CN108388683A - A kind of vegetation pattern spatial simulation method based on factor of the habitat - Google Patents
A kind of vegetation pattern spatial simulation method based on factor of the habitat Download PDFInfo
- Publication number
- CN108388683A CN108388683A CN201710450532.9A CN201710450532A CN108388683A CN 108388683 A CN108388683 A CN 108388683A CN 201710450532 A CN201710450532 A CN 201710450532A CN 108388683 A CN108388683 A CN 108388683A
- Authority
- CN
- China
- Prior art keywords
- habitat
- factor
- vegetation
- vegetation pattern
- data
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of new vegetation pattern spatial simulation methods based on factor of the habitat.This method is directed to the space complexity and nonlinear characteristic of vegetation distribution, introduces SVM algorithm, constructs the vegetation pattern spatial simulation model based on factor of the habitat, includes mainly three steps:First, using remote sensing and GIS spacial analytical methods, the pre- processing of factors of the habitat basic data such as case area natural climate, soil relief, population traffic are realized;Secondly, using Radial basis kernel function(RBF)As the kernel function of SVM, SVM models are trained with by pretreated basic data, and use grid data service search model optimized parameter, and then build the vegetation pattern spatial simulation model based on factor of the habitat;Finally, model is run and verified, and then realizes the spatial distribution simulation of case area vegetation pattern.This method not only facilitates the quantitative simulation of vegetation pattern and its spatial distribution in realization in macro-scale, and has fine simulation precision and very strong operability.
Description
Technical field
The vegetation pattern spatial simulation method based on factor of the habitat that the present invention relates to a kind of, be suitable for Global climate change and
Research on Ecological Effect field.
Background technology
Relationship and Study on mechanism between vegetation distribution and related factor of the habitat are vegetation Spatial Distribution Pattern moulds
Quasi- precondition.For the topography and landform character of different zones, natural climate condition and mankind's activity annoyance level, quantitatively take off
Show interaction volume, the mechanism of action between analysis vegetation distribution and factor of the habitat, is understanding vegetation distribution shape
Condition, sunykatuib analysis vegetation pattern spatial distribution, variation tendency and its future scenarios core content.At present in vegetation spatial distribution
It is main using the method for including field sample prescription-line-transect on-site inspection method, RS and GIS technology combination in research, to vegetation class
Type spatial distribution is interpreted and delineates.Conventional method not only needs to carry out a large amount of field to sample and expend a large amount of manpowers on the spot
Material resources, and the region extremely complex in natural conditions and topography and landform character, can not carry out sample activity, so that traditional
The vegetation pattern in these regions can not be identified and analyzed in method.In addition, conventional method can not be to climate change and the mankind
Vegetation pattern future scenarios under activity-driven are simulated, but are limited only to present situation and past trend comparative analysis.With
Remote sensing technology and GIS technology plant information extract and classify etc. fast development, high-precision, high spatial resolution and
The development and application of high spectrum resolution remote sensing technique are sentenced by monitoring vegetation structure information and analysis vegetation index to carry out vegetation pattern
Not, have the characteristics that speed is fast, robustness is high, stablize, improve the precision of identification and classification, to carry out vegetation pattern spatial mode
Quasi- driving parameter provides high feasible data basis.
Vegetation pattern spatial distribution characteristic is that various factors of the habitat interact in specific region as a result, this phase interaction
With being an extremely complex various nonlinear system.And existing research method often samples data mapping and carries out vegetation class
Type classify or spatial analysis, can neither embody vegetation pattern be by the mechanism of influence and the interaction of many habitat factors,
Vegetation pattern sunykatuib analysis in can not meeting in macro-scale.Therefore, how habitat element and vegetation class to be analysed in depth
On the basis of type and its spatial distribution interaction mechanism, using the vegetation pattern sampling point data of finite discrete, in conjunction with high-resolution
Rate remote sensing image data, and consider various factors of the habitat, build vegetation pattern spatial distribution simulation model, Jin Ershi
The sunykatuib analysis of existing vegetation pattern spatial distribution is asked vegetation pattern spatial distribution simulation urgent need to resolve in large scale
Topic.Support vector machines (Support Vector Machine, abbreviation SVM) theory is one based on structural risk minimization theory
Kind machine learning algorithm(Vapnik V,1998, Statistical Learning Theory. Wiley, New York),
Solution to solve nonlinear regression provides a kind of new thinking, all more has in speed and precision compared to more traditional method
Apparent advantage is succeeded application in many fields.Map construction especially by low-dimensional to the higher-dimension inner product space is linear
Grader has training speed and convergence rate quickly, together in solving the problems, such as small sample, non-linear and high dimensional pattern identification
When have preferable generalization ability and higher nicety of grading, be very suitable for utilize more factors of the habitat and small discrete data
Carry out the application of large scale vegetation pattern spatial distribution simulation.
