def analyze(threshold, geojson):
"""For a given Hansen threshold mask on WHRC biomass data
and geometry return a dictionary of total t/ha.
"""
# logging.info('[Soil Carbon Service]: In Soil carbon service')
try:
d = {}
hansen_asset = SETTINGS.get('gee').get('assets').get('hansen')
soil_carbon_asset = SETTINGS.get('gee').get('assets').get(
'soils_30m')
region = get_region(geojson)
reduce_args = {
'reducer': ee.Reducer.sum().unweighted(),
'geometry': region,
'bestEffort': True,
'scale': 30
}
tc_mask = ee.Image(hansen_asset).select('tree_' +
str(threshold)).gt(0)
sc = ee.Image(soil_carbon_asset).multiply(
ee.Image.pixelArea().divide(10000)).mask(tc_mask)
# Identify soil carbon value
sc_value = sc.reduceRegion(**reduce_args).getInfo()
d['total_soil_carbon'] = sc_value
# logging.info(f'[Soil Carbon Service]:d = {d}')
return d
except Exception as error:
logging.error(str(error))
raise soilCarbonError(message='Error in soil carbon analysis')
def _gee_biomass(geom, thresh):
image1 = get_thresh_image(
str(thresh),
SETTINGS.get('gee').get('assets').get('biomassloss_v1').get(
'hansen_loss_thresh'))
image2 = ee.Image(
SETTINGS.get('gee').get('assets').get('biomassloss_v1').get(
'biomass_2000'))
region = get_region(geom)
# Reducer arguments
reduce_args = {
'reducer': ee.Reducer.sum(),
'geometry': region,
'bestEffort': True,
'scale': 90
}
# Calculate stats 10000 ha, 10^6 to transform from Mg (10^6g) to Tg(10^12g) and 255 as is the pixel value when true.
area_stats = image2.multiply(image1) \
.divide(10000 * 255.0) \
.multiply(ee.Image.pixelArea()) \
.reduceRegion(**reduce_args)
carbon_stats = image2.multiply(
ee.Image.pixelArea().divide(10000)).reduceRegion(**reduce_args)
area_results = area_stats.combine(carbon_stats).getInfo()
return area_results
def analyze(threshold, geojson):
"""For a given Hansen threshold mask on WHRC biomass data
and geometry return a dictionary of total t/ha.
"""
try:
d = {}
hansen_asset = SETTINGS.get('gee').get('assets').get('hansen')
biomass_asset = SETTINGS.get('gee').get('assets').get('whrc_biomass')
region = get_region(geojson)
reduce_args = {'reducer': ee.Reducer.sum().unweighted(),
'geometry': region,
'bestEffort': True,
'scale': 30}
tc_mask = ee.Image(hansen_asset).select(['tree_' + str(threshold)], ['tree_cover']).gt(0)
biomass = ee.ImageCollection(biomass_asset).max().multiply(ee.Image.pixelArea().divide(10000)).mask(tc_mask)
tree_cover = tc_mask.multiply(ee.Image.pixelArea().divide(10000)).mask(tc_mask)
# Identify thresholded biomass value
biomass_value = biomass.reduceRegion(**reduce_args).getInfo()
tree_cover_value = tree_cover.reduceRegion(**reduce_args).getInfo()
d['biomass'] = biomass_value.get('b1', 0)
d['tree_cover'] = tree_cover_value.get('tree_cover', 0)
return d
except Exception as error:
logging.error(str(error))
raise WHRCBiomassError(message='Error in WHRC Biomass Analysis')
def analyze(geojson):
"""For a given geometry and mangrove biomass data
return a dictionary of total t/ha.
