Quantitative monitoring method of tea gardens damaged by drought and high temperature combined disaster using satellite remote sensing

Abstract

The present invention discloses a satellite remote sensing quantitative monitoring method for tea gardens affected by drought and high temperature combined disasters. The method first obtains Sentinel‑2 remote sensing images and drone data of tea garden CDH-affected areas and preprocesses them. Secondly, the preprocessed remote sensing images are used to construct a feature space, and the tea garden planting area is extracted in combination with XGBoost. According to the preprocessed drone data, the CDH damage index CDH_DSI reference data of the tea garden is calculated. Then, the vegetation index change ΔVI before and after the tea garden is subjected to CDH stress is calculated, and the CDH_DSI reference data and ΔVI are used to construct a tea garden CDH_DSI estimation model. Finally, based on the tea garden estimation model and remote sensing images, satellite remote sensing-based CDH_DSI quantitative mapping of the tea garden is completed. The present invention realizes satellite remote sensing quantitative estimation and satellite remote sensing quantitative mapping of the tea garden damage index.

Description

Technical Field

[0001] The present invention relates to the intersecting fields of optical satellite remote sensing data processing, machine learning, gardening, and natural disaster monitoring, and in particular to a satellite remote sensing quantitative monitoring method for tea gardens affected by drought and high temperature combined disasters. Background Art

[0002] Satellite remote sensing has been widely used for crop area estimation and agricultural disaster monitoring. Compared to major grain crops like rice, corn, and wheat, research on extracting information about tea plantations using satellite remote sensing technology started relatively late. However, there is relatively little research on remote sensing monitoring of crop temperature stress, especially high-temperature stress, which remains limited to field and laboratory experiments, with limited satellite remote sensing research. However, to date, there are no reports on using satellite remote sensing technology to monitor damage to tea plantations under the combined stress of drought and high temperature.

[0003] In general, there has been more research on agricultural disaster remote sensing monitoring, especially for major crops, but less on tea gardens. The satellite remote sensing data used are mostly from the Advanced Very High Resolution Radiometer (AVHRR), Moderate-resolution Imaging Spectroradiometer (MODIS), Tropical Rainfall Measuring Mission (TRMM), Advanced Microwave Scanning Radiometer (AMSR), Soil Moisture Active Passive (SMAP), and Soil Moisture and Ocean Salinity (SMOS) satellites. These satellites have the advantage of high temporal resolution, but low spatial resolution, which cannot meet the needs of agricultural production. With the development of satellite remote sensing technology, medium- and high-resolution satellite data such as the US Landsat, the European Space Agency's Sentinel series, and my country's high-resolution, resource, and environmental disaster reduction satellites are becoming increasingly abundant. It is necessary to further explore their potential in remote sensing monitoring of tea garden disasters.

[0004] Traditionally, field surveys for disaster data collection are time-consuming, labor-intensive, and inefficient. However, the development of drone technology has provided a reliable means for ground-based disaster surveys. Therefore, there is an urgent need to leverage the advantages of drones and medium- to high-resolution satellite remote sensing to develop satellite remote sensing quantitative monitoring technology for combined drought and high-temperature stress in tea gardens to meet the needs of disaster loss assessment and field management. Summary of the Invention

[0005] Drought and high temperature combined stress is a natural disaster that is becoming increasingly serious in the context of global climate change. The purpose of the present invention is to address the problem of lack of regional quantitative monitoring and assessment products for damage to tea gardens under drought and high temperature combined stress. By making full use of the advantages of drones, machine learning, and satellite remote sensing, a satellite remote sensing quantitative monitoring method for tea gardens suffering from drought and high temperature combined disasters is proposed, as well as a reference data acquisition technology for the tea garden damage index under drought and high temperature combined disaster stress, a satellite remote sensing quantitative estimation technology for the tea garden damage index, and a satellite remote sensing quantitative mapping technology for the tea garden damage index. This forms a satellite remote sensing quantitative monitoring technology system for tea gardens suffering from drought and high temperature combined disasters, and realizes satellite remote sensing quantitative monitoring of the severity of damage to tea gardens under drought and high temperature combined disaster stress.

[0006] The quantitative monitoring method of compound drought–heatwave (CDH) disaster in tea gardens using satellite remote sensing includes the following steps:

[0007] Step 1: Acquire Sentinel-2 remote sensing images and drone data of CDH-affected areas in tea gardens and perform preprocessing.

