<?xml version="1.0"?>
<api>
  <query-continue>
    <allpages gapfrom="Relascope" />
  </query-continue>
  <query>
    <pages>
      <page pageid="1462" ns="0" title="Reducing Emissions from Deforestation and Forest Degradation (REDD)">
        <revisions>
          <rev xml:space="preserve">{{Ficontent}}{{author|name=[[User:Pbecksc|Philip Beckschäfer]]}}
The [[Kyoto Protocol]] requires Annex-I countries to account for the carbon changes associated with [[afforestation]], [[reforestation]], [[deforestation]] and all land use activities undertaken since 1990. Developing countries can, however, only claim credits generated from afforestation and reforestation through the CDM but not from avoided deforestation. REDD addresses the issue that almost 20% of global GHG emissions results from deforestation in developing countries. The basic concept of REDD is simple: governments, companies or forest owners in developing countries should be rewarded for keeping their forests instead of cutting them down.

Under the [[UNFCCC]] REDD was first discussed at COP 11 (2005) but it took until COP16 (2010) in Cancún to reach an agreement on REDD. Issues on financing, governance in developing countries or the non-permanence of carbon credits based on forestry projects are still under discussion. The term “non-permanence” refers to the risk that a forestry project for which carbon credits were issued burns down or is destroyed by pests and consequently the stored carbon is released back to the atmosphere. But there are as well crucial issues related to forest inventory. How to adequately estimate and monitor changes in forest cover, associated carbon stocks and greenhouse gas emissions, incremental changes due to sustainable forest management, reductions in emissions from deforestation, and reductions in emissions from forest degradation is one still unsolved problem. In this context, it is especially forest degradation that causes problems. There is still no agreement on a definition of what a degraded forest is. Also the detection of forest degradation by means of remote sensing remains demanding and costly. A further question that needs to be answered in order to implement a successful REDD scheme is how to establish the relevant baseline or reference emissions levels against which reductions will be measured? To calculate the amount of credits which are issued it needs to be measured, how much the REDD project contributed to the decrease in emissions from deforestation and forest degradation. Therefore, it is necessary to know what would have happened without the project, what would be the business-as-usual scenario (BAU scenario)? The calculation of a BAU scenario is done on the basis of historical deforestation rates, which are prolonged into the future. A case from Australia showed that the creation of a reliable BAU scenario is highly complex and harbors a lot of uncertainties. According to The Australia Institute (2010), the prediction of Australia’s future deforestation rates was not adequately possible even over short time frames. For some years, the actual deforestation emissions have been in excess of 80% higher than the BAU projections. For the Australian government to have accurately predicted these emissions, it would have had to foresee changes in commodity prices, rainfall and other relevant social and economic factors. If we take into account that Australia has one of the most advanced satellite-based systems for monitoring deforestation emissions in the world, the National Carbon Accounting System (NCAS), it seems unrealistic that the prediction of deforestation in developing countries can be more accurate. 

==REDD and REDD+==
The terms REDD and REDD+ describe different forms of a scheme aiming at the reduction of emissions from deforestation and forest degradation in developing countries. The pure REDD focuses only on the reduction of emissions from deforestation and forest degradation. REDD+ broadens the scope and adds conservation, sustainable management of forests, and the enhancement of forest carbon stocks to REDD. REDD++ or AFOLU aims at including all transitions in land cover that affect carbon storage (e.g. peat land, trees-outside-forests, agroforestry system) into the final scheme. At COP 16 in Cancún, considerable progress has been made towards making REDD an operational instrument. The Cancún Agreement does also contain the statement that countries shall implement national forest monitoring systems that allow a transparent verification of the forest carbon dynamics.

[[Category:UNFCCC Implications for Forest Monitoring]]</rev>
        </revisions>
      </page>
      <page pageid="1950" ns="0" title="Region Growing Segmentation">
        <revisions>
          <rev xml:space="preserve">=Region Growing Segmentation with Saga's Seeded Region Growing Tool=

The following tutorial by Sebastian Kasanmascheff explains how to delineate tree crowns, using SAGA's Seeded Region Growing Tool. The product, a polygon shapefile, can then be used in an object-based classification, f.ex. in order to classify different tree species.

=Material you need to complete the tutorial=

'''Data'''

A multispectral image of the forest canopy

A canopy height model [[Canopy_Height_Model_based_on_Airborne_Laserscanning_using_LAStools#Create_a_CHM_directly_from_height-normalized_points:_lasthin_and_las2dem|CHM]] 

'''Software''' 

QGIS 2.18.11

SAGA 2.3.2



=Split multiband image into several raster images=

SAGA's Region Growing Algorithm works only with single band images. Therefor, we have to split our multiband image into its individual bands following [[Split_stack|these instructions]].