Therefore, it for the space complexity and nonlinear characteristic due to vegetation distribution, is obtained with linear classification method
To simulation precision be difficult meet demand problem in science, make full use of SVM methods in the case where sample size is less, can also obtain
The algorithm advantage of good classification result and statistical law is obtained, the present invention is considering natural climate, soil relief, ma n-made factor
On the basis of equal factors of the habitat, SVM algorithm is introduced, the sample of each vegetation pattern is trained, and then determines kernel function
Optimized parameter, it is final to build the vegetation pattern spatial simulation model based on factor of the habitat.This method not only facilitates macro in realization
The quantitative simulation of the vegetation pattern and its spatial distribution on scale is seen, and with good simulation precision and very strong operable
Property.
Invention content
The present invention is directed to consider with vegetation be distributed the habitats such as relevant natural climate, soil relief, ma n-made factor because
Son automatically extracts model, the plant in realization in macro-scale using vegetation pattern spatial distribution of the SVM structures based on factor of the habitat
The quantitative simulation being distributed by type space.To achieve the goals above, the key technology scheme that the present invention realizes includes following several
A step:
The first step, data collection and pretreatment
1-1)For research object and research range, case area is selected.To case area natural climate, soil relief, ma n-made factor
The historical statistics observation data and case area vegetation pattern history field investigation data of equal factors of the habitat, are collected respectively
And pretreatment realizes that its high-precision spatialization is handled with GIS spacial analytical methods;
1-2)Spatial distribution data based on treated factor of the habitat and vegetation pattern, for lacking for history surveyed and statistic data
It falls into, designed in case area and arranges that specimen sample investigates scheme, in conjunction with remote sensing image, carry out field sampling investigation, supplement is perfect
Entire case area factor of the habitat and vegetation pattern sample data;
1-3)Unified normalized is carried out to all factor of the habitat data and vegetation pattern data.
Second step, vegetation pattern spatial simulation model construction and parameter selection
2-1)The normalization data of factor of the habitat and vegetation pattern sampling point is obtained about based on the first step, selection is using radial base
Kernel function(RBF)As the kernel function of SVM, factor of the habitat and vegetation sampled data of the application case area Jing Guo normalized,
Model training and study are carried out to SVM models;
2-2)The optimized parameter of SVM models is searched for grid data service;
2-3)On the basis of determining optimized parameter, realize that the vegetation pattern based on factor of the habitat automatically extracts the structure of model.
Third step, the simulation of vegetation pattern spatial distribution
3-1)It is based on habitat in conjunction with by the unified pretreated factor of the habitat model parameter of normalization and vegetation sampled data, operation
The vegetation pattern of the factor automatically extracts model, preliminary to realize automatically extracting for case area's vegetation pattern spatial distribution;
3-2)The vegetation pattern spatial distribution as-is data obtained is sampled using based on high-resolution remote sensing image and field investigation,
Verification analysis is carried out to model simulation results;
3-3)Binding model verification result repeats second step content, carries out repetition learning and amendment to SVM models, and then obtain
The optimized parameter of vegetation pattern classification, until by precision test;
3-4)Using the final SVM models debugged and by the optimized parameter collection of inspection, in conjunction with the factor of the habitat number in case area
According to the automatic classification and spatial distribution simulation of realization case area vegetation pattern.
Specific implementation mode
In order to make the purpose of the present invention, technical solution be more clear, the present invention is described in more detail below.It should manage
Solution, specific real-time mode described herein are only used to explain the present invention, be not intended to limit the present invention.
The vegetation pattern spatial simulation method based on factor of the habitat that the invention discloses a kind of, includes the following steps:
The collection and pretreatment of the first step, the related factor of the habitat data in research area;
Second step, vegetation pattern spatial simulation model construction and parameter selection;
Third step, the simulation of vegetation pattern spatial distribution.
