How to use

AcATaMa is divided and ordered by four tabs/sections:

  1. Thematic
  2. Sampling
  3. Classification
  4. Accuracy Assessment

At the bottom of the plugin there are 3 buttons:

1. Thematic map

The Thematic section you can set the thematic map (needed for some process, see more in use cases).

Optionally you can clip the thematic raster selected in an area of interest. After clipping the new thematic raster cut is automatically loaded.

Important!

Clip the thematic map in an area of interest (or load the thematic map clipped) can be (or not) very important for the sampling design and the accuracy assessment result, because the area by classes changes and some parts of AcATaMa depend on the area by classes.

Types of thematic rasters accepted in AcATaMa

It must be a categorical thematic layer with byte or integer as data type with a specific pixel-value/color associated. There are two types, respect to pixel-value/color associated, accepted in AcATaMa:

  1. Raster with pseudocolor style:

    You can use any raster (of byte or integer as data type) with specific style so that AcATaMa acquire the categorical information from the raster. Go to properties of the raster, then go to style, select singleband pseudocolor and generate the desired pixel-value/color associated (manually or generated automatically using the several options that Qgis have to do this) with only one requirement: the pixel-values associated must be integers.

    (Optional) After configure the style in Qgis for the raster is recommended save it in .qml Qgis style file, else Qgis save it in temporal file (or on the fly) and if you restart the Qgis and load the raster again you lost the pixel-value/color style associated. For save the style go to Style menu and click in Save as default Qgis save it in the same place and name of the raster with extension .qml.

    (Optional) Alternative (or additional) to the above, you can save all layers style and config saving it in a Qgis project.

  2. Raster with color table:

    You can use any raster (of byte or integer as data type) with pixel-values/color associated through a color table inside it as metadata. You can see it using gdalinfo or in style in layer properties this is shown as paletted.

Note

The thematic map is the raster layer to which the accuracy assessment will be applied (for example a land cover map) and also is the base to generate the random sampling.

2. Sampling desing

The sampling design defines how to select the sampled for the accuracy assessment (or any others uses). The Sampling section you can make and design the sampling using two categories for that: Simple Random Sampling and Stratified Random Sampling:

Simple Random Sampling

In the simple random sampling, every points (each x, y coordinates combination) has an equal chance of being selected. The size sample (number of points) that you define (Field: “Number of samples”), will be created pick randomly coordinates into the area of the thematic map, without taking into account the class or category to which it belongs. You can restrain the sites where the points will be crated with the follow options:

Stratified Random Sampling

In the stratified random sampling you divide the area of interest into smaller areas (strata), with a specific number of samples for each stratum. The strata need to be mutually exclusive and inclusive of the entire study area (FAO, 2016).

According to Olofsson et al. (2013, 2014), the stratification is recommended to improve the precision of the accuracy and area estimates. When strata are created for the objective of reporting accuracy by strata, the stratified design allows ensuring that a precise estimate is obtained for each stratum. In this way, a land change or other category that occupies a small proportion of the landscape can be identified and the sample size allocated to this stratum can be large enough to produce a small standard error for the user’s accuracy estimate.

In the basic case, each stratum could be a class or category of the thematic map; for this option you should select the thematic raster in the drop-down menu in “Categorical raster” block. If you want generate a stratified sample using other criteria (geographical sub-regions for example), you have to include an additional thematic raster layer with classes representing the strata and select it in the drop-down menu.

The stratified random may be:

If you want to define a specific sample size for one or more stratum, you can write it in the table and AcATaMa will modify the number of points in the others strata proportionally to the area, in order to keep the overall sample size; this allow perform the simplified approach of sample size allocation suggested by Olofsson et al.(2014), in which you define a specific sample size for the rare classes and the remain samples is allocated proportionally to the area of each other strata. If you want allocate a equal size sample for all strata, you can use this option to calculate the overall sample size, and assigning the number of points in each stratum in the Fixed values by category (manually) option.

You can “turn off” strata by deselecting in the last column of the table (“On”); these strata will not be taking into account to calculate the proportion of the area (Wi) or the sample size.