=Seed points=


The first step here is to extract  the position of the tree tops, which are going to be the starting point for the region growing algorithm. You find a description of how to derive the seed points from a CHM [[Individual_Tree_Detection_(ITC)|here]] .
Now have a look at how the seeds align with your multiband image


[[File:Unfiltered_seed_points.PNG]]


In this picture you see, that the points align well with the tree tops. However, there are many points that represent small trees (in the lower left corner) which you might not be so interested in. In order to correct that, we are going to filter the points by their height in the seed shapefile and save the shapefile with the selected seeds only, as a new shapefile.
If the seed points do not fit the image well, it might be due to the fact that you are using an orthophoto and not a true-orthophoto. If you are working only with a smaller dataset you can help it by using the {{button|text=Georeferencer}} in the {{button|text=Raster menu}} by rubbersheeting your orthophoto to make it fit. You'll find the instructions for doing so [[Georeferencing_of_UAV photos|here]].


This looks much better:


[[File:Filtered_seed_points.PNG]]

=Seed point rasterization=


In order to use the seed points in the region growing algorithm in SAGA, we have to convert the vector file to a raster file using the {{button|text=Rasterize}} module from the GDAL Conversions in QGIS. It is critical that we create a raster file that only contains the seed points and no-data. So every pixel outside the seed points has to be no-data. It is also critical that our rasterized vector image has the '''exact same''' CRS (Coordinate reference system), extent and pixel size as the single band images.

* In the {{button|text=Processing Toolbox}}, type 'rasterize' and select the {{button|text=Rasterize}} tool ind the SAGA geoalgorithms submenu
* Select the seed point shapefile as input
* Select {{button|text=ID}} as {{button|text=Attribute}} and the other parameters according to the screenshot below
* Select the extent of one of the band splits as extent by clicking {{button|text=select canvas/ layers extent}} in the {{button|text=Output extent}} line
* Copy the '''exact''' cellsize from either the metadata of one of your split images and set output raster size to {{button|text= Output resolution in map units per pixel}}
* Save to a local file



[[File:Rasterize.PNG]]



The result should look like this:



[[File:Universe2.PNG]]







Now verify that the newly created seed point raster aligns 100% with one of the band split rasters by checking the metadata or the {{button|text=Save raster layer as..}} mask  and by zooming into the seed pixels. Also check with the Info button[[File:Info.PNG]], if any pixel which is not a seed point, is no-data. 


[[File:Saveas.PNG]]


Your raster should align like this:


[[File:Seedpixel.PNG]]

=Superimpose Vector=

It happens regularly that, although following the previous steps meticulously, that the extent and pixel size of the rasterized vector file does not match the splitted band images 100 %. In this case we have to use the OTB  {{button|text=Superimpose Vector}} tool that you can find in the {{button|text=Processing Toolbox}}.

* Open the  {{button|text=Superimpose Vector}} tool in the {{button|text=Processing Toolbox}}.
* Select one of the split band images as {{button|text=Reference input}}
* Select the rasterized seed point shapefile as {{button|text=The image to reproject}}
* Fill in the other parameters as shown here:

[[File:Superimpose.PNG]]


Next, go to the {{button|text=Properties}} of the newly created layer in the {{button|text= Layers Panel}}

* Check the extent of the layer in the {{button|text=Metadata}} tab.
* Go to {{button|text=Styles}}, load max/min values and set the {{button|text=Color Gradient}} to {{button|text=White to black}}

The result should look very much the same as the first rasterized vector, but the extent and pixel size are now 100 % the same as in 

the split band images.

=Region Growing in SAGA=

Now we finally made it to the fun part of the whole exercise, which is creating the segments around the tops of the trees i.e. the seed points. Therefor we open SAGA GIS.

* When prompted, select an empty Startup Project

[[File:SAGA1.PNG|300px|]]


* Click the [[File:SAGA2.PNG]] button

* Change [[File:SAGA3.PNG|150px|]] to {{button|text=All files}}


* Open all split band images and the superimposed seed raster '''one after another'''
* The layer panel on the left should now look like this:


[[File:SAGA4.PNG]]


* In the line below {{button|text=Grids}} you see the extent of your raster files. If you happen to see two different of those lines, your raster images do not match 100 %. You can fix that again with the {{button|text=Superimpose sensor}} Tool in the OTB library in {{button|text=QGIS}}

=Seeded Region Growing Algorithm=


* Go to the {{button|text=Geoprocessing}} menu and select {{button|text=Find and Run tool}}

* Write ' region growing' and select [[File:SAGA6.PNG]]



* Select the one grid on offer
* Select your seed point raster as {{button|text=Seeds}}

* Select your splitted band images as {{button|text=Features}}
* Select {{button|text=create}} in {{button|text=Segments}}, {{button|text=Similarity}} and {{button|text=Tables&lt;&lt;Seeds}}
* Select {{button|text=Normalize}}
* Select {{button|text=4 (von Neumann}}
* Select {{button|text= feature space and position}}
* Insert the values shown in the picture and click {{button|text=Okay}} twice. We will talk about the input values later on...

