Surveillance of Pine Wood Nematode Disease Based on Satellite Remote Sensing Images

Surveillance of pine wood nematode disease based on satellite remote sensing images

author: sherry

2023-01-16

I Preface
The forest environment is an important part of the biological environment in the human natural environment, an important component of the earth's biosphere, and the main body of the earth's terrestrial ecosystem. The health of the forest system is of great significance to the sustainable development of human beings. However, with the development of human activities, natural forests are gradually replaced by artificial secondary forests with relatively single species of trees. Compared with natural forests, artificial secondary forests have weaker resistance to pests and diseases. The prevention, monitoring and management of forest pests and diseases has become one of the priorities of the forestry management department.
Pine forests have strong adaptability, can resist drought, strong wind and cold climate, and can grow in various types of soil, so they are widely distributed in my country. However, the tree species is fatally threatened by the pine wood nematode. Pine wilt disease caused by pine wood nematode is called cancer of pine trees. It is a devastating disease with many transmission routes, hidden disease sites, fast onset speed, long incubation time, and difficult control. It is a major invasive alien species in my country and has been included in the domestic and foreign forest plant quarantine objects. Since the disease was discovered in 1982, it has spread rapidly and occurred in 14 provinces (cities, districts) including Jiangsu, Anhui, Shandong, Zhejiang, Guangdong and other parts of the country, covering an area of more than 80,000 hectares, resulting in the death of a large number of pine trees. It has caused serious damage to my country's pine forest resources, natural landscape and ecological environment, and caused serious economic and ecological losses.

Schematic diagram of pine forest and diseased trees
The current forest management and disease monitoring and prevention methods based on manual field survey and plot method can no longer meet the requirements of the severe situation of forest disease intensification. On the basis of fully improving the awareness of forest pest control and control, it is urgent to introduce and develop new monitoring technologies to realize early detection of forest pests and diseases, and accurate monitoring of disease location and distribution range. With the rapid development of space technology, satellite remote sensing data sources can be obtained from multiple scales, and the comprehensive use of satellite remote sensing data of different scales for quantitative monitoring of forest diseases and insect pests has become more and more popular among experts and scientific researchers. Remote sensing technology can explore the incidence and pathogenesis of forest pests and diseases from multiple angles and levels. Compared with traditional monitoring methods, the information is more comprehensive and accurate, and the conclusions drawn are more convincing.

II Theoretical Basis
2.1 Causes of Pine Wood Nematode Disease
The transmission of pine wood nematode disease is mainly based on crustacean insects such as longhorn longhorn (Pine Beetle) and relies on the flight of beetles to carry out the transmission of nematode disease. The pine beetle eats the twig bark of the pine tree, and the pine xylophilus enters the pine tree through the wound. In the process of feeding and growing, pine xylophilus will secrete a large amount of cellulase and toxic substances, destroy the cell wall and resin channel of the pine epidermis, cause the metabolism disorder of the pine tree, hinder the water transport and die. Therefore, the onset of pine wood nematode disease has two characteristics. First, it is distributed discretely; second, there is a high probability of diseased trees around dead trees. These two characteristics guide the field investigation and satellite remote sensing monitoring research work of this study.

Typical characteristics of pine wood nematode disease transmission

After the pine wood nematode invades the trees, the needles of the pine wood lose water, chlorosis, and then turn brown, and finally the whole plant dies, and the needles are reddish-yellow. Its development process can be divided into four stages:
ØNormal appearance, reduced or stopped resin secretion, decreased transpiration;
ØThe needles begin to change color, resin secretion stops, and traces of beetle or other beetle infestation and egg laying can usually be observed;
ØMost of the needles turn yellowish brown, wilting, and beetle burrs are usually visible;
ØThe needle leaves all turn yellowish brown, and the diseased tree withers and dies, but the needle leaves do not fall off. At this time, there are generally many kinds of pests living on the tree.
The time from infection to death of pine trees will vary with the time and quantity of pine wood insect infestation. From the onset of symptoms to death, it takes about one month to several months. In summer, when the outside temperature rises, the attack of diseases and insect pests will gradually increase, the incidence rate will become higher and higher, and the speed of infection and death will be faster and faster.