Specific steps are described in detail below:
Study the collection and pretreatment of the related factor of the habitat data in area.Based on having between vegetation pattern space and its factor of the habitat
There is very strong correlative character, therefore in the factor of the habitat data in collection research area, has considered landform, the gas in research area
The environmental elements such as time, population.First, dem data, vegetation sample investigation data, Eco-hydrological belt transect survey data, multi-source are collected
High-definition remote sensing image data, and to the digitlization on each plant covering boundary in vegetation distribution map, complete data update, obtain
Vegetation pattern spatial distribution;Secondly, the meteorological measuring to the weather routine observation website for studying area, extracts the daily of record
Precipitation, temperature value are arranged and obtain ten annual means of two periods of each site location respectively, built using high speed high-precision curved
Mould method(HASM)Method simulates to obtain the spatial resolution in research area as the mean annual precipitation of 100m, the sky of average organism temperature
Between distributed data.Further include the population distribution data and traffic accessibility data for studying area's 100m resolution ratio simultaneously;Finally, for
The ecology ring such as dem data, for many years mean precipitation data, for many years average organism temperature data, demographic data, traffic accessibility data
The otherness feature of the numberical range of border supplemental characteristic, to avoid partial factors from occupying excessive power in model construction process
Weight, influences model performance, data normalization processing is carried out to basic data machine, to solve the comparativity between data target.
Data normalization processing method is as follows:
(1)
In formula,For the maximum value of all kinds of factor sample datas,For the minimum value of sample data.
Vegetation pattern spatial simulation model construction and parameter selection.First, it is collected for the spatial data of vegetation pattern
The features such as difficult and space complexity, the normalization number of factor of the habitat and vegetation pattern sampling point is obtained about based on the first step
According to selection uses Radial basis kernel function(RBF)Kernel function as SVM.Radial basis kernel function(Gaussian kernel function)Theory it is public
Formula is:
(2)
Secondly, regression function is constructed:
(3)
And construct its Lagrange function:
(4)
In formula,, dual problem is:
(5)
(6)
Solving acquisition nonlinear solshing is:
(7)
Finally, selection is optimized to SVM parameters.The selection of SVM parameters is structure vegetation pattern spatial distribution simulation model
Key, the present invention use Radial basis kernel function(RBF)Vegetation distribution model is established as basic function.Therefore, it is necessary to Gauss
The optimum choice of the width parameter σ and error punishment parameter C of distribution.The selection of optimal nuclear parameter is carried out with grid data service,
Detailed process includes:A) range of SVM parameters σ and penalty coefficient C, the value of setting grid search space C values and σ values are set
Range (C ∈ [2C1,2C2], σ∈[2σ1,2σ2], b) it constructs using C values and σ values as the two-dimensional grid search space of coordinate, d)Setting
The step-size in search of its grid, and traverse grid search space and be trained for each pair of (C, σ) parameter, determine optimized parameter group
It closes, using K-fold cross validations and records its classification accuracy.
Vegetation pattern spatial distribution is simulated.On the basis of data set is normalized, wrapped with each grid lattice point
Each environmental element data value contained constitutes vector and is used as sample data, using the stochastical sampling method that do not put back to, according to fixed
Oversampling ratio obtains training sample data collection, and for determining the parameter in model and building model, remaining data collection is as test
Classifying quality.Since different vegetation types are different for the sensibility of environmental factor, the otherness of optimized parameter in model construction
Clearly, it is therefore desirable to it sets different parameters according to different vegetation types and is simulated to establish multiple spatial distribution models,
Its result more has characteristic and reasonability.
The simulation model of Heihe River basin vegetation pattern spatial distribution is constructed using the present invention, and on vegetation group level
The spatial distribution of each vegetation pattern is simulated, the overall accuracy of analog result is 76%.The highest vegetation pattern of precision
It is the short fruticuli desert of fruticuli-, temperate zone dogstail-forb meadow grassland, alpine cushion-like vegetation, has respectively reached 90%.Show
Method proposed by the present invention has preferable feasibility and operability on vegetation distribution simulation, is suitable under big regional scale
The spatial simulation of vegetation distribution.
Claims (1)
1. a kind of vegetation pattern spatial simulation method based on factor of the habitat includes mainly following three steps:
The first step, data collection and pretreatment
1-1)For research object and research range selection case area, to case area natural climate, soil relief, ma n-made factor etc.
