CN112415537B

CN112415537B - Model for depicting forest photosynthetic effective radiation distribution by using three-dimensional point cloud data

Info

Publication number: CN112415537B
Application number: CN202010396373.0A
Authority: CN
Inventors: 马利霞; 于东升; 陈洋
Original assignee: Institute of Soil Science of CAS
Current assignee: Institute of Soil Science of CAS
Priority date: 2020-05-08
Filing date: 2020-05-08
Publication date: 2022-07-12
Anticipated expiration: 2040-05-08
Also published as: CN112415537A

Abstract

The invention provides a model for depicting forest Photosynthetically Active Radiation (PAR) distribution by utilizing three-dimensional point cloud data, belonging to the research field of vegetation remote sensing modeling. Aiming at the problem that the PAR three-dimensional distribution of the forest canopy cannot be effectively estimated by utilizing laser radar data on a fine scale at present, the invention provides a forest PAR three-dimensional space-time distribution estimation model based on forest ground laser radar data acquired by multiple stations, wherein the model separately considers the transmission process of direct solar radiation and scattering PAR in forest and simultaneously considers the influence of element angles. The model constructed by the invention can quantitatively depict the three-dimensional space-time distribution of PAR in forest on a fine scale, has higher precision and is beneficial to the research of the forest physiological process.

Description

Model for depicting forest photosynthetic effective radiation distribution by using three-dimensional point cloud data

One, the technical field

The invention relates to a model for depicting three-dimensional space-time distribution of forest photosynthetically active radiation on a fine scale through three-dimensional point cloud data, in particular to a method for accurately estimating three-dimensional distribution of forest photosynthetically active radiation at any time on the fine scale by acquiring three-dimensional point cloud data of a forest canopy through a ground laser radar technology, belonging to the research field of vegetation remote sensing modeling.

Second, background Art

The Photosynthetically Active Radiation (PAR) refers to the part of solar radiation with the wavelength of 400-700nm which can drive vegetation to carry out photosynthesis, and is an important environmental factor influencing the photosynthesis and transpiration process of the vegetation. The difference of the PAR received by different positions of the forest canopy is obvious, the difference changes along with the change of the position of the sun, and meanwhile, the photosynthetic speed of the leaves has a nonlinear characteristic on the PAR response, so that the three-dimensional space-time distribution of the photosynthetically active radiation of the forest canopy has important significance for researching the material and energy exchange process between the forest and the atmosphere.

The PAR of the forest canopy at any time and specific position can be measured by using a quantum sensor or an optical instrument, however, when the random distribution degree of canopy elements is high, the cost for obtaining the three-dimensional distribution of the PAR is high, and meanwhile, the distribution of the canopy elements can be influenced in the measurement process, so that the method is usually used for measuring the distribution of the PAR below the canopy and is less used for obtaining the three-dimensional space-time distribution of the PAR of the forest canopy. The three-dimensional space-time distribution of the PAR of the forest canopy is closely related to the canopy structure, and in order to overcome the limitation of ground measurement, some researches describe the three-dimensional space-time distribution of light on the forest canopy by constructing a forest canopy three-dimensional model and further utilizing methods such as ray tracing and the like. The three-dimensional model of the forest canopy can be divided into four types according to the complexity. The simplest type is that a forest canopy is assumed to be a mixed medium layer with elements distributed randomly, and the penetration coefficient of light is calculated by utilizing a Leaf Area Index (LAI) through a Bell's law theory, so that the energy value of the corresponding position of the canopy is obtained. For example, Myneni et al, in "A procedure for interpreting the radiation region of defined and truncated foil sites, part I. The mathematical expression of bell's law is relatively simple and widely used, but only for forest canopies with a uniform distribution of elements. The second method is to assume the forest canopy as a geometric body, e.g. a sphere, ellipsoid. For example, Chen et al in "A four-scale bidirectional reflectance model based on selectivity architecture" describes the reflectivity information of the canopy by assuming the canopy is a cone. The third method is to consider information such as leaf area density and leaf angle distribution in the voxel and the surface element, assuming that the forest canopy is composed of the voxel or the surface element. For example, in Huanghua and other countries, the RAPID principle and application of a three-dimensional remote sensing mechanism model are that a forest canopy is assumed to be composed of surface elements, and the radiation distribution in the canopy is described by using a radiometric method. The fourth method is to simulate the three-dimensional scene of the canopy with a fine structure by a computer, and the model can obtain more accurate PAR three-dimensional distribution. For example, Zhongjie in "simulation study of the three-dimensional distribution of direct photosynthetic active radiation based on virtual plant canopy", i.e. the three-dimensional distribution of the direct PAR of a single wood is calculated by simulating its real three-dimensional structure. However, as the forest canopy is complex and changeable, high cost and time are required for constructing a real and detailed forest three-dimensional scene. The three-dimensional forest structure model is still different from the real forest canopy structure in the nature.