Change process of pine wood nematode disease infection
2.2 Remote Sensing Technology
Remote sensing is "remote perception". Literally, it is to obtain information from "objects" without touching them at a long distance. It collects the data of the target object "remotely" through the sensor, and obtains a science and technology related to the ground object, or region, or phenomenon through the analysis of the data.
In recent years, with the development of satellite remote sensing technology, various satellite remote sensing satellites such as optical and microwave have been successfully launched and operated, making it possible for satellite remote sensing methods to replace traditional field survey methods for large-scale routine monitoring of different ground objects or emergency observations for emergencies . At present, there are many ways to classify optical satellite remote sensing satellites. From the perspective of spatial resolution, they can be divided into medium and low resolution satellite remote sensing satellites. Common ones include NOAA meteorological observation satellites, GOCI static water color satellites, etc.; medium resolution satellite remote sensing satellites Terra and Aqua; Medium and high-resolution land satellites Landsat, Sentinel-2; and high-resolution satellite remote sensing satellites Quick Bird, SPOT-6, Worldview4 and China's Gaofen-2, etc.

Satellite remote sensing images have the following characteristics:
(1) Macro: The remote sensing detection range is wide, especially satellite remote sensing, and the detection range can not be limited by political regions and geographical conditions. The higher the remote sensing platform is, the larger the detection range will be.
(2) Multi-band: The sensor can detect and record information in different bands from ultraviolet, visible light, near-infrared to microwave, far exceeding the range of visible light used in photogrammetry.
(3) Periodicity: Remote sensing satellites have the characteristics of periodic and repeated acquisition of images, and can make repeated observations of the same area in a short period of time, thereby enabling forecasting and forecasting.
Economy: Based on the above characteristics of satellite remote sensing, it is concluded that in a large range, if you want to obtain 3D data with the same accuracy as that produced by aerial photographs, satellite remote sensing image acquisition is more economical than aerial photograph acquisition, and it does not need to spend too much manpower and material resources.

3 Data Acquisition and Preprocessing
3.1 Satellite remote sensing image data acquisition
The data acquisition methods and websites of different satellite remote sensing images are different. For example, the Landsat series satellite images are the satellite data of the United States Geological Survey and the Space Agency. You can obtain Landsat8 images with a resolution of 30 meters for free on the USGS website, while the Sentinel series satellite images The image is the satellite image data of the European Space Agency. The Sentinel-2 image with a resolution of 10 meters can be obtained free of charge from the European Space Agency. As for the higher resolution remote sensing image data Gaofen-2 or Quick Bird, SPOT-6, Worldview4 Need to pay for purchase.
3.2 Satellite image preprocessing
The preprocessing of remote sensing images mainly includes preprocessing processes such as radiometric calibration, atmospheric correction, orthorectification, image fusion, image mosaic, and image cropping. The preprocessing operations are mainly carried out in ENVI software。

Atmospheric Correction

Since the radiance finally measured by the sensor is not the real reflection of the ground surface, the radiation amount error will be caused by scattering during the transmission process, and the atmosphere will also absorb part of the radiation. The effect of reflection. Atmospheric correction is mainly divided into two types: based on statistical models and based on physical radiation transmission. The FLAASH model is commonly used, and it is also the preferred correction model for hyperspectral radiation energy image reflectivity inversion.

Orthorectification

Orthophoto production generally selects some ground control points on the photo, and uses the digital elevation model (DEM) data that has been acquired in the photo range to simultaneously correct the tilt and projection difference of the image, and reconstruct the image. Sampling into an orthophoto.

Image Fusion

Image fusion refers to the process of image processing and computer technology to maximize the extraction of beneficial information in each channel from the image data collected by multi-source channels about the same target, and finally synthesize high-quality images to improve the accuracy of image information. Utilization, improve the accuracy and reliability of computer interpretation, improve the spatial resolution and spectral resolution of the original image, and facilitate monitoring. The fusion methods of remote sensing images include GS fusion, CN fusion, PC fusion and so on.