The historical statistics observation data and case area vegetation pattern history field investigation data of factor of the habitat, be collected respectively and
Pretreatment realizes that its high-precision spatialization is handled with GIS spacial analytical methods;
1-2)Spatial distribution data based on treated factor of the habitat and vegetation pattern, for lacking for history surveyed and statistic data
It falls into, designed in case area and arranges that specimen sample investigates scheme, in conjunction with remote sensing image, carry out field sampling investigation, supplement is perfect
Entire case area factor of the habitat and vegetation pattern sample data;
1-3)Unified normalized is carried out to all factor of the habitat data and vegetation pattern data
Second step, vegetation pattern spatial simulation model construction and parameter selection
2-1)The normalization data of factor of the habitat and vegetation pattern sampling point is obtained about based on the first step, selection is using radial base
Kernel function(RBF)As the kernel function of SVM, factor of the habitat and vegetation sampled data of the application case area Jing Guo normalized,
Model training and study are carried out to SVM models;
2-2)The optimized parameter of SVM models is searched for grid data service;
2-3)On the basis of determining optimized parameter, realize that the vegetation pattern based on factor of the habitat automatically extracts the structure of model
Third step, the simulation of vegetation pattern spatial distribution
3-1)It is based on habitat in conjunction with by the unified pretreated factor of the habitat model parameter of normalization and vegetation sampled data, operation
The vegetation pattern of the factor automatically extracts model, preliminary to realize automatically extracting for case area's vegetation pattern spatial distribution;
3-2)The vegetation pattern spatial distribution as-is data obtained is sampled using based on high-resolution remote sensing image and field investigation,
Verification analysis is carried out to model simulation results;
3-3)Binding model verification result repeats second step content, carries out repetition learning and amendment to SVM models, and then obtain
The optimized parameter of vegetation pattern classification, until by precision test;
3-4)Using the final SVM models debugged and by the optimized parameter collection of inspection, in conjunction with the factor of the habitat number in case area
According to the automatic classification and spatial distribution simulation of realization case area vegetation pattern.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710450532.9A CN108388683A (en) | 2017-06-15 | 2017-06-15 | A kind of vegetation pattern spatial simulation method based on factor of the habitat |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710450532.9A CN108388683A (en) | 2017-06-15 | 2017-06-15 | A kind of vegetation pattern spatial simulation method based on factor of the habitat |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108388683A true CN108388683A (en) | 2018-08-10 |
Family
ID=63076426
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710450532.9A Pending CN108388683A (en) | 2017-06-15 | 2017-06-15 | A kind of vegetation pattern spatial simulation method based on factor of the habitat |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108388683A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113095619A (en) * | 2021-03-04 | 2021-07-09 | 广东省科学院广州地理研究所 | Method and system for simulating vegetation productivity space pattern based on climate and soil |
CN113348473A (en) * | 2019-01-24 | 2021-09-03 | Abb瑞士股份有限公司 | Installation foundation for managing artificial intelligence module |
CN113361350A (en) * | 2021-05-25 | 2021-09-07 | 东南大学 | Quantitative analysis method for surface habitat factors |
CN113919226A (en) * | 2021-10-15 | 2022-01-11 | 中国矿业大学(北京) | Mining vegetation ecological cumulative effect disturbance range identification method based on weight |
CN113348473B (en) * | 2019-01-24 | 2024-05-28 | Abb瑞士股份有限公司 | Management artificial intelligence module installation foundation |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102682202A (en) * | 2012-04-28 | 2012-09-19 | 清华大学 | Green space system community planning method for urban and rural ecological planning |
CN103164608A (en) * | 2011-12-15 | 2013-06-19 | 交通运输部科学研究院 | Judgment method of animal traffic hotspots based on habitat factors |
CN104778451A (en) * | 2015-03-31 | 2015-07-15 | 中国科学院上海技术物理研究所 | Grassland biomass remote sensing inversion method considering grassland height factor |
CN106204611A (en) * | 2016-07-19 | 2016-12-07 | 中国科学院地理科学与资源研究所 | A kind of LiDAR point cloud data processing method based on HASM model and device |
-
2017
