Amazon Biome

Developed by Imazon.

About

This repository contains the scripts to classify and filter the Amazon biome.

We highly recommend the reading of Amazon's Appendix of the Algorithm Theoretical Basis Document (ATBD). The fundamental information about the classification and methodology is there.

How to use

Create an account in Google Earth Engine plataform.
Install the python version 3.x.
Install the Earth Engine python API and get the credentials.
Download or clone this repository to your local workspace.

Example of the samples

// This script is executed only in the code editor

// asset containing sample points for training
var assetSamples = "projects/mapbiomas-workspace/AMOSTRAS/Amazonia/Colecao4/samples-collection-4-2019-5";

var samples = ee.FeatureCollection(assetSamples);

Map.addLayer(samples, {}, 'samples 2019', true);

Link to script

Classification

Set the configuration parameters and the paths on the script mapbiomas-classification-amazon.py
Define the number of samples to be used in each image classification.
Define the area estimated table.
Set the list of years, period to filter the image collection and satellite id.

params = [
    [2010, '2010-01-01', '2010-12-31', 'l5'],
    [2013, '2013-01-01', '2013-12-31', 'l8'],
    [2019, '2019-01-01', '2019-12-31', 'l7'],
]

Create a list with some landsat path/row ids.

pathRowsList = [
    "227/68", "225/69",
    "226/69", "224/69",
    "221/65", "222/66",
    "223/65", "223/66",
    "225/68", "226/68",
]

Define a list of imagens you don't want to use. Some Landsat imagery are faulty or have bad metadata.

trash = [
    "LT05_231068_19850823", "LT05_231068_19850908",
    "LT05_227067_19850827", "LT05_227067_19850912", 
    "LT05_224065_19850822", "LT05_224065_19851025",
    "LT05_227067_19870817", "LT05_226068_19870911",
    "LT05_226068_19870826", "LT05_226068_19870810",
    "LT05_232068_19870921", "LT05_232068_19870905",
    "LT05_232068_19870820", "LT05_232068_19870804",
    "LT05_230069_19880925", "LT05_230069_19880909",
    "LT05_230069_19880824", "LT05_229069_19880902",
    "LT05_227063_19880819", "LT05_227063_19880904",
    "LT05_227063_19881006", "LT05_229069_19890820",
    "LT05_229069_19890921", "LT05_229068_19890921",
    "LT05_231060_19901008", "LT05_233060_19910822",
    "LT05_229069_19910927", "LT05_229069_19910911",
    "LT05_221063_19911106", "LT05_221063_19911021",
    "LT05_221063_19910919", "LT05_221063_19910903",
    "LT05_223067_19940925", "LT05_223067_19940808",
    "LT05_223067_19941027", "LT05_223066_19940925",
    "LT05_229071_19951008", "LT05_229071_19950906",
    "LT05_228071_19951001", "LT05_228071_19950830",
    "LT05_228071_19950814", "LT05_230069_19970902",
]

Define the Randon Forest parameters

randomForestParams = {
    'numberOfTrees': 50,
    'variablesPerSplit': 4,
    'minLeafPopulation': 25
}

featureSpace = ["gv", "gvs", "soil", "npv", "shade", "ndfi", "csfi"]

There are some others parameters you can explore.
Run the code on the terminal.

$ python3 mapbiomas-classification-amazon.py

Reduce the collection of classifications

This methodology is processing and classifying every single image of Landsat collection with less than 50% of cloud cover for the Amazon biome region between 1985 and 2019. We are generating tons of data and we need to extract the best information possible for each pixel. For that reason, we reduce the classification collection by year using an adjustment rule.

For set up the script, define the parameters on the script mapbiomas-classification-mode.py.

assetClassification = "projects/imazon-simex/LULC/classification"
assetBiomes = "projects/mapbiomas-workspace/AUXILIAR/biomas-raster-41"
assetScenes = 'projects/mapbiomas-workspace/AUXILIAR/cenas-landsat'
assetOutput = 'projects/imazon-simex/LULC/integration-scenes'

version = '2'
outputVersion = '7'

years = [
    '2017',
    '2018',
    '2019',
]

pathRows = [
    "223/63", "223/64",
    "224/63", "224/64",
    "224/65", "227/65",
]

By the end, you can extend the trash variable with more Landsat images and run the code.

$ python3 mapbiomas-classification-mode.py

Spatial and temporal filter

In order to reduce the misclassified pixels, inconsistent transitions, or fill the gaps, we apply a spatial and temporal filter.

Create a table for temporal filter rules.

rulesTable = './csv/temporal-filter-rules.csv'

Define the spatial filter params

filterParams = [
    {
        'classValue': 3,
        'maxSize': 5
    },
    {
        'classValue': 4,
        'maxSize': 5
    },
    {
        'classValue': 12,
        'maxSize': 5
    },
    {
        'classValue': 15,
        'maxSize': 5
    },
    {
        'classValue': 19,
        'maxSize': 5
    },
    {
        'classValue': 25,
        'maxSize': 5
    },
    {
        'classValue': 27,
        'maxSize': 5
    },
    {
        'classValue': 33,
        'maxSize': 5
    },
]

Run the spatial and temporal script.

$ python3 mapbiomas-classification-filter.py

Kerenvascs/amazon