Image Mosaic

Image mosaic refers to the process of connecting a series of flat images into a continuous panoramic image (mosaic image). Image mosaic requires that the mosaic image has no mosaic seam, no image quality degradation, color balance, and no ghosting.
4 Research Methods and Results
4.1 Pine original spectral curve
An important condition for carrying out research on satellite remote sensing monitoring is to analyze the spectrum of the endmember of the research object. Based on the end members of pine wood in different health conditions, 100 pine wood samples with different pest levels were selected for statistical analysis, and the spectral curves of different levels of pest damage were extracted as shown in Figure 7. It can be seen that the spectra of pine leaves with different health levels show great differences, among which the spectra of healthy pine leaves show a steep slope (that is, "red edge") near the wavelength of 680-760 nm, indicating that the leaves have higher chlorophyll content and Healthy cell structure. Some leaves of the infected pine leaves are healthy, and a small number of leaves are dehydrated and yellow. The slope of the "red edge" of the corresponding spectral curve decreases, and the reflectance of the yellow and red light bands (about 530-680 nm) increases, which is related to the yellowing caused by the decrease in chlorophyll content of the leaves. . For dead leaves, there is no obvious reflectance feature, and the spectral curve shows a gentle upward trend.
4.2 Various index spectra of pine wood
In order to further widen the difference between different levels of pests, various spectral indices can be constructed by using the original spectrum. The vegetation index inverts information related to vegetation physiology or physics, so as to evaluate the growth status of vegetation. At present, the vegetation indexes that are widely used in the monitoring research of forestry diseases and insect pests include normalized difference vegetation index, ratio vegetation index, greenness index, vegetation decay index and anthocyanin reflection index, etc.
Vegetation index table

Vegetation Index Name	Calculation formula
GI	R544/R677
SIPI	(R800-R445)/(R800-R680)
NPCI	(R680-R430)/(R680+R430)
MSR	(R800/R670- 1)/(R800/R670+l)^1/2
NRI	(R570-R670)/(R570+R670)
PRI	(R570-R531)/(R570+R531)
TCARI	3[(R700-R670)-0.2(R700-R550)*(R700/R670)]
PSRI	(R800-R445)/(R800-R680)
PHRI	(R550-R531)/(R550+R531)
ARI	(R550)^(- 1)-(R700)^(-1)
TVI	0.5[120(R750-R550)-200*(R670-R550)]
RVSI	[(R712+R752)/2]-R732
MCARI	[(R701-R671)-0.2*(R701-R549)]/(R701/R671)
AR VI	R800-(2R700-R436)]/[R800+2R700-R436)
DVI	R800-R700
EVI	2(R800-R700)/(R800+6R700-7.5*R436+1)
GNDVI	(R546-R700)/(R546+R700)
LMI	R1650/R830
OSAVI	[(R800-R700)/(R800+R700+0.16)]*(1+0.16)
NDVI	(R800-R700)/(R800+R700)
RVI	R800/R700
SAVI	1.5*(R800-R700)/(R800+R700+0.5)
SLAVI	R800/(R700+R800)
VARI	(R546-R700)/(R546+R700-R436)
YI	(R580-2*R630+R680)/2500
WBI	R950/R900

Spectral curves of pine leaves in different health conditions
4.2 Pine Extraction Results
Based on the analysis of the original spectrum of pine leaf endmembers and various spectral indices, different grades of healthy pine wood and pine leaves are extracted. The research methods and process mainly include the following two aspects:
Classification of ground features: The purpose of classification of ground features is mainly to better observe the boundaries of vegetation, soil, etc. The classification method adopts supervised classification, and uses random forest algorithm for supervised learning. First, manually mark the types of objects to be classified and the object endmembers (ROI). In this paper, the objects are divided into three categories: hardened surface (blue), grassland bare soil (light green) and tree body (green), and then input into RF classification Automatic classification in the machine to obtain tree distribution map.
(2) Endmember detection: Endmember detection is used to extract different grades of pine wood in the image, using the original spectrum and various index spectral data of different grades of pests as the input data of endmember detection, and then select the spectral angle detection method to identify the image Different grades of pine xylophilus in , get grade results of different grades of pine xylophilus.
Post-test processing. There are scattered pixels in the original detection results, which need to be optimized. Using post-classification processing tools, including maximum analysis and cluster processing, the goal is to remove small patch noise. The optimization results are shown in Figure 8.