Tree Cover Loss Analysis

This project performs a polygonal summary on tree cover loss and intersecting layers for a given input feature using SPARK and Geotrellis.

Analysis

Currently the following analysis are implemented

• Tree Cover Loss (for ArcPy client)
• Annual Update minimal
• Carbon Flux Full Standard Model
• Carbon Flux Sensitivity Analysis
• Glad Alerts
• Viirs/ MODIS Fire Alerts
• Forest Change Diagnostic
• GFW Pro Dashboard
• Integrated Alerts

Tree Cover Loss

A simple analysis which only looks at tree cover loss, tree cover density (2000 or 2010), aboveground biomass, gross GHG emissions, gross carbon removals, and net GHG flux. Aboveground carbon, belowground carbon, and soil carbon stocks in 2000 and annual emissions from aboveground biomass loss (old emissions model) can be optionally output. Users can select one or many tree cover thresholds. Output will be a flat file, with one row per input feature and tree cover density threshold. Optional contextual analysis layers include plantations, humid tropical primary forests, global peat, drivers of tree cover loss, and tree cover loss due to fires. Contextual layers are lazily analyzed, meaning that they are not loaded into the analysis if the user does not request them (to save memory). The emissions outputs (annual and total) are from the forest carbon flux model (Harris et al. 2021 Nature Climate Change) forest carbon flux model. Outputs also include carbon removals and carbon net flux (total only) (Harris et al. 2021). Emissions, removals, and net flux are reported for (> tree cover density 2000 threshold OR Hansen gain OR mangrove extent NOT pre-2000 plantations) because the model results include not just pixels above a certain tree cover density threshold, but also Hansen gain pixels. Other outputs from this analysis (loss, gain, biomass, carbon pools, etc.) use the simple tree cover density threshold.

This type of analysis only supports simple features as input. Best used together with the ArcPY Client.

test:runMain org.globalforestwatch.summarystats.SummaryMain treecoverloss --feature_type feature --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix --tcd 2000 --threshold 0 --threshold 30 --contextual_layer is__gfw_plantations --contextual_layer tsc_drivers__class --carbon_pools

Annual Update minimal

An analysis intersecting Tree Cover Loss data with a number of layers. It is used to compute statistics for the GFW country and user dashboards, including carbon flux outputs (emissions, removals, net flux).

Supported input features are

• GADM
• Geostore
• WDPA
• Simple Feature

Output are Whitelist, Summary and Change tables for input features. For GADM, there will be seperate sets for ISO, ADM1 and ADM2 areas. For GADM there will also be summary tables with one row per ISO, ADM1, ADM2 and Tree Cover Density, which is used for to prepare the Download Spreatsheets on the GFW country pages. To produce final spreadsheets you will need to add another post processing step.

test:runMain org.globalforestwatch.summarystats.SummaryMain annualupdate_minimal --feature_type gadm --tcl --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix
test:runMain org.globalforestwatch.summarystats.SummaryMain annualupdate_minimal --feature_type wdpa --tcl --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix
test:runMain org.globalforestwatch.summarystats.SummaryMain annualupdate_minimal --feature_type geostore --tcl --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix
test:runMain org.globalforestwatch.summarystats.SummaryMain annualupdate_minimal --feature_type feature --tcl --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix

Carbon Flux Full Standard Model

Carbon Flux (full standard model) analysis is used to produce detailed statistics for carbon flux research, with additional model-specific contextual layers. It uses the same approach as the annual update analysis, but with all stock and flux outputs from the forest carbon flux model. It also analyzes several contextual layers that are unique to the carbon flux model and does not include many contextual layers used in the annual update minimal analysis. It currently only works with GADM features.

test:runMain org.globalforestwatch.summarystats.SummaryMain carbonflux --feature_type gadm --tcl --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix

Carbon Flux Sensitivity Analysis

Carbon Flux analysis is used to produce statistics from sensitivity analysis runs of the forest carbon flux model. It uses the same approach as the Carbon Flux Full Standard Model, but with fewer model outputs analyzed. The sensitivity analysis argument needs to match the folder names in the input tile folders (e.g., maxgain, biomass_swap). To run this model with the standard flux model output for an analysis using fewer flux model outputs, use standard as the sensitivity_type argument. It currently only works with GADM features.

test:runMain org.globalforestwatch.summarystats.SummaryMain carbon_sensitivity --feature_type gadm --tcl --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix --sensitivity_type sensitivity_analysis_type
test:runMain org.globalforestwatch.summarystats.SummaryMain carbon_sensitivity --feature_type gadm --tcl --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix --sensitivity_type standard

Glad Alerts

Glad alert analysis computes whitelist, summary, daily and weekly change data for given input features and intersects areas with the same contextual layers as in annual update minimal. It is used to update the country and user dashboards for the GFW website.