- 2017-06-15 CN CN201710450532.9A patent/CN108388683A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103164608A (en) * | 2011-12-15 | 2013-06-19 | 交通运输部科学研究院 | Judgment method of animal traffic hotspots based on habitat factors |
CN102682202A (en) * | 2012-04-28 | 2012-09-19 | 清华大学 | Green space system community planning method for urban and rural ecological planning |
CN104778451A (en) * | 2015-03-31 | 2015-07-15 | 中国科学院上海技术物理研究所 | Grassland biomass remote sensing inversion method considering grassland height factor |
CN106204611A (en) * | 2016-07-19 | 2016-12-07 | 中国科学院地理科学与资源研究所 | A kind of LiDAR point cloud data processing method based on HASM model and device |
Non-Patent Citations (1)
Title |
---|
周勋 等: "黑河流域植被类型分布模拟分析", 《地球信息科学学报》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113348473A (en) * | 2019-01-24 | 2021-09-03 | Abb瑞士股份有限公司 | Installation foundation for managing artificial intelligence module |
CN113348473B (en) * | 2019-01-24 | 2024-05-28 | Abb瑞士股份有限公司 | Management artificial intelligence module installation foundation |
CN113095619A (en) * | 2021-03-04 | 2021-07-09 | 广东省科学院广州地理研究所 | Method and system for simulating vegetation productivity space pattern based on climate and soil |
CN113361350A (en) * | 2021-05-25 | 2021-09-07 | 东南大学 | Quantitative analysis method for surface habitat factors |
CN113919226A (en) * | 2021-10-15 | 2022-01-11 | 中国矿业大学(北京) | Mining vegetation ecological cumulative effect disturbance range identification method based on weight |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109902885A (en) | Typhoon prediction method based on deep learning mixed CNN-LSTM model | |
Jiang et al. | Estimation of land production and its response to cultivated land conversion in North China Plain | |
Song et al. | Impacts of environmental heterogeneity on moss diversity and distribution of Didymodon (Pottiaceae) in Tibet, China | |
Wieland et al. | Using Shapley additive explanations to interpret extreme gradient boosting predictions of grassland degradation in Xilingol, China | |
Guo et al. | Partitioning beta diversity in a tropical karst seasonal rainforest in Southern China | |
CN113592132B (en) | Rainfall objective forecasting method based on numerical weather forecast and artificial intelligence | |
CN108388683A (en) | A kind of vegetation pattern spatial simulation method based on factor of the habitat | |
Yaojie et al. | A data‐mining‐based approach for aeolian desertification susceptibility assessment: A case‐study from Northern China | |
Zhang et al. | Spatial domain bridge transfer: An automated paddy rice mapping method with no training data required and decreased image inputs for the large cloudy area | |
Liu et al. | Change trend and restoration potential of vegetation net primary productivity in China over the past 20 years | |
LI et al. | Spatio-temporal changes in rice area at the northern limits of the rice cropping system in China from 1984 to 2013 | |
Murakami et al. | Prediction of municipality-level winter wheat yield based on meteorological data using machine learning in Hokkaido, Japan | |
Wang et al. | Prediction of potential distribution areas and priority protected areas of Agastache rugosa based on Maxent model and Marxan model | |
Zhang et al. | Inversion models of aboveground grassland biomass in Xinjiang based on multisource data | |
Sheng et al. | Predicting isoscapes based on an environmental similarity model for the geographical origin of Chinese rice | |
Wang et al. | Mapping the ratoon rice suitability region in China using random forest and recursive feature elimination modeling | |
CN116128183B (en) | Method for evaluating influence of diversion engineering on lake wetland habitat change | |
CN115983478A (en) | Distributed photovoltaic power generation power prediction analysis method, system, terminal and medium | |
Zhang et al. | An agent-based reasoning of impacts of regional climate changes on land use changes in the Three-River Headwaters Region of China | |
CN109960781A (en) | A method of updating world crops total output raster data | |
CN108985493A (en) | A kind of ground class variation prediction method based on self-adapting changeable filter | |
Inoue et al. | Effects of individual size, local competition and canopy closure on the stem volume growth in a monoclonal Japanese cedar (Cryptomeria japonica D. Don) plantation | |
Xiao et al. | A spatialization method for grain yield statistical data: A study on winter wheat of Shandong Province, China | |
Sun et al. | Evaluating the net primary productivity in the grasslands of southern China from 2001 to 2010 using a new land portfolio assessment model | |
Wang et al. | Spatiotemporal variations in drought and waterlogging and their effects on maize yields at different growth stages in Jilin Province, China |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180810 |