Sampling options

Optionally, in any type of sampling you can restrain the allocation of the points according to these criteria::

Other options

Both Simple Random Sampling and Stratified Random Sampling at bottom has the following options:

3. Classification

For the classification follow these steps:

Classification dialog

For the classification follow these steps (in classification dialog):

Nota: Samples order

For the classify the samples AcATaMa shuffles the list of samples for each file (the samples ID are not in order), this is to ensure randomization in the classification.

Configuration buttons

The configuration buttons dialog you can set all buttons for classify the samples, you can set the name, the color and (optionally) the thematic raster class.

Important!

The column Thematic raster class is available only if you set the thematic raster in Thematic tab before open the dialog. You must configure the thematic raster class for all buttons if you want accuracy assessment result.

4. Accuracy Assessment

Accuracy is defined as the degree to which the map produced agrees with the reference classification.

AcATaMa calculated the accuracy assessment using the formulas included in Olofsson et al. (2013, 2014), the results are presented in five tables:

  1. Error Matrix: The error matrix is a cross-tabulation of the class labels allocated by the classification of the remotely sensed data against the reference data (thematic map) for the sample sites. The main diagonal of the error matrix highlights correct classifications while the off-diagonal elements show omission and commission errors. The values of the error matrix are fundamental to both accuracy assessment and area estimation Olofsson et al. 2014). The column Wi is the area proportion of each stratum in the map.

  2. Error matrix of estimated area proportion: The absolute counts of the sample are converted into estimated area proportions using the equation (9) in Olofsson et al. (2014) for simple random, systematic or stratified random sampling with the map classes defined as the strata.

  3. Quadratic error matrix for estimated area proportion: Correspond to the standard error estimated by the equation (10) in Olofsson et al. (2014)

  4. Accuracy Matrices:

    • User´s accuracy matrix of estimated area proportion: User´s accuracy is the proportion of the area mapped as a particular category that is actually that category “on the ground” where the reference classification is the best assessment of ground condition. User’s accuracy is the complement of the probability of commission error (Olofsson et al. 2013). The user´s accuracy is calculated by the equation (2) in Olofsson et al. (2014). In the report, the user´s accuracy for each class or category correspond to the diagonal of the matrix, that means, the fields in which the class of the thematic raster map and the classified category (reference) are equals.

    • Producer´s accuracy matrix of estimated area proportion: Producer’s accuracy is the proportion of the area that is a particular category on the ground that is also mapped as that category. Producer’s accuracy is the complement of the probability of omission error (Olofsson et al. 2013). The producer’s accuracy is calculated by the equation (3) in Olofsson et al. (2014). In the report, the accuracy for each class or category correspond to the diagonal of the matrix, the fields in which the class of the thematic raster map and the classified category (reference) are equals.

    • Overall accuracy: Is the proportion of the area mapped correctly. It provides the user of the map with the probability that a randomly selected location on the map is correctly classified (Olofsson et al. 2013). It is important use carefully this value because the overall map accuracy is not always representative of the the accuracy of the individual classes or strata (FAO, 2016). The overall map accuracy is calculated by the equation (1) in Olofsson et al. (2014)

  5. Classes area adjusted table: The accuracy assessment serves to derive the uncertainty of the map area estimates. Whereas the map provides a single area estimate for each class without confidence interval, the accuracy estimates adjusts this estimate and also provides confidence intervals as estimates of uncertainty . The adjusted area estimates can be considerably higher or lower than the map estimates (FAO, 2016).

    The estimated area for each class or stratum and the standard error of the estimated area is given by the equation (11) in Olofsson et al. (2014); they allow to obtain the confidence interval with the percent defined by the z-score value. By default AcATaMa calculate a 95% confidence interval (Z=1,96), but you can modify the z- score value according to the desired percent (Settings options in the report of results).

Tips

1. Save and restore classification status

In some case when you have several samples to classified, you want save all status and configuration of AcATaMa and close Qgis, and after you want to load it again in new Qgis instance. For that we recommend to do:

References