"""
try:
d = {}
biomass_asset = SETTINGS.get('gee').get('assets').get(
'mangrove_biomass')
region = get_region(geojson)
reduce_args = {
'reducer': ee.Reducer.sum().unweighted(),
'geometry': region,
'bestEffort': True,
'scale': 30
}
biomass = ee.Image(biomass_asset).multiply(
ee.Image.pixelArea().divide(10000))
# Identify thresholded biomass value
biomass_value = biomass.reduceRegion(**reduce_args).getInfo()
d['biomass'] = biomass_value
return d
except Exception as error:
logging.error(str(error))
raise MangroveBiomassError(
message='Error in Mangrove Biomass Analysis')
def analyze(geojson):
"""For a given geometry and population density data
return a dictionary of total people in a region.
"""
try:
d = {}
# The number of people per cell
population_asset = SETTINGS.get('gee').get('assets').get(
'population')
region = get_region(geojson)
reduce_args = {
'reducer': ee.Reducer.sum().unweighted(),
'geometry': region,
'bestEffort': True,
'scale': 30
}
# Convert m2 to ha
scale_factor = ee.Number(1e4)
# The number of persons per cell
population = ee.Image(population_asset)
# .multiply(ee.Image.pixelArea().divide(scale_factor))
# Total population value within region
population_value = population.reduceRegion(**reduce_args).getInfo()
d['population'] = population_value
return d
except Exception as error:
logging.error(str(error))
raise PopulationError(message='Error in Population Analysis')
def analyze(geojson, start_date, end_date):
"""Forma250 microservice class. This service uses the latest image in
'projects/wri-datalab/FormaGlobalGFW' image collection. The bands of
that image contain 'alert_delta': the percent of clearing per pixel that occured
within the last 3 months, 'alert_near_term_delta': the percent of clearing
which occured within the last 1 month, 'alert_date': the first date when
the delta passed a threshold of an ecogroup. 'alert_clearing': the %
clearing in pixel over the past year, and 'alert_accuracy': the error of
clearing based on historical performance.
We use the alert_date to identify pixels of alert_delta that correspond
to a specific date range. Mask out the rest. And then calculate both a
weighted area in ha (weighted by the fractional percent alert_delta), and also
a simple count of pixels where clearing occured over a date range.
"""
try:
region = get_region(geojson)
asset_id = SETTINGS.get('gee').get('assets').get('forma250GFW')
logging.info(asset_id)
ic = ee.ImageCollection(asset_id).sort('system:time_start', False)
latest = ee.Image(ic.first())
alert_date_band = latest.select('alert_date')
milisec_date_start = ee.Date(start_date).millis()
milisec_date_end = ee.Date(end_date).millis()
date_mask = alert_date_band.gte(milisec_date_start).And(
alert_date_band.lte(milisec_date_end))
reduce_sum_args = {
'reducer': ee.Reducer.sum().unweighted(),
'geometry': region,
'bestEffort': True,
'scale': 231.65635826395828,
'crs': "EPSG:4326",
'maxPixels': 9999999999
}
area_m2 = latest.select('alert_delta').mask(date_mask).divide(
100).multiply(ee.Image.pixelArea()).reduceRegion(
**reduce_sum_args).getInfo()
alert_area_ha = squaremeters_to_ha(area_m2['alert_delta'])
tmp_counts = date_mask.gt(0).reduceRegion(
**reduce_sum_args).getInfo()
alert_counts = int(tmp_counts['alert_date'])
# logging.info(f"Number of alerts over time period = {alert_counts}")
# logging.info(f"Estimated area loss over time period = {alert_area_ha} ha")
#
# Need to pass the area from the geojson object to area_ha, and also add the
# 'area_ha_loss' key/value into the json that is passed to the front-end.
return {
'area_ha_loss': alert_area_ha,
'alert_counts': alert_counts
}
except Exception as error:
logging.error(str(error))
raise FormaError(message='Error in Forma250 Analysis')
def latest():
"""Gets the date of the latest image
"""
try:
asset_id = SETTINGS.get('gee').get('assets').get('forma250GFW')
logging.info(asset_id)
ic = ee.ImageCollection(asset_id)
latest_im = ic.toList(ic.size()).get(-1).getInfo()
latest_date = latest_im['properties']['date']
logging.info('Retreiving latest date: ')
logging.info(latest_date)
return {'latest': latest_date}
except Exception as error:
logging.error(str(error))
raise FormaError(message='Error in Forma250 Analysis')
def _gee_hansen(geojson, thresh):
image = get_thresh_image(str(thresh),
SETTINGS.get('gee').get('assets').get('biomassloss_v1').get('hansen_loss_thresh'))
region = get_region(geojson)
# Reducer arguments
reduce_args = {
'reducer': ee.Reducer.sum(),
'geometry': region,
'bestEffort': True,
'scale': 90
}
# Calculate stats
area_stats = image.divide(10000 * 255.0) \
.multiply(ee.Image.pixelArea()) \
.reduceRegion(**reduce_args)
return area_stats.getInfo()
def analyze(threshold,
geojson,
begin,
end,
aggregate_values=True,
n_divisions='n4',
method='reduce_regions',
num_pixels=10000,
best_effort=False):
try:
# logging.info("Starting hansen analysis")
# Set constants for the analysis
# make all objects SERVER
# returns ee.Number
num_pixels = ee.Number(num_pixels).long()
threshold = ee.Number(threshold)
begin = ee.Number.parse(ee.Date(begin).format('yy').slice(0, 2))
end = ee.Number.parse(ee.Date(end).format('yy').slice(0, 2))
# returns ee.String
n_divisions = ee.String(n_divisions)
lossyear_band = ee.String('lossyear_').cat(threshold.format())
treecover2000_band = ee.String('treecover2000_').cat(
threshold.format())
treecover2010_band = ee.String('treecover2010_').cat(
threshold.format())
gain20002012_band = ee.String('gain20002012')
asset_id = ee.String(
SETTINGS.get('gee').get('assets').get('hansen_optimised'))
# Get the feature collection of geometries
# returns ee.FeatureCollection
region_nosplit = get_region(geojson)
# logging.info("Got the feature collection of geometries")
def divide_geometry_ntimes(feature, n_divisions=n_divisions):
"""
Optionally divide each of the fc's geoms by nX and return fc.
returns ee.FeatureCollection
"""
fc4 = ee.FeatureCollection(divide_geometry(feature))
fc16 = ee.FeatureCollection(fc4.map(divide_geometry)).flatten()
fc64 = ee.FeatureCollection(
fc16.map(divide_geometry)).flatten()
fc256 = ee.FeatureCollection(
fc64.map(divide_geometry)).flatten()
out = ee.Dictionary({
'n0': ee.FeatureCollection(feature),
'n4': fc4,
'n16': fc16,
'n64': fc64,
'n256': fc256,
})
return ee.FeatureCollection(out.get(n_divisions))
region_split = ee.FeatureCollection(region_nosplit).map(
divide_geometry_ntimes).flatten()
# logging.info("Divided the feature collection of geometries")
# Choose if to split region
# Note code for region_split OR region_nosplit is called depending
# value of Boolean split_geometry=True
# returns ee.FeatureCollection
region_fc = ee.Algorithms.If(n_divisions, region_split,
region_nosplit)
n_features = ee.Number(ee.FeatureCollection(region_fc).size())
# logging.info(f'Number of features is {n_features.getInfo()}')
def get_area(f):
return ee.Feature(
None, {
'area_ha': ee_squaremeters_to_ha(f.geometry().area()),
'px': f.geometry().area().divide(30.0)
})
area_feature = ee.FeatureCollection(region_fc).map(get_area)
total_area = area_feature.aggregate_sum('area_ha')
# logging.info(f"Total area (Ha) is {total_area.getInfo()}")
# logging.info(f"Area (Ha) per feature is {area_feature.aggregate_array('area_ha').getInfo()}")
# logging.info(f"Pixels per feature is {area_feature.aggregate_array('px').getInfo()}")
# if split_geometry divide numPixels by number of splits
num_pixels = ee.Algorithms.If(
n_divisions,
ee.Number(num_pixels).divide(n_features).long(), num_pixels)
# logging.info("Divided the feature collection of geometries")
# Get the optimised hansen asset
# these are all binary bands, except lossyear
hansen_optimised = ee.Image(asset_id)
# Identify year 2000 tree cover at given threshold
# returns ee.Image
treecover2000_image = ee.Image(
hansen_optimised.select(treecover2000_band))
treecover2000_image = treecover2000_image.updateMask(
treecover2000_image)
# returns ee.Number
extent2000 = get_extent_fc(treecover2000_image,
region_fc,
method=method,
bestEffort=best_effort,
scale=False,
numPixels=num_pixels)
extent2000 = ee_squaremeters_to_ha(extent2000)
# logging.info(f"Calculated tree cover extent in year 2000: {type(extent2000)}")
# Identify 2010 tree cover at given threshold
# returns ee.Image
treecover2010_image = ee.Image(
hansen_optimised.select(treecover2010_band))
treecover2010_image = treecover2010_image.updateMask(
treecover2010_image)
# returns ee.Number
extent2010 = get_extent_fc(treecover2010_image,
region_fc,
method=method,
bestEffort=best_effort,
scale=False,
numPixels=num_pixels)
extent2010 = ee_squaremeters_to_ha(extent2010)
# logging.info(f"Calculated tree cover extent in year 2010: {type(extent2010)}")
# Identify tree gain over data collection period
# returns ee.Image
# NOTE GAIN IS NOT THRESHOLDED BY TREECOVER2000!
# FIXME IT MAKES NO SENSE TO EXPORT THRESHOLDED GAIN!
gain20002012_image = ee.Image(
hansen_optimised.select('gain20002012'))
gain20002012_image = gain20002012_image.updateMask(
gain20002012_image)
# returns ee.Number
gain = get_extent_fc(gain20002012_image,
region_fc,
method=method,
bestEffort=best_effort,
scale=False,
numPixels=num_pixels)
gain = ee_squaremeters_to_ha(gain)
# logging.info(f"Calculated tree cover gain between 2000 and 2012:")
# Identify loss
# returns ee.Image
lossyear_image = ee.Image(hansen_optimised.select(lossyear_band))
lossyear_image = lossyear_image.updateMask(lossyear_image)
# Select loss pixels from begin year till end year (0-18)
# returns ee.Image
loss_image_ag = lossyear_image.gte(begin).And(
lossyear_image.lte(end))
# returns ee.Number
loss_ag = get_extent_fc(loss_image_ag,
region_fc,
method=method,
bestEffort=best_effort,
scale=False,
numPixels=num_pixels)
loss_ag = ee_squaremeters_to_ha(loss_ag)
# logging.info("Calculated aggregated tree cover loss in period")
# logging.info("Begin calculating yearly tree cover loss during period")
# Identify loss area per year from beginning year to end year (inclusive)
# Calculate annual biomass loss - add subset images to a collection
# and then map a reducer over image collection
# returns ee.List()
year_list = ee.List.sequence(begin, end, 1)
# returns ee.ImageCollection
def tmp_f(year):
year = ee.Number(year)
return lossyear_image \
.updateMask(lossyear_image.eq(ee.Image.constant(year))) \
.divide(year) \
.set({'year': ee.Number(2000).add(year)})
yearly_loss_collection = ee.ImageCollection(year_list.map(tmp_f))
# logging.info("Created annual biomass loss image collection")
# returns ee.FeatureCollection
def reduceFunction(img):
tmp = get_extent_fc(img,
region_fc,
method=method,
bestEffort=best_effort,
scale=False,
numPixels=num_pixels)
out = ee_squaremeters_to_ha(tmp)
year = ee.Number(img.get('year')).format('%.0f')
return ee.Feature(None, {'year': year, 'loss': out})
output = yearly_loss_collection.map(reduceFunction)
# returns ee.Dictionary
loss_years = ee.Dictionary.fromLists( \
output.aggregate_array('year'), \
output.aggregate_array('loss'))
# logging.info("Calculated yearly tree cover loss during period")
# Choose which loss type to return
# Note code for loss_ag OR loss_years is called depending
# value of Boolean aggregate_values=True
loss = ee.Algorithms.If(aggregate_values, loss_ag, loss_years)
# Create dictionary of results
# ee.Dictionary
d = ee.Dictionary({
'areaHa': total_area,
'loss_start_year': begin,
'loss_end_year': end,
'treeExtent': extent2000,
'treeExtent2010': extent2010,
'gain': gain,
'loss': loss
})
# Evaluate the dictionary of results
tmp_d = d.getInfo()
return tmp_d
except Exception as error:
logging.error(str(error))
raise HansenError(message='Error in Hansen Analysis')
import ee
from flask import Flask
from gfwanalysis.config import SETTINGS
from gfwanalysis.routes.api import error
from gfwanalysis.routes.api.v1 import hansen_endpoints_v1, forma250_endpoints_v1, \
biomass_loss_endpoints_v1, landsat_tiles_endpoints_v1, histogram_endpoints_v1, \
landcover_endpoints_v1, sentinel_tiles_endpoints_v1, highres_tiles_endpoints_v1, \
recent_tiles_endpoints_v1, whrc_biomass_endpoints_v1, mangrove_biomass_endpoints_v1, \
population_endpoints_v1, soil_carbon_endpoints_v1, mc_analysis_endpoints_v1, \
recent_tiles_classifier_v1, composite_service_v1, geodescriber_endpoints_v1
from gfwanalysis.routes.api.v2 import biomass_loss_endpoints_v2, landsat_tiles_endpoints_v2, nlcd_landcover_endpoints_v2
from gfwanalysis.utils.files import load_config_json
logging.basicConfig(
level=SETTINGS.get('logging', {}).get('level'),
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
datefmt='%Y%m%d-%H:%M%p',
)
# Initializing GEE
gee = SETTINGS.get('gee')
ee_user = gee.get('service_account')
private_key_file = gee.get('privatekey_file')
if private_key_file:
logging.info(
f'Initializing EE with privatekey.json credential file: {ee_user} | {private_key_file}'
)
credentials = ee.ServiceAccountCredentials(ee_user, private_key_file)
ee.Initialize(credentials)
ee.data.setDeadline(60000)
"""EE LANDSAT TILE URL SERVICE"""
import json
import logging
import ee
import redis
from gfwanalysis.config import SETTINGS
from gfwanalysis.errors import LandsatTilesError
r = redis.StrictRedis.from_url(url=SETTINGS.get('redis').get('url'))
class RedisService(object):
@staticmethod
def check_year_mapid(year):
text = r.get(year)
if text is not None:
return json.loads(text)
return None
@staticmethod
def get(year):
text = r.get(year)
if text is not None:
return text
return None
@staticmethod
def set_year_mapid(year, mapid, token):
def analyze(geojson):
"""
Analyze NLCD Landcover
"""
try:
logging.info(f'[nlcd-landcover-service]: Initialize analysis')
# nlcd land cover
d = {}
# Extract int years to work with for time range
valid_years = [{
'id': 'NLCD2001',
'year': 2001
}, {
'id': 'NLCD2006',
'year': 2006
}, {
'id': 'NLCD2011',
'year': 2011
}, {
'id': 'NLCD2016',
'year': 2016
}]
# Gather assets
band_name = 'landcover'
landcover_asset = SETTINGS.get('gee').get('assets').get(
'us_landcover')
image_list = [
ee.Image(f"{landcover_asset}/{year.get('id')}")
for year in valid_years
]
us_landcover = ee.ImageCollection(image_list).select(band_name)
# region = ee.Feature(geojson).geometry()
region = get_region(geojson)
scale = 30
logging.info(
f'[nlcd-landcover-service]: built assets for analysis, using {band_name}'
)
# Calculate landcover with a collection operation method
stats = us_landcover.map(
ImageColIntersect(region, scale,
ee.Reducer.frequencyHistogram())).getInfo()
logging.info(f'[nlcd-landcover-service]: retreived {stats}')
# Format data structure
data = [{
'id': d['id'],
'stats': {
k: v * 30 * 30 * 1e-4
for k, v in d['properties'][band_name].items()
}
} for d in stats['features']]
tmp = {}
for el in data:
year = [y['year'] for y in valid_years
if y['id'] == el['id']][0]
tmp[year] = lookup('nlcd_landcover', el['stats'], False)
d['nlcd_landcover'] = tmp
return d
except Exception as error:
logging.error(str(error))
raise NLCDLandcoverError(message='Error in LandCover Analysis')
def analyze(threshold, geojson, begin, end):
"""
Takes WHRC Carbon and Hansen loss data and returns amount of biomass, carbon and co2 lost
by time.
"""
try:
logging.info(f'[biomass-loss-service]: with properties passed: {threshold}, {begin}, {end}')
# hansen tree cover loss by year
d = {}
# Extract int years to work with for time range
start_year = int(begin.split('-')[0]) - 2000
end_year = int(end.split('-')[0]) - 2000
# Gather assets
band_name = f'loss_{threshold}'
hansen_asset = SETTINGS.get('gee').get('assets').get('hansen')
biomass_asset = SETTINGS.get('gee').get('assets').get('whrc_biomass')
hansen = ee.Image(hansen_asset).select(band_name)
biomass = ee.ImageCollection(biomass_asset).max()
region = get_region(geojson)
logging.info(f'[biomass-loss-service]: built assets for analysis, using Hansen band {band_name}')
# Reducer arguments
reduce_args = {
'reducer': ee.Reducer.sum().unweighted(),
'geometry': region,
'bestEffort': True,
'scale': 30,
'tileScale': 16
}
# mask hasen data with itself (important step to prevent over-counting)
hansen = hansen.mask(hansen)
# Calculate annual biomass loss with a collection operation method
def reduceFunction(img):
out = img.reduceRegion(**reduce_args)
return ee.Feature(None, img.toDictionary().combine(out))
# Calculate annual biomass loss - add subset images to a collection and then map a reducer to it
# Calculate stats 10000 ha, 10^6 to transform from Mg (10^6g) to Tg(10^12g) and 255 as is the pixel value
collectionG = ee.ImageCollection([biomass.multiply(ee.Image.pixelArea().divide(10000)).mask(
hansen.updateMask(hansen.eq(year))).set({'year': 2000 + year})
for year in range(start_year, end_year + 1)])
output = collectionG.map(reduceFunction).getInfo()
# # Convert to carbon and co2 values and add to output dictionary
biomass_to_carbon = 0.5
carbon_to_co2 = 3.67
d['biomassLossByYear'] = {}
d['cLossByYear'] = {}
d['co2LossByYear'] = {}
for row in output.get('features'):
yr = row.get('properties').get('year')
d['biomassLossByYear'][yr] = row.get('properties').get('b1')
d['cLossByYear'][yr] = float(f"{d.get('biomassLossByYear').get(yr) * biomass_to_carbon :3.2f}")
d['co2LossByYear'][yr] = float(f"{d.get('cLossByYear').get(yr) * carbon_to_co2 :3.2f}")
# # Calculate total biomass loss
d['biomassLoss'] = sum([d.get('biomassLossByYear').get(y) for y in d.get('biomassLossByYear')])
return d
except Exception as error:
logging.error(str(error))
raise BiomassLossError(message='Error in BiomassLoss Analysis')
