the corresponding scatter points as the moisture upper boundary point and lower boundary point of the divided interval crops respectively; Fit the moisture upper boundary point and lower boundary point to obtain the upper and lower boundary lines of the NDMI-LST space.

3. A method for detecting farmland plot irrigation by integrating multi-source remote sensing data according to claim 2, characterized in that, The upper and lower boundary lines of the NDMI-LST space are specifically expressed as: f u = a u × NDMI + b u f b = a b × NDMI + b b where f u is the upper and lower boundary lines of the NDMI-LST space, a u , b u are the slope and intercept of the upper boundary line obtained by fitting, f b is the upper and lower boundary lines of the NDMI-LST space, a b , b b are the slope and intercept of the lower boundary line obtained by fitting.

4. A method for detecting farmland plot irrigation by integrating multi-source remote sensing data according to claim 3, characterized in that, In step S3, calculate the temperature-vegetation drought index TVDI as follows: Ts = f u (NDMI)-LST MI = f u (NDMI) - f b (NDMI TVDI = Ts / MI In the formula, Ts is the distance between the plot coordinates (NDMI, LST) and the upper boundary line in the vertical axis direction, and MI is the distance between the upper and lower boundaries in the vertical axis direction.

Images