The laser radar technology (LiDAR) with the three-dimensional drawing capability can acquire the real three-dimensional structure of the natural canopy and provides a foundation for describing the three-dimensional space-time distribution of the PAR of the forest canopy. The most common method is by voxelizing the canopy point cloud, each voxel calculating its PAR value using bell's law. To simplify the calculation, the PAR value is usually calculated using the path length instead of the LAI. The penetration factor of light at different locations in the canopy is calculated using the path length of light in the canopy by Peng in a "Modeling spatial patterns of understory light intensity using air laser scanner (LiDAR)". Bell's Law assumes that the canopy elements are randomly distributed, so voxel size is generally more difficult to determine. The second method is to estimate the penetration coefficient of the light at different locations of the canopy using the number of points or energy information of the points of the LiDAR data. For example, Bode in "Subcanopy Solar Radiation model: the method comprises the steps of Predicting the specific radiation of a crown layer, estimating the penetration coefficient of light at different positions of the crown layer by using the difference between the total number of emission points and the number of interception points of the crown layer, and estimating the penetration coefficient of the light by Chen at the position of "Sensitivity of direct radiation gap fraction from air turbine for laser to laser and reflection characteristics" by using the energy information emitted by LiDAR and the energy information returned.

However, existing research is primarily directed to estimating the three-dimensional spatiotemporal distribution of the underlying PAR of the canopy based on Airborne Lidar (ALS) data using bell's law theory. First, bell's law only applies to the case where elements are randomly distributed, with limitations. Secondly, the ALS data points are of low density and do not accurately describe the three-dimensional spatiotemporal distribution of the forest canopy PAR. Meanwhile, because the photosynthetic efficiency of the vegetation is higher under the condition of scattering PAR than under the condition of direct incidence PAR, the transmission process of the direct solar radiation and the scattering PAR in the canopy needs to be considered separately, but many researches do not consider the factor. In addition, the PAR of the element surface of the canopy is influenced by the included angle between the normal vector of the element and the incident direction of the sun, the smaller the included angle is, the larger the element light-intercepting area is, and the larger the surface PAR is, but the research for describing the three-dimensional distribution of the PAR of the canopy by using LiDAR data at present cannot consider the influence of the real element angle distribution of the canopy.

Third, the invention

The invention aims to provide a model for estimating the three-dimensional space-time distribution of the PAR of a forest canopy on a fine scale by using ground laser radar (TLS) data. The principle of the invention is as follows:

the solar incident PAR is composed primarily of direct and diffuse components. The model will consider the transmission of the direct and scattered PAR inside the canopy separately. The penetration coefficients of direct incidence and scattering PAR at different positions of the canopy are the key for estimating the three-dimensional space-time distribution of the forest PAR, the direct incidence and scattering PAR of each point can be obtained by calculating the penetration coefficient of the direct incidence and scattering PAR of each point in TLS data and further utilizing the measured incident direct incidence and scattering PAR values of the sun, and finally the three-dimensional space-time distribution result of the forest canopy PAR is obtained.

The technical scheme of the invention mainly comprises the following steps:

(1) firstly, a TLS scanner is used for acquiring complete point cloud data of a forest canopy in a multi-station mode. Namely, TLS and targets are erected at different positions of the canopy, and the station number of the scanner is adjusted according to the density of the sample. And finally, splicing the TLS data of different stations through the target to obtain complete three-dimensional point cloud data of the canopy. The values of direct solar radiation and the scattered PAR were measured with a BF5 radiometer.

(2) The direct PAR for each point is estimated. Assuming that each point in the canopy corresponds to a micro-plane, when the micro-plane is perpendicular to the incident direction of the sun, the PAR value of the surface can be expressed as the product of the transmission coefficient of the incident direct solar PAR and the direct solar PAR of the point. Therefore, the determination of the direct PAR penetration coefficient of each point is very critical. Assuming direct solar incident PAR as a collimated light source, the point cloud is rotated according to the direction of the sun so that the solar incident rays are parallel to the-z axis. The rotated point cloud data is subjected to three-dimensional meshing, and the length and the width (V) of the mesh₀) Greater than the canopy average dot spacing (fig. 1 a). Searching for a point q to be solved_j(x₀，y₀，z₀) The point set of the grid (FIG. 1b) is higher than z₀Is projected onto the XY plane and rasterized (fig. 1 c). The grid size is the mean point spacing (V)₁). The grid without points is identified as pores, the proportion of which is the penetration coefficient of the direct PAR at that point. And then obtaining the PAR value of the micro plane perpendicular to the solar incidence direction by using the penetration coefficient of the direct incident PAR of the solar and the direct incident PAR of the point. When an included angle exists between the micro plane of each point and the incident direction of the sun, the normal vector of each point is calculated by using a method of a near point structural surface, namely, 6 points around the point to be solved are searched to calculate the covariance matrix of the point to be solved, and then the eigenvalue and the eigenvector of the point are calculated, wherein the vector corresponding to the minimum eigenvalue is the normal vector of the point. Calculating a cosine value of an included angle between a normal vector of a point to be solved and the incident direction of the sun, and multiplying the cosine value by the calculated direct incidence PAR to obtain the final direct incidence PAR of each pointAnd (6) obtaining the result.