[0008] Step 2: Use the preprocessed remote sensing images to construct a feature space and use the XGBoost algorithm to extract the tea garden planting area.

[0009] Step 3: Calculate the reference data of the CDH damage severity index (CDH_DSI) of the tea garden based on the drone data according to the pre-processed drone data;

[0010] Step 4: Calculate the vegetation index change (ΔVI) of the tea garden before and after CDH stress based on the Sentinel-2 remote sensing image of the tea garden planting area and its feature space.

[0011] Step 5: Use the tea tree CDH_DSI damage index reference data and ΔVI to construct a tea garden CDH_DSI estimation model.

[0012] Step 6: Based on the tea garden CDH_DSI estimation model and the global Sentine-2 remote sensing image, complete the tea garden CDH_DSI quantitative mapping based on satellite remote sensing.

[0013] The following are the preferred technical solutions of the present invention:

[0014] In step 1, the acquired Sentinel-2 remote sensing images and drone data are preprocessed, which specifically includes the following steps:

[0015] The remote sensing image data were pre-processed by stitching, cropping, cloud removal and resampling to 10 meters.

[0016] The drone photo data is spliced using professional software to obtain regional drone images.

[0017] In step 2, a tea garden extraction feature space including original spectral bands and vegetation indices is first constructed. The original spectral bands include: blue light band (MSI2), green light band (MSI3), red light band (MSI4), red edge band 1 (MSI5), red edge band 2 (MSI6), red edge band 3 (MSI7), near infrared band 1 (MSI8), near infrared band 2 (MSI8A), short wave infrared band 1 (MSI11), and short wave infrared band 2 (MSI12).

[0018] Vegetation indices include the Normalized Difference Vegetation Index (NDVI), the Green Normalized Difference Vegetation Index (GNDVI), the Ration Vegetation Index (RVI), and the Enhanced Vegetation Index (EVI). The calculation formulas are as follows:

[0019]

[0020]

[0021]

[0022]

[0023] Among them, ρ MSI2 , ρ MSI3 , ρ MSI4 , ρ MSI8 Refers to the blue, green, red and near-infrared bands respectively.

[0024] Then the tea garden planting area is extracted based on the XGBoost algorithm.

[0025] Step 3: Calculate the reference data of the tea tree CDH damage severity index (CDH_DSI) based on the drone data, including the following steps:

[0026] First, the damaged tea leaves were identified based on the high spatial resolution images from drones.

[0027] Secondly, based on the Sentinel-2 pixel size (10 meters), the damage ratio of tea leaves within a Sentinel-2 pixel is calculated.

[0028] As reference data for the CDH damage index in tea gardens.

[0029] In step 4, the vegetation index change ΔVI before and after the tea garden suffered a drought and high temperature combined disaster was calculated.

[0030]

[0031] Among them, VI before Represents the vegetation index before damage, VI after represents the vegetation index after damage;

[0032] Step 5: Using the tea tree CDH_DSI reference data as the dependent variable and ΔVI (%) as the independent variable, the XGBoost algorithm was used to construct a tea garden CDH_DSI remote sensing estimation model.

[0033] In step 6, the CDH_DSI remote sensing estimation model constructed in step 5 is used to calculate the CDH_DSI of all tea garden pixels in the study area, and the CDH_DSI spatial distribution mapping of tea gardens at the county scale is realized.

[0034] Compared with the prior art, the advantages of the present invention are:

[0035] 1. Introducing machine learning into satellite remote sensing monitoring of tea gardens affected by combined drought and high temperature disasters to improve monitoring accuracy; 2. Introducing drones into disaster surveys of tea gardens affected by combined drought and high temperature disasters as ground truth data for satellite remote sensing modeling and verification, with more samples, higher accuracy, and more reliable data; 3. Traditional satellite remote sensing disaster monitoring is targeted at single disasters such as drought, high temperature, low temperature, and floods. This invention is the first to use satellite remote sensing technology to monitor the degree of damage to tea gardens under combined drought and high temperature stress; 4. The proposed disaster severity index has a strong correlation with the change in vegetation index before and after the disaster, and has a clear physical meaning, thus realizing satellite remote sensing quantitative estimation of the tea garden damage index and satellite remote sensing quantitative mapping; 5. This method is applicable to all satellite remote sensing monitoring of the severity of damage to tea gardens affected by combined drought and high temperature disasters. Its ideas are also applicable to satellite remote sensing monitoring of tea gardens affected by drought, floods, high temperature, and low temperature, and it is also of reference significance for remote sensing monitoring of disasters of other crops. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Sentinel-2 image of Wuyi County, Zhejiang Province, the study area selected for this invention;

