ENM resolution mix - multiple runs version

Created: 2014-11-13 15:20:55 Last updated: 2015-05-12 05:28:09

Download Workflow

This workflow is used to compare the results of two kinds of models: one using only low resolution layers and the other using a random mix of low and high resolution layers. Users select the study region and the environmental variables considered to be the main drivers of a virtual species niche. The workflow is all based on the ENM components, which use the openModeller Web Service (OMWS). After getting initial parameters from the user, the workflow generates a random virtual niche. A random threshold within a given prevalence range is then used to separate suitable from unsuitable areas, from where random training and testing points are sampled. Training points are used to generate the models and independent testing points are used for external tests measuring the AUC. All steps (except mask and environmental layer selection) are repeated a given number of times to check if the model using mixed resolution layers performs significantly better than the one with only low resolution layers.

Preview

Download as scalable diagram (SVG)

Run

Run this Workflow in the Taverna Workbench...

Option 1:

Copy and paste this link into File > 'Open workflow location...'
http://www.myexperiment.org/workflows/4535/download?version=3
[ More Info Expand ]

Workflow Components

Authors (1)

Titles (1)

Descriptions (1)

Dependencies (0)

Inputs (7)

Name	Description
min_virtual_niche_prevalence	Minimum virtual niche prevalence (proportion of suitable area in the native range).
max_virtual_niche_prevalence	Maximum virtual niche prevalence (proportion of suitable area in the native range).
num_background_points	Number of background points
num_testing_points	Number of points to be used for testing (an equal number will be used for presences and absences).
min_number_training_points	Minimum number of training points to be used when creating models for the virtual species. The real number will be a random value between minimum and maximum.
max_number_training_points	Maximum number of training points to be used when creating models for the virtual species. The real number will be a random value between minimum and maximum.
replicates	Number of times to repeat the experiment: generate virtual niche, generate models with different resolutions and test them.

Processors (8)

Name	Type	Description
choose_layers_and_mask	workflow
create_and_test_models	workflow
for_loop_triggering	beanshell	Script import java.util.ArrayList; ArrayList trigger_list = new ArrayList(); int end_value = Integer.parseInt(replicates); for (int i = 1; i <= end_value; i += 1) { trigger_list.add(i); }
merge_all_layers	beanshell	Script all_layers = high_resolution_layers + "\n" + low_resolution_layers;
sample_background_points	component	Randomly sample background points on the specified mask using the intersection between all environmental layers (high and low resolution). Points are returned in the absences port. Component sample_points
no	stringconstant	Value no
1	stringconstant	Value 1
demo	stringconstant	Value demo

Beanshells (7)

Name	Inputs	Outputs
for_loop_triggering	replicates	trigger_list
merge_all_layers	high_resolution_layers low_resolution_layers	all_layers
mix_layers	low_resolution_layers high_resolution_layers	mixed_resolution_layers
suggest_random_suitable_threshold	model_values min_prevalence max_prevalence	threshold proportion
define_random_number_of_training_points	min_points max_points	num_points
get_last_iteration_result	random_proportion_list point_list threshold_list model_xml_list model_values_list	random_proportion point threshold model_xml model_values
empty_list		empty_list

Outputs (12)

Name	Description
mask_id
layer_labels
origin_point
model_xml
suitability_threshold
low_resolution_auc
mixed_resolution_auc
num_training_points
training_points_csv
testing_presence_points_csv
testing_absence_points_csv
prevalence

Datalinks (33)

