HII POWER DRIVER

What does this task do?

This task calculates the impact of power on the terrestrial surface as one of the key drivers for a combined Human Impact Index⁠. "Impact" is a pressure score based on the intensity of electricity usage as measured by "Night Time Lights" datasets. An inter-calibrated nighttime lights dataset with values ranging from 0 - 63 is transformed to a pressure score of 0 - 10 using equal interval quantiles, calculated from the earliest available image (1992) using the quantile_calc function in input_preprocess.py. These are:

quantiles = {
    "0": {"value": 0, "min": 0, "max": 0},
    "1": {"value": 1, "min": 1, "max": 4},
    "2": {"value": 2, "min": 5, "max": 5},
    "3": {"value": 3, "min": 6, "max": 6},
    "4": {"value": 4, "min": 7, "max": 7},
    "5": {"value": 5, "min": 8, "max": 9},
    "6": {"value": 6, "min": 10, "max": 11},
    "7": {"value": 7, "min": 12, "max": 16},
    "8": {"value": 8, "min": 17, "max": 30},
    "9": {"value": 9, "min": 31, "max": 62},
    "10": {"value": 10, "min": 63, "max": 63},

The output HII driver calculated by this task is, like all other HII drivers, unitless; it refers to an absolute 0-10 scale but is not normalized to it, so the actual range of values may be smaller than 0-10.

Input Dataset Calibration

The HII power driver (0-10) is calculated from the previous calendar year's calibrated nightlights dataset, which is produced on demand by the task by using the HIIPowerPreprocessTask task. Two distinct source nightlight datasets are used to calculate this calibrated version following the methods below; these are the Defense Meteorological Satellite Program (DMSP)/Operational Linescan System (OLS) and the Visible Infrared Imaging Radiometry Suite (VIIRS) on the Suomi National Polar-orbiting Partnership Satellite. DMSP provides data from 1992 - 2013 and VIIRS provides data from 2012 through the present.

Inconsistencies within the DMSP time series require implementing intra-calibration within the DMSP dataset. Here the intra-calibrated dataset produced by Li et al. 2020⁠ is used. Key differences between the DMSP and VIIRS datasets also require inter-calibration of the two datasets. In addition to the calibration process to match VIIRS to DMSP, VIIRS requires substantial noise reduction. These key differences are:

Noise removal

1. Annual composites are created using the median value of all pixels for a given year with raw values greater than 0.
2. General noise (which generally increases with distance from the equator) is removed by creating an image mask using a pixel latitude image. An exponential function is used to stretch the threshold values from a minimum to maximum threshold of 0.1 and 0.75 respectively from 0° to ±60°. The formula for this is:

      t = (|lat|⁴ / 60⁴) x (t_max - t_min) + (t_min)

        Where:
        t is the final threshold image
        lat is a latitude image
        t_min is the minimum threshold
        t_max is the maximum threshold

Calibration steps

1. Regression coefficients are calculated to convert VIIRS values (beginning with 2014) to DMSP values by comparing pixel values of areas with stable lights through time. These are determined statically by selecting pixels from the DMSP time series from 2000 - 2013 that have a standard deviation less than 2. A stratified sample of 300 pixel values for each available DMSP value as of the latest available image (2013) is calculated and used to derive constant slope and intercept coefficients. See the stable_light_points function in input_preprocess.py.
2. The resulting equation to transform VIIRS values to match DMSP is determined to be:
         calibrated_viirs = log(viirs) * 11.62 + 19.4 [R²: 67.16]
3. Values in the calibrated image greater than 63 are set 63, matching the range of values from DMSP.

The result and goal is an annual night lights dataset derived from VIIRS that is calibrated to the pre-2013 DMSP values, which is then transformed into the HII power driver by the main task.

Variables and Defaults

Environment variables

SERVICE_ACCOUNT_KEY=<GOOGLE SERVICE ACCOUNT KEY>

Class constants

scale=300

Usage

All parameters may be specified in the environment as well as the command line.

/app # python task.py --help
usage: task.py [-h] [-d TASKDATE] [--overwrite]

optional arguments:
  -h, --help            show this help message and exit
  -d TASKDATE, --taskdate TASKDATE
  --overwrite           overwrite existing outputs instead of incrementing

In addition, the HIIPowerPreprocessTask task can be called to run one of three preprocessing steps used in the main task.py, with the following constants available as arguments:

BUCKET = "hii-scratch"
CALC_PREVIOUS_ANNUAL_VIIRS = "viirs"
CALC_CALIBRATION_COEFFICIENTS = "coefficients"
CALC_WEIGHTING_QUANTILES = "quantiles"

The first of these, viirs, is used on demand by the main power task.py.

/app# python input_preprocess.py --help
usage: input_preprocess.py [-h] [-d TASKDATE] [-j [{viirs,coefficients,quantiles