(3) The scatter PAR for each spot was estimated. The scattering PAR for each point in the canopy can be expressed as the product of the solar incident scattering PAR and the transmission coefficient of the scattering PAR for that point. Therefore, the determination of the scattering PAR penetration coefficient of each point is very critical. The solar incident scattering PAR is assumed to be isotropic. And translating the TLS data to enable the point to be solved to be a central point, converting a coordinate system into a polar coordinate system, and setting an inclination angle interval and an azimuth angle interval to be V to divide the upper hemisphere of the canopy into [90/V ] × [360/V ] conical bodies. When there is no point in the cone, it is identified as a void. The ratio of the pores is the transmission coefficient of the scattering PAR of the point to be determined. The product of the penetration coefficients of the incident solar scattering PAR and the scattering PAR is the scattering PAR of the point to be determined.

(4) And estimating the three-dimensional distribution of the forest PAR at a specific moment. The sum of the direct PAR and the scattered PAR of each point in the TLS data is the total PAR of each point, and thus the three-dimensional spatial distribution of the corresponding temporal canopy total PAR, direct PAR, and scattered PAR can be obtained.

Description of the drawings

FIG. 1 direct PAR penetration coefficient calculation

(a) For gridding TLS data after rotating according to solar direction

(b) To select the grid where the point to be solved is located

(c) Projecting the points on an XY plane and rasterizing

FIG. 2 forest canopy TLS data and results of three-dimensional distribution of PAR every two hours

(a) TLS data for forest canopy

(b) - (f) canopy in 2017, 7, 27, 8: 00-16: 00 results of PAR three-dimensional distribution every two hours, the brighter the color, the higher the PAR value, the minimum value of 0. mu. mol. m^-2·s^-1Maximum value of 2036.6 μmol · m^-2·s^-1

Fifth, detailed description of the invention

The invention is further described below by way of specific examples:

and (2) acquiring forest sample point cloud data (figure 2a) by using a ground laser radar scanner according to the technical scheme step (1). And (3) setting the sampling interval of TLS to be 1cm at a position of 10m according to the density of the sample, and totally setting 5 stations, one station at the center and one station at each of four corners of the sample. Three positions are selected to place targets for data stitching processing. And after the data are acquired, splicing by using Cyclone software, and cutting the center area of the sample. Noise points in the data were manually removed. The TLS data is rotated according to the reference point such that the y-axis is in the due north direction. Values for direct solar incidence PAR and scattering PAR were measured using a BF5 radiometer.

Calculating direct PAR of each point according to the step (2) of the technical scheme, calculating scattering PAR of each point according to the step (3) of the technical scheme, and finally obtaining three-dimensional space-time distribution of total PAR of the canopy according to the step (4) of the technical scheme. The accuracy of the model was verified by comparing the model calculation results with the radiometer measurement results for 9 specific locations (table 1). The results of the three-dimensional distribution of the PAR every two hours of the canopy are finally obtained (fig. 2b-2 f).

TABLE 1 comparison of model calculations with radiometer measurements

Claims

1. A method for depicting a forest Photosynthetically Active Radiation (PAR) distribution model by using three-dimensional point cloud data mainly comprises the following steps:

(1) acquiring three-dimensional point cloud data of multiple stations of a forest canopy by using a ground laser radar;

(2) calculating direct projection PAR of each point in forest canopy three-dimensional point cloud data: rotating the point cloud according to the incident direction of the sun to enable the solar ray to be parallel to the-z axis direction, calculating the penetration coefficient of each point directly irradiating PAR by using a voxel method, and calculating the normal vector of each point by a near point texture surface mode; obtaining the direct incidence PAR of each point according to the penetration coefficient of the direct incidence PAR of each point, the direct incidence PAR of the sun measured by a radiometer and the cosine value of the included angle between the normal vector of each point and the incident direction of the sun;

(3) calculating the scattering PAR of each point in the three-dimensional point cloud data of the forest canopy: in order to obtain the scattering PAR of a certain point in the point cloud data, the three-dimensional point cloud data of the canopy layer needs to be moved so that the point to be obtained is a central point, a coordinate system is converted into spherical coordinates, the upper hemisphere of the canopy layer is divided into different cone-shaped body areas by setting the inclination angle and azimuth angle resolution, and the scattering PAR penetration coefficient of each point is calculated according to the proportion of the cone-shaped body without the points; obtaining the scattering PAR of each point according to the penetration coefficient of the scattering PAR of each point and the incident scattering PAR of the sun measured by a radiometer;

(4) and finally obtaining a three-dimensional space-time distribution result of the forest canopy PAR at a specific moment according to the direct PAR and the scattering PAR of each point.

2. The method for characterizing the distribution model of the forest photosynthetically active radiation by using the three-dimensional point cloud data as claimed in claim 1, wherein in the steps (1) (2) (3), the three-dimensional distribution of the forest canopy PAR at any moment is calculated on a fine scale by using the ground lidar data by considering the transmission process of the direct solar radiation and the scattering PAR in the canopy and considering the influence of the canopy element angle on the PAR three-dimensional distribution.