In MartinezLaura/CanHeMonR.MaxEnt: Application of MaxEnt for canopy health monitoring through airbone image analysis.

This is an R Markdown Notebook. When you execute code within the notebook, the results appear beneath the code.

Try executing this chunk by clicking the Run button within the chunk or by placing your cursor inside it and pressing Ctrl+Shift+Enter.

print("Hello World!")

• Packages required:
• raster
• doParallel
• foreach
• rJava
• dismo, with maxent >= 3.4
• maptools
• rgeos
• DiagrammeR
• sp

For maxent installation please go to: https://biodiversityinformatics.amnh.org/open_source/maxent/

Please run the following code to check the avaiable packages and install if necessary

RequiredPackages <- c("rgeos", "raster","doParallel", "foreach", "rJava", "dismo", "maptools", "DiagrammeR", "sp", "sf", "rmarkdown", "CanHeMonR.MaxEnt", "dplyr", "ggplot2") 
for (i in RequiredPackages) { #Installs packages if not yet installed
    if (!require(i, character.only = TRUE)) install.packages(i)
}
dismo::maxent()

Algorithm flowchart

DiagrammeR::DiagrammeR("
graph TB
    A>Point shapefile with precences] --> D[Sample presences and backgroud points]
    B>Raster files and textures to run] --> D
    D --> E>Serialized object with the presence and background points]
    E --> F[Calibration and creation of MaxEnt model]
    F --> G>Serialized object with the model]
    G --> H[Run model over the tiles]
    B --> H
    H --> I>Tif files with the MaxEnt predictions]
    I --> J[Post-process of gemoetry operations]
    J --> K>Point shapefiles]
    K --> L[Visual inspection od the output model]
    L --> D
")

Sampling presence and pseudo-absent data for maxent

Function that creates a sample dataframe for the selected classes with the presences and background points for the selected rasters. This sample will be used as training data for image classification through MaxEnt

Inputs

• Points shapefile with the presence data.
• Path to the rasters and the associated textures of the area that we want to predict. It is important to add all the areas that we want to predict and not only a sample of it in order to help the model describe the variation of the data.
• A vector with the classes that we want to model. this vector has to have the same fields as the ones described on the shapefile
• An integer describing the number of background points that we want to sample for each tile

More information and examples of the function in the code

Output

• Serialized object (rdsdata file) with the data frame with the presences and absences points of the selected classes

Function

You can set the parameters and run it in the R console with:

render("sample_points.r", params = list(r_train_dir= "" ,
                                        text_train_dir= "",
                                        tile= "",
                                        text= "",
                                        prob_tifs= "",
                                        vuln_classes= "",
                                        training_pol_filename= "", 
                                        field_name= "",
                                        nimputs_tile="",
                                        data_outp_dir="",
                                        abs_samp= "",
                                        parallel= "",
                                        nWorkers= "",
                                        data_outp_name= "",
                                        randompt= "", 
                                        EOS= ""))

Calibrate and creation of the MaxEnt model

Calibration and creation of the model for each class selected in the previous step, using a raster brick as predictors. If desired, it can output also the stadisctics of the model to measure the preformace.

Inputs

• A vector with the classes that we want to model. this vector has to have the same fields as the ones described in the previus function
• Serialized object (rdsdata file) with the data frame with the presences and absences points of the selected classes

More information and examples of the function in the code

Output

• Serialized object (rdsdata file) with class-specific distribution MaxEnt models

Function

You can set the parameters and run it in the R console with:

render("calibrate_model.r", params = list(vuln_classes = "", 
                                          training_path = "", 
                                          model_outp_dir = "", 
                                          name = "",
                                          stadistics = "", 
                                          myargs = "", 
                                          model_type = "",
                                          stadisticspath = ""))

Run model

Run a saved MaxEnt model on the image data you want to predict.

Inputs

• Path to the rasters and the associated textures of the area that we want to predict.
• Serialized object (rdsdata file) with the MaxEnt model you want to use

More information and examples of the function in the code

Output

• A raster image for each tile with the probabilities or cummulative probabilities of presence for each class

Function

You can set the parameters and run it in the R console with:

render("run_model.r", params = list(predictor_dir = "",
                                         text_train_dir = "",
                                         MaxEntmodel_dir = "",
                                         fname_MaxEntmodel_r = "",
                                         output_dir = "",
                                         rastername = "",
                                         model_type = "",
                                         EOS = ""))

Creation of the thresholds for rasters

Creation of several serialized objects with different thresholds in order to select the one that maximized the retrival of Pine trees with the less false positive possible.

Inputs

• Path to the outputted rasters from the previous function.
• A Point shapefile with the detected points
• The minimum and maximim threshold that you want to study. Remeber that MaxEnt outputs a raster file with probablities between 0 and 1.
• Number of 'absence' points to be randomly selected from each tile.

More information and examples of the function in the code

Output

• A set of serialized objects for each the threshold.

Function

You can set the parameters and run it in the R console with:

render("sick_tree_errors.r", params = list(r_pred_dir  = "", 
                                           tile  = "", 
                                           prefix = "", 
                                           vuln_classes = "", 
                                           training_pnt_filename = "", 
                                           radius = "", 
                                           field_name = "", 
                                           abs_samp  = "",
                                           minthresh = "", 
                                           maxthresh = "", 
                                           stepthresh = "",
                                           parallel = "", 
                                           nWorkers = "", 
                                           data_outp_dir = ""))

Selection of the best threshold for rasters and poligonization

Selection of the best cut-off probability thresholds that selects the one that maximized the retrival of Pine trees with the less false positive possible. Selection of the smallest number of pixels detected inside a true tree-disk. Poligonization of the raster files with the selected parameters.

Inputs

• Path to the serialized objects for each the threshold.
• Path to the outputted rasters outputted by MaxEnt.
• A Point shapefile with the detected points
• The minimum and maximim threshold that you want to stdy. Remeber that MaxEnt outputs a raster file with probablities between 0 and 1.
• Number of 'absence' points to be randomly selected from each tile.

More information and examples of the function in the code

Output

• A polygon shapefile for each raster.

Function

You can set the parameters and run it in the R console with:

render("MaxCutoff", params = list(input_dir  = "", 
                                  plots  = "", 
                                  stadisticspath  = "", 
                                  prefix  = "", 
                                  max_error  = "", 
                                  cla  = "", 
                                  r_pred_dir  = "", 
                                  tile  = "", 
                                  max_area  = "", 
                                  res  = "", 
                                  outp_dir  = ""))

MartinezLaura/CanHeMonR.MaxEnt documentation built on May 17, 2019, 6:21 p.m.