Source	Sink
choose_layers_and_mask:low_resolution_layers_ids	create_and_test_models:low_resolution_layers
choose_layers_and_mask:high_resolution_layers_ids	create_and_test_models:high_resolution_layers
choose_layers_and_mask:mask_id	create_and_test_models:mask_id
min_number_training_points	create_and_test_models:min_number_training_points
max_number_training_points	create_and_test_models:max_number_training_points
num_testing_points	create_and_test_models:num_testing_points
max_virtual_niche_prevalence	create_and_test_models:max_virtual_niche_prevalence
min_virtual_niche_prevalence	create_and_test_models:min_virtual_niche_prevalence
for_loop_triggering:trigger_list	create_and_test_models:loop_iterator
sample_background_points:absence_points_xml	create_and_test_models:background_points
merge_all_layers:all_layers	create_and_test_models:all_layers
replicates	for_loop_triggering:replicates
choose_layers_and_mask:high_resolution_layers_ids	merge_all_layers:high_resolution_layers
choose_layers_and_mask:low_resolution_layers_ids	merge_all_layers:low_resolution_layers
merge_all_layers:all_layers	sample_background_points:layers
num_background_points	sample_background_points:num_points
choose_layers_and_mask:mask_id	sample_background_points:mask_id
no:value	sample_background_points:environmentally_unique
1:value	sample_background_points:proportion_of_absences
no:value	sample_background_points:spatially_unique
demo:value	sample_background_points:species_label
choose_layers_and_mask:mask_id	mask_id
choose_layers_and_mask:high_resolution_layers_labels	layer_labels
create_and_test_models:origin_point	origin_point
create_and_test_models:model_xml	model_xml
create_and_test_models:suitability_threshold	suitability_threshold
create_and_test_models:low_resolution_auc	low_resolution_auc
create_and_test_models:mixed_resolution_auc	mixed_resolution_auc
create_and_test_models:num_training_points	num_training_points
create_and_test_models:training_points_csv	training_points_csv
create_and_test_models:testing_presence_points_csv	testing_presence_points_csv
create_and_test_models:testing_absence_points_csv	testing_absence_points_csv
create_and_test_models:prevalence	prevalence

Coordinations (3)

Controller	Target
merge_all_layers	sample_background_points
choose_layers_and_mask	merge_all_layers
sample_background_points	create_and_test_models

Information Workflow Type

Taverna 2

Information Uploader

Renato De Giovanni

Information License

All versions of this Workflow are licensed under:

Information Version 3 (of 10)

Information Credits (2)

(People/Groups)

Information Attributions (0)

(Workflows/Files)

None

Information Tags (0)

None

Log in to add Tags

Information Shared with Groups (1)

BioVeL internal

Information Featured In Packs (0)

None

Log in to add to one of your Packs

Information Attributed By (0)

(Workflows/Files)

None

Information Favourited By (0)

No one

Information Statistics

5452 viewings

1427 downloads

[ see breakdown ]

Citations (0)

None

Version History

In chronological order:

Compare ENM models created with layers with different spatial resolutions.

Created by Renato De Giovanni on Thursday 13 November 2014 15:20:51 (UTC)
Compare ecological niche models created with layers with different spatial resolutions.

Created by Renato De Giovanni on Friday 14 November 2014 08:46:24 (UTC)

Revision comment:

Removed virtual niche test.
Compare ecological niche models created with layers with different spatial resolutions.

Created by Renato De Giovanni on Friday 14 November 2014 10:18:20 (UTC)

Revision comment:

Increased default number of replicates and removed redundant parameter in nested workflow.
ENM resolution mix - multiple runs version

Created by Renato De Giovanni on Tuesday 18 November 2014 17:16:00 (UTC)

Revision comment:

Updated versions of some components and improved workflow description.
ENM resolution mix - multiple runs version

Created by Renato De Giovanni on Wednesday 19 November 2014 11:40:23 (UTC)

Revision comment:

Upgraded some of the components.
ENM resolution mix - multiple runs version

Created by Renato De Giovanni on Thursday 04 December 2014 16:16:21 (UTC)

Revision comment:

Updated components versions.
ENM resolution mix - multiple runs version

Created by Renato De Giovanni on Friday 05 December 2014 18:43:04 (UTC)

Revision comment:

Initial R code to calculate probability of getting a better AUC when mixing resolutions and to plot a graph comparing the results. Also fixed a bug in threshold calculation and increased the default number of replicates.
ENM resolution mix - multiple runs version

Created by Renato De Giovanni on Tuesday 09 December 2014 20:34:17 (UTC)

Revision comment:

Updated samplePoints component, removed unnecessary loop feedback and fixed R code.
ENM resolution mix - multiple runs version

Created by Renato De Giovanni on Saturday 04 April 2015 21:27:30 (UTC)

Revision comment:

Upgraded interaction components to use github pages.
ENM resolution mix - multiple runs version

Created by Renato De Giovanni on Tuesday 12 May 2015 05:28:09 (UTC)

Revision comment:

This version plots a different graph to compare results and calculates mean and standard deviation of the AUCs differences.