Spatial simulation model acquisition future land cover tutorial

origin

I've been reading my previous blog these days , Turn to this one ：

At that time, I didn't ask for an understanding , No deep exploration , Until I saw it YouTube There is one. up The main course of spatial prediction simulation （ In the reference website ）, So it aroused my interest .

brief introduction

Spatial prediction model has the advantage of spatial simulation , Reflect position change information , Predict the distribution change of spatial pattern , It can simulate the spatiotemporal evolution process of complex systems. Such models are mainly used to determine land use ／ Patterns and processes of land cover change , And predict the location of future changes .

This kind of model is in time 、 Discrete grid dynamic model in space and state , Be able to deduce the utilization of excavated land from empirical data ／ Transition rules covering changes , It can complete a series of tasks in a complex and dynamic environment , And can adapt to and perceive changes in the environment , These features are widely used to simulate the interaction between different activities . The relevant principles of spatial prediction model can be referred to ：

technology roadmap

Operation process

Image preprocessing

Data preparation

The spatial model prediction of land cover data needs to prepare land cover data for many years , Road data and DEM data .

The selected study area is Lezhi county, Ziyang City, Sichuan Province （ Yes? ** Here again ？** Because Lezhi county is my hometown , ha-ha ）.

DEM The data is 30m Resolution NASA DEM, The road data is 2020 year open street map Road vector of . Land cover data is 2009 year 、2014 year 、2019 Wuhan University data set in .

The image factor can include not only the traffic distance 、dem, You can also add soil data 、 Meteorological data, etc . Here is to show , Don't delve into .

The purpose of the study is to 2009 Years and 2014 Land cover data of Lezhi county in trained spatial prediction model , get 2019 Lezhi county in Forecast land cover image , And evaluate the results .

Projection and resolution processing

Raster data processing

Use GCS_WGS_1984 Geographical coordinate system and WGS_1984_UTM_Zone_48N Projection coordinate system , Coordinate system conversion and re projection of all data ：

The projection conversion of raster data uses the data frame included in the export function , And set the resolution to 30m：

Processing of vector data ：

The conversion of vector data uses the projection function

After the road data is projected , The traffic distance map needs to be calculated , That is to calculate the distance from each pixel in Lezhi county to the nearest road . The Euclidean distance calculation tool is used here ：

Note that the resolution is set to 30m：

And set the space processing range in the environment tool to Lezhi county ：

To prevent bug, You need to set the parallel processing factor to... In the environment settings 0：

Click on the run , Get traffic distance grid ：

Data clipping and pixel alignment

Use the grid clipping tool （ Data management - grid - Grid processing - tailoring ）, Don't choose Mask extraction ：

Yes 2009 Annual land cover 、2014 Annual land cover image 、dem Image and traffic distance image are cropped

To ensure that the number of pixels of each data is the same and to achieve pixel alignment , When cutting, you must check Crop geometry and maintain crop range . At this time, open the properties of all images , At this time, the input data are the same row and column number and resolution .

Space simulation

Tool is introduced

MOLUSCE,Methods Of Land Use Chande Evaluation, It is used for land use change assessment QGIS plug-in unit , It mainly has the following functions ：

• Get time periods A（ In the past ） Land use category grid 、 Period of time B（ Now? ） Land use category grid and explanatory variable grid .
• Train a model to predict land use change from the past to the present .
• Use derived models 、 Land use status and current factors predict future land use changes .

model training

install qgis（ Need to downgrade to 2 Series version , Do not use 3 edition ）, Install the plug-in in the plug-in center .

Add land cover data and spatial impact elements ,

Click to check Geometry . The checking geometry is mainly the coordinate system of the comparison data 、 The resolution of the 、 Pixel alignment of these three contents , Be sure to follow the data processing preprocessing steps above , This step can pass the inspection . If there is no problem, you can go to the next step of prediction .

Choose check all raster, And calculate the correlation coefficient of each image factor , The choice of algorithm can be arbitrary , The algorithm I choose here is the Pearson product moment correlation coefficient ：

Go to the next step , Calculate Lezhi county 2009 Year to 2014 Land change matrix in , The change image needs to be exported ：

Complete the appeal steps , You can start training the model ：

Model to predict

After data training , Can simulate the predicted image ：

This simulation process takes a long time , Wait patiently ：

Accuracy evaluation

When the forecast is completed , The tool can be used for accuracy evaluation ：

Operation video

It is divided into two parts: image preprocessing and spatial simulation , Because my headset is noisy , I will not introduce the relevant language ：

The pre processed video is as follows ：

Spatial prediction simulation - Data preprocessing

The video of space simulation is as follows ：

Spatial prediction simulation - Model prediction part

At the end

Because I haven't touched the field of spatial prediction simulation before , There is inevitably something wrong , Feel free to leave a comment .
View more data tutorials , have access to “ Remote sensing geoscience data collection ”：
https://www.gisrsdata.com/

Reference resources

[1] Luyuting . Land use of Hefei Based on remote sensing spatial-temporal fusion / Analysis and Simulation of land cover change [D]. Anhui University ,2019.

[2] Li Jing , Fanzemeng , Yuetianxiang . Spatiotemporal simulation of land cover in Southwest China [J]. Journal of Ecology ,2014,34(12):3266-3275.

[3]Prediction of Land Use/Land Cover Change using QGIS and ArcGIS (2010-2020-2030).https://www.youtube.com/watch?v=07_Z4JraSFc&list=PL-oOgWm8Lw-9Ck1vAmku9m1mZNACrzXz7&index=7&ab_channel=GIS%26RSSolution