Users can select, if they want to run the full analysis, or only look at change data. Computing only change data makes sense, if neither input features, nor contextual layers have changed, but only glad alerts. In that case, only the daily and weekly change tables will be updated.

Supported input features are

• GADM
• Geostore
• WDPA
• Simple Feature

test:runMain org.globalforestwatch.summarystats.SummaryMain gladalerts --feature_type gadm --glad --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix [--change_only]
test:runMain org.globalforestwatch.summarystats.SummaryMain gladalerts --feature_type wdpa --glad --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix [--change_only]
test:runMain org.globalforestwatch.summarystats.SummaryMain gladalerts --feature_type geostore --glad --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix [--change_only]
test:runMain org.globalforestwatch.summarystats.SummaryMain gladalerts --feature_type feature --glad --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix [--change_only]

Viirs/ MODIS Fire Alerts

Viirs/ MODIS Fire alert analysis computes whitelist, daily and weekly change data for given input features and intersects areas with the same contextual layers as in glad alerts. It also enriches all fire data with intersecting contextual information. It is used to update the country and user dashboards for the GFW website.

Supported input features are

• GADM
• Geostore
• WDPA
• Simple Feature

test:runMain org.globalforestwatch.summarystats.SummaryMain firealerts --fire_alert_type MODIS/VIIRS --fire_alert_source s3://bucket/prefix/file.tsv --feature_type gadm --glad --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix [--change_only]
test:runMain org.globalforestwatch.summarystats.SummaryMain firealerts --fire_alert_type MODIS/VIIRS --fire_alert_source s3://bucket/prefix/file.tsv --feature_type wdpa --glad --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix [--change_only]
test:runMain org.globalforestwatch.summarystats.SummaryMain firealerts --fire_alert_type MODIS/VIIRS --fire_alert_source s3://bucket/prefix/file.tsv --feature_type geostore --glad --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix [--change_only]
test:runMain org.globalforestwatch.summarystats.SummaryMain firealerts --fire_alert_type MODIS/VIIRS --fire_alert_source s3://bucket/prefix/file.tsv --feature_type feature --glad --features s3://bucket/prefix/file.tsv --output s3://bucket/prefix [--change_only]

Forest Change Diagnostic

Forest Change Diagnostic computes forest loss and fire alerts for an input set of lists of geometries. It is used to compute statistics for lists of a GFW Pro customer. It requires the firealerts options and optionally takes any feature options. The only supported input feature is gfwpro. Automatically turns on the --split_features option.

test:runMain org.globalforestwatch.summarystats.SummaryMain forest_change_diagnostic --feature_type gfwpro --features s3://bucket/prefix/file.tsv --fire_alert_source s3://bucket/prefix/file.tsv --output s3://bucket/prefix

There are two test input files available in the source tree, so you can run a sample forest change diagnostic locally without s3 via a command like:

test:runMain org.globalforestwatch.summarystats.SummaryMain forest_change_diagnostic --split_features --feature_type gfwpro --features src/test/resources/palm-oil-32.tsv --fire_alert_type modis --fire_alert_source src/test/resources/sample-fire-alert.tsv --output testout

GFW Pro Dashboard

GFW Pro Dashboard computes summary statistics for the GFW Pro Dashboard. It optionally takes firealert options and feature options.

Supported input features are:

• GADM
• Geostore
• WDPA
• Simple Feature
• GFW Pro

Integrated Alerts

Integrated Alerts computes a combination of deforestation alerts based on GLAD-L, GLAD-S2, and RADD systems. It is used to compute these alerts for GFW. It optionally takes any feature, gadm, or wdpa options.