[0037] Figure 2 The comparison of the F1 value and overall accuracy of tea garden extraction discovered by the present invention with different classification methods with different sample numbers;

[0038] Figure 3 This is a tea plantation distribution map of Wuyi County, extracted based on the optimal sample number and classification method discovered by this invention. a1, b1, c1, d1, e1, and f1 are high-resolution Google Earth images; a2, b2, c2, d2, e2, and f2 are the tea plantation images extracted using Sentinel-2 and superimposed on a1, b1, c1, d1, e1, and f1 images.

[0039] Figure 4 The scatter plot of the drought and high temperature combined disaster severity index (CDH_DSI) of tea garden estimated by the present invention based on Sentinel-2 data using different methods, namely XGBoost, random forest and multivariate linear regression, and the CDH_DSI measured by drone;

[0040] Figure 5 The spatial distribution of the severity index of the drought and high temperature combined disaster in Wuyi County is estimated based on the XGBoost model and Sentinel-2 data according to the technical system of the present invention. DETAILED DESCRIPTION

[0041] The present invention is further described below with reference to specific figures and implementation examples. The present invention is a satellite remote sensing quantitative monitoring method for tea gardens affected by drought and high temperature combined disasters. The specific calculation method includes the following steps:

[0042] Step 1: First, download the Sentinel-2A / B satellite images of the study area from January 1, 2019 to February 28, 2023; perform cloud removal processing to remove clouds and shadows, interpolate the pixels affected by clouds and shadows, and obtain cloud-free images covering the entire study area, such as Figure 1 As shown in FIG. 1 , the research area selected for the present invention is a satellite Sentibel-2 image of Wuyi County, Zhejiang Province.

[0043] Step 2: Using the processed remote sensing images, construct a tea garden extraction feature space including the original spectrum and vegetation index to extract the tea garden planting area.

[0044] The original spectral bands include Sentinel-2 MSI2, MSI3, MSI4, MSI5, MSI6, MSI7, MSI8, MSI8A, MSI11, and MSI12. The spatial resolution of MSI2 (496.6 nm), MSI3 (560.0 nm), MSI4 (664.5 nm), and MSI8 (835.1 nm) is 10 m × 10 m, while the spatial resolution of MSI5 (703.9 nm), MSI6 (740.2 nm), MSI7 (782.5 nm), MSI8A (864.8 nm), MSI11 (1613.7 nm), and MSI12 (2202.4 nm) is 20 m × 20 m. MSI5, MSI6, MSI7, MSI8A, MSI11, and MSI12 need to be resampled to 10 m × 10 m.

[0045] Vegetation indices include NDVI, GNDVI, RVI, and EVI, and their calculation formulas are as follows;

[0046]

[0047]

[0048]

[0049]

[0050] Then the effects of different numbers of training samples on the tea garden extraction accuracy are analyzed, and the tea garden extraction accuracy of XGBoost, random forest, logistic regression, and naive Bayesian regression are compared. Figure 2 As shown in the figure, overall, machine learning methods such as XGBoost and random forest are better than traditional logistic and naive Bayes regression methods. The classification accuracy of machine learning methods such as XGBoost and random forest increases with the increase of sample number, but XGBoost performs best. Finally, the optimal algorithm and number of labels for satellite remote sensing extraction of tea gardens are determined.

[0051] Figure 3 This is the tea garden distribution map of Wuyi County extracted based on the optimal number of samples and classification method. The overall classification accuracy, F1 value, precision and recall value are 0.9243, 0.9237, 0.9236 and 0.9237, respectively. Among them, a1, b1, c1, d1, e1, and f1 are high-spatial-resolution Google Earth images, and a2, b2, c2, d2, e2, and f2 are the results of tea gardens extracted based on Sentinel-2 and superimposed on the a1, b1, c1, d1, e1, and f1 images. Comparing the Google Earth images with the extracted results, the effect is satisfactory.

[0052] Step 3: Identify tea leaves damaged by CDH stress based on high spatial resolution UAV images and calculate the CDH_DSI reference data of tea gardens at the Sentinel-2 pixel scale (10 meters).

[0053] Step 4: Calculate the vegetation index change ΔVI before and after the tea garden is subjected to drought and high temperature combined disaster stress.

[0054]

[0055] Among them, VI before Represents the vegetation index before damage, VI after represents the vegetation index after damage;

[0056] Step 5: Using CDH_DSI reference data as the dependent variable and ΔVI (%) as the independent variable, a CDH_DSI remote sensing estimation model was constructed by comparing XGBoost, random forest, logistic regression, naive Bayesian regression and other algorithms, and the best model was determined based on its prediction accuracy. Figure 4 The scatter plot of CDH_DSI estimated based on Sentinel-2 data and CDH_DSI measured based on drones shows that XGBoost performs best, R 2 It reaches 0.82 and the RMSE is 7.61.

[0057] Step 6: Using the CDH_DSI remote sensing estimation model and ΔVI (%) constructed in step 4, the CDH_DSI of all tea garden pixels in the study area is calculated to achieve the CDH_DSI spatial distribution mapping of tea gardens at the county scale. The results are as follows: Figure 5 As shown, for the first time, quantitative remote sensing mapping of the spatial distribution of the severity index of the combined disaster of drought and high temperature in tea gardens was achieved.

Claims

1. A satellite remote sensing quantitative monitoring method for tea gardens affected by drought and high temperature combined disasters, characterized by: The following steps are involved: Step 1: Acquire Sentinel-2 remote sensing images and UAV data of tea gardens affected by drought, high temperature and CDH and perform preprocessing; Step 2: Use the pre-processed remote sensing images to construct feature space and extract the tea garden planting area using the XGBoost algorithm; Step 3: Calculate the reference data of the CDH damage index (CDH_DSI) for tea gardens based on the pre-processed drone data. The specific process is as follows: First, the damaged tea leaves are identified based on the pre-processed drone data; Secondly, based on the size of Sentinel-2 pixels, the damage ratio of tea leaves within a Sentinel-2 pixel is calculated as the reference data of the CDH damage index of the tea garden; Step 4: Calculate the vegetation index change ΔVI before and after the tea garden suffered CDH based on the Sentinel-2 remote sensing image of the tea garden planting area and its feature space; Step 5: Using the tea tree damage index CDH_DSI reference data and ΔVI, a tea garden CDH_DSI estimation model is constructed. Specifically, the tea tree CDH_DSI reference data is used as the dependent variable, ΔVI (%) is used as the independent variable, and the XGBoost algorithm is used to construct the tea garden CDH_DSI estimation model. Step 6: Based on the tea garden CDH_DSI estimation model and the global Sentine-2 remote sensing image, complete the tea garden CDH_DSI quantitative mapping based on satellite remote sensing.

2. The satellite remote sensing quantitative monitoring method for tea gardens affected by drought and high temperature combined disasters according to claim 1, characterized in that: The specific process of the pretreatment is as follows: Stitching, cropping, cloud removal, and resampling of Sentinel-2 remote sensing image data to 10 meters; The drone data were stitched together to obtain regional drone images.

3. The satellite remote sensing quantitative monitoring method for tea gardens affected by drought and high temperature combined disasters according to claim 1, characterized in that: The feature space constructed in step 2 includes the tea garden extraction feature space of the original spectral band and vegetation index; The original spectrum bands include: blue light band MSI2, green light band MSI3, red light band MSI4, red edge band 1 MSI5, red edge band 2 MSI6, red edge band 3 MSI7, near infrared band 1 MSI8, near infrared band 2 MSI8A, short wave infrared band 1 MSI11, and short wave infrared band 2 MSI12; The vegetation index includes the normalized difference vegetation index NDVI, the green normalized difference vegetation index GNDVI, the ratio vegetation index RVI, and the enhanced vegetation index EVI. The calculation formula is as follows: Among them, ρ MSI2 , ρ MSI3 , ρ MSI4 , ρ MSI8 Refers to the blue, green, red and near-infrared bands respectively.

4. The satellite remote sensing quantitative monitoring method for tea gardens affected by drought and high temperature combined disasters according to claim 1, characterized in that: The calculation of the vegetation index change ΔVI before and after the tea garden was subjected to CDH stress was as follows: Among them, VI before Represents the vegetation index before damage, VI after Represents the vegetation index after damage.

Images