[[File:Region_growing_mask.PNG]]


After a short while, SAGA will have finished the calculation and  you'll notice two new layers in the layer panel. Double-click on {{button|text= Segments}} and you hopefully see the first fruit of all the effort that you put into this exercise:


[[File:Region_growing_segments.PNG]]



=Vectorising the raster data=

In the next step we are going to create polygons in a shapefile which we can then open in {{button|text=QGIS}}...
* In the {{button|text=Find and Run tool}} search, enter 'vectorising grid' and select [[File:Vectorising_Saga.PNG|150px|]]


* Fill in the mask as follows and double-click {{button|text= Okay}} :

[[File:Vectorising_Saga2.PNG]]


* Right click on the newly created {{button|text=Segments}} file under {{button|text=Shapes}} in the Layer panel
* Select {{button|text=Save as}} and save the file as ESRI shapefile



=Opening the polygon file in QGIS=

Now you can open the shapefile in QGIS and adjust the look of the segments in the {{button|text= Styles}} tab in the properties of the layer:

[[File:Segments2.PNG]]


=The math part=

As you see in the last image, the segments fit the crowns quite well, but there is room for improvement. You can start to fine-tune the segmentation by changing the different parameters in the segmentation tool: '''Variance in Feature Space, Variance in Position Space, (Similarity Treshold), Method and Neighbourhood'''. You can also try to '''include or exclude certain data'''. 

You can do that by trial and error, however, it helps to understand the math that lies behind these algorithms.

What the algorithm in the Seeded Region Growing module does, is that it starts with the seed point, goes to the adjacent pixels and joins them together if they are similar enough. 'Enough' in this case is determined by the similarity-condition which is expressed in the following formula:

[[File:Formula_(1).PNG]]


The meaning of the parameters are easy to understand: ''''f'''' is a value (f.ex. a spectral value or a height value) ''''r'''' is a location. The indices designate which value it is, the one from the seed point or from the pixel that is being investigated (should it be joined to the segment of the seed or not).
''''t'''' indicates vector transposition and '''σ&lt;sub&gt;1&lt;/sub&gt;''' and '''σ&lt;sub&gt;2&lt;/sub&gt;''' are variances in colour and position space. The latter normalize the value differences between seed point and pixel.
'''α''' is the similarity criterion. It not need to be normalized and is set arbitrarily at 0.15.

'''σ&lt;sub&gt;1&lt;/sub&gt;''' and '''σ&lt;sub&gt;2&lt;/sub&gt;''' control how far away the values (feature space / position space) are allowed to be between each other. For a maximum distance of 40 % of the peak intensity and three color bands, the formula for '''σ&lt;sub&gt;1&lt;/sub&gt;''' is:

[[File:Formula_(2).PNG]]


'''σ&lt;sub&gt;2&lt;/sub&gt;''' determines the maximum spatial distance '''δ''' which is set to the radius of the largest crown (for spherical crowns):

[[File:Formula_(2).PNG]]



(BECHTEL 2008&lt;ref name=&quot;bechtel&quot;&gt;Bechtel, B. (2008).&quot;Segmentation for Object Extraction of Trees using MATLAB and SAGA.&quot; SAGA–Seconds Out, Hamburger Beiträge Zur Physischen Geographie Und Landschaftsökologie. Univ. Hamburg, Inst. für Geographie (2008): 1-12..&lt;/ref&gt;)



The algorithm in the Seeded Region Growing module is based on the work of ERIKSON (2003&lt;ref name=&quot;Erikson2003&quot;&gt;Erikson, M. (2003). Segmentation of individual tree crowns in colour aerial photographs using region growing supported by fuzzy rules. Canadian Journal of Forest Research, 33(8), 1557-1563.&lt;/ref&gt;, 2004&lt;ref name=&quot;Erikson2004&quot;&gt;Erikson, M. (2004). Segmentation and classification of individual tree crowns (Vol. 320).&lt;/ref&gt;, ERIKSON &amp; OLOFSSON 2005&lt;ref name=&quot;Eriolof2005&quot;&gt;Erikson, M., &amp; Olofsson, K. 2005). Comparison of three individual tree crown detection methods. Machine Vision and Applications, 16(4), 258-265.&lt;/ref&gt;)

=References=
&lt;references/&gt;

[[Category:QGIS Tutorial]]</rev>
        </revisions>
      </page>
    </pages>
  </query>
</api>