Supported input features are

• GADM
• Geostore
• WDPA
• Simple Feature
• GFW Pro

Inputs

Use Polygon Features encoded in TSV format. Geometries must be encoded in WKB. You can specify one or many input files using wildcards:

ex:

• s3://bucket/prefix/gadm36_1_1.tsv
• s3://bucket/prefix/geostore_*.tsv

Make sure features are sufficiently small to assure a well balanced partition size and workload. Larger features should be split into smaller features, prior to running the analysis. Also make sure, that features do not overlap with tile boundaries (we use 10x10 degree tiles). For best performance, intersect input features with a 1x1 degree grid. If you are not sure how to best approach this, simply use the ArcPY Client Alternatively, use the --split_features opton.

Options

The following options are supported:

Option	Type	Analysis or Feature Type	Description
features	string	all (required)	URI of features in TSV format
output	string	all (required)	URI of output dir for CSV files
feature_type	string	all (required)	Feature type: one of 'gadm', 'wdpa', 'geostore', 'feature', or 'gfwpro'
limit	int	all	Limit number of records processed
iso_first	string	gadm or wdpa features	Filter by first letter of ISO code
iso_start	string	gadm or wdpa features	Filter by ISO code larger than or equal to given value
iso_end	string	gadm or wdpa features	Filter by ISO code smaller than given value
iso	string	gadm or wdpa features	Filter by country ISO code
admin1	string	gadm features	Filter by country Admin1 code
admin2	string	gadm features	Filter by country Admin2 code
id_start	int	feature analysis	Filter by IDs greater than or equal to given value
id_end	int	feature analysis	Filter by IDs less than or equal to given value
wdpa_status	string	wdpa features	Filter by WDPA Status
iucn_cat	string	wdpa features	Filter by IUCS Category
tcd	int	treecoverloss analysis	Select tree cover density year
threshold	int	treecoverloss analysis	Treecover threshold to apply (multiple)
contextual_layer	string	treecoverloss analysis	Include (multiple) selected contextual layers: is__umd_regional_primary_forest_2001, is__gfw_plantations, is__global_peat, tsc_drivers__class, is__tree_cover_loss_fire
carbon_pools	flag	treecoverloss analysis	Optionally calculate stock sums for multiple carbon pools in 2000 (aboveground, belowground, soil)
tcl	flag	all	Filter input feature by TCL tile extent, requires boolean tcl field in input feature class
glad	flag	all	Filter input feature by GLAD tile extent, requires boolean glad field in input feature class
change_only	flag	all except treecover	Process change only
sensitivity_type	string	carbon_sensitivity	Select carbon sensitivity model
fire_alert_type	string	firealerts	Select Fire alert type
fire_alert_source	string	firealerts	URI of fire alert TSV file
overwrite	flag	all	Overwrite output location if already existing
split_features	flag	all	Split input features along 1x1 degree grid
no_output_path_suffix	flag	all	Do not autogenerate output path suffix at runtime
intermediate_list_source	flag	forest_change_diagnostic analysis	URI of intermediate list results in TSV format
contextual_feature_type	flag	gfwpro_dashboard analysis	type of contextual feature
contextual_feature_url	flag	gfwpro_dashboard analysis	URI of contextual feature in TSV format

Inventory

• build.sbt: Scala Build Tool build configuration file
• .sbtopts: Command line options for SBT, including JVM parameters
• project: Additional configuration for SBT project, plugins, utility, versions
• src/main/scala: Application and utility code
• src/test/scala: Unit test files

Spark Job Commands

Local

sbt:geotrellis-wri> test:runMain org.globalforestwatch.summarystats.SummaryMain annualupdate_minimal --features file:/Users/input/ten-by-ten-gadm36/wdpa__10N_010E.tsv --output file:/User/out/summary

EMR

Before running review sbtlighter configuration in build.sbt, reload SBT session if modified.

sbt:geotrellis-wri> sparkCreateCluster
sbt:treecoverloss> sparkSubmitMain org.globalforestwatch.summarystats.SummaryMain annualupdate_minimal --features s3://gfw-files/2018_update/tsv/gadm36_1_1.csv --output s3://gfw-files/2018_update/results/summary --feature_type gadm --tcl
sbt:treecoverloss> sparkSubmitMain org.globalforestwatch.summarystats.SummaryMain gladalerts --features s3://gfw-files/2018_update/tsv/wdpa__*.tsv --output s3://gfw-files/2018_update/results/summary  --feature_type wdpa --tcl --iso BRA