Land use land cover classification using QuantumGIS; Study case mangrove forest in Segara Anakan, Indonesia

This time I would like to present you how to deal with QuantumGIS, especially how to produce land use land cover (LULC) map extracted from Landsat 8 OLI

In this case, I would like to measure the present condition of mangrove forest only based on the reflectance characteristic, study case mangrove forest in Segara Anakan, Indonesia.

Study area: Segara Anakan, Indonesia
Study area: Segara Anakan, Indonesia

The algorithm of the image interpretation will be as follow;

1. DN conversion into TOA (Top of Atmospheric Correction)

2. Preparing band input for used as dataset

3. Create training area, to “guide” the software

4. Perform LULC classification

5. Calculate the mangrove forest area

 

Before we begin, its better for you to check your computer specification. If your computer is Windows 64bit, you need to download and install first the WinPython. Then you need to extract the installer with 7-zip (see Figure 1 to teach you how to extract the installer). Inside $_OUTDIR/python-2.7.6.amd64/Lib/site-packages you can find the scipy , the matplotlib, and the numpy directories that you can copy to the QGIS installation directory (see Figure 2); then install the Semi-Automatic Classification Plugin for QGIS.

Figure 1. How to extract the installer of WinPython
Figure 1. Right click, and the installer of WinPython into the folder

 

Figure 2. Directory to save the folders (scipy , matplotlib, and numpy)

To install the Semi-Automatic Classification Plugin, run your QGIS and click Plugins – Manage and Install Plugins.. – type “semi” in the Plugins window as shown in Figure 3. Then click “Install plugin”

Figure 3. How to install the Semi-Automatic Classification Plugin in QGIS
Figure 3. How to install the Semi-Automatic Classification Plugin in QGIS

1. DN conversion into TOA (Top of Atmospheric Correction)

In the satellite imagery interpretation, the conversion of DN into TOA is crucial since we need the real physical reflectance values of the specific land use land cover. In this study we employ DOS (Dark Object Subtraction). If you are interested more in DOS method, please read it from the following paper.

Run your QGIS and start the Semi-Automatic Classification Plugin; in the main interface select the tab Pre processing > Landsat > select the directory which is consist of your Landsat 8 OLI dataset with the MTL.txt file. Also select the “Output directory” of the converted bands.

Check (x) the option of “Apply DOS1 atmospheric correction” and click “Perform conversion”

Figure 4. Activating the plugin and converting the DN into TOA
Figure 4. Activating the plugin and converting the DN into TOA

After the conversion process finished, you could check the result using the “identify feature” modul. Move the mouse to the particular area and just give single click, then the new window will be appear and shows the reflectance value (see Figure 5)

Figure 5. The value of DN and reflectance
Figure 5. The value of DN and reflectance

2. Preparing band input for used as dataset

We need to produce the color composite of the study area, so it will make us easier to select the training data. We will produce the composite RGB of 543 (RGB = 432 for Landsat 7) is useful for the interpretation of mangrove forests in the image, because of healthy vegetation reflects a large part of the incident light in the near-infrared wavelength.

Go to the main toolbar of QGIS, and select Raster – Miscellaneous – Build Virtual Raster (catalog); click the button “Select” in the input file section… and select the bands 5, 4, and 3 (do not forget to hold +Ctrl); click the button “Select” in the output file section, and give the output name (for example RGB.vrt). Check “Separate” and click “OK”.

Until this step, you have been successfully produced the RGB image of the study area.

Figure 6. Producing dataset for multispectral data
Figure 6. Producing dataset for multispectral data

In the window of  Semi-Automatic Classification Plugin, click the button “Band set” then the tab Band set will open; click the button “Refresh list”, then Add rasters to set (order the band names in ascending order, from top to bottom using ↑ and ↓ arrows. Select only the dataset with RT_ in the file name. You can also check the center wavelength of the data as shown in Figure 7.

Figure 7. Producing dataset and check the center wavelength
Figure 7. Producing dataset and check the center wavelength

3. Create training area, to “guide” the software

To create the training area, in the vertical-right side of QGIS window, under the “Training shapefile” click the button “New shp”, and select where to save the shapefile, and give the name of your ROI-Region of Interest (for example ROI.shp).

In order to “guide” the software, we have to to create some ROI, considering the reflectance variability of LULC classes. It is important that ROIs represent homogeneous areas of the image, therefore we are going to draw the ROIs using a region process (i.e. a segmentation of the image, grouping similar pixels).

The following are the land cover classes that we are going to identify in the image:
1. Water (e.g. sea water, river, and lake)
2. Agriculture (e.g. paddy field areas)
3. Another crops
4. Mangrove forests
5. Forest
6. Fish pond (e.g. aquacultures)
7. Cloud cover

When you are selecting the training area, please consider to select more than 3 of the training data of each LULC. Based on my experience, if we only select a few of them, it will lead to some error classification.

Figure 8 shows how to create training data. Click the (+) symbol to add the “box” training data and click the “Create ROI polygon” to produce free shapefile. Give it the name and ID value to each training data. After finished it, do not forget to click “Save ROI” and “Add to signature”.

Figure 8. Creating shapefiles as training areas
Figure 8. Creating shapefiles as training areas

If you would like to produce some spectral signature plot or scatter plot, just select all the training data and click the two symbol next to “Add to signature” as shown in the Figure 9 below. You can also produce particular training data, for instance only “mangrove”. Then just select the “mangrove”, and click the spectral signature symbol.

Figure 9. Producing scatter plot and signature plot of the training data
Figure 9. Producing scatter plot and signature plot of the training data

4. Perform LULC classification

The remotely-sensed image classification can be performed through several algorithms; in this study we will use the Maximum Likelihood classification.

Before the software produce the final result of the classification, it is possible to preview the classification. In the “Classification algorithm”, under “Classification preview” type the “Set” Size = 500 then click the button (+) and then click on the image; instantly, the classification preview image will be appeared.

If you already satisfied with the result, then you could click “Perform classification” and give the output name. However, if you would like to modify the training data, you could go back to the stage 3.

Its done! Congratulation, you successfully produce the LULC of Segara Anakan, Indonesia.

Figure 10. Performing classification of remotely-sensed data of the study area
Figure 10. Performing classification of remotely-sensed data of the study area. Mangrove forests represented in light-green color.

 

5. Calculate the mangrove forest area

To calculate the LULC, especially the mangrove forest, go back to the Semi-Automatic Classification Plugin main window and select “Post processing – Classification report”.

In the drop down menu select the “Class_Result”, this is the result of your classification. Then click “Calculate classification report”.

 

Figure 11. Calculating the LULC
Figure 11. Calculating the LULC

This study successfully documented the present (May 1st, 2014) area of mangrove forests in Segara Anakan Indonesia. There are about 33.13 square kilometers of mangrove forests in the study area. However, this is not reflected the real number since I did not conduct any field work or accuracy assessment. This is only to show you how to work in QuantumGIS environment to produce image classification.

If you would like to do some accuracy assessment measurement, you could use the “Accuracy” modul in this software. Just input your reference data from the field work or from the official database and then calculate the accuracy of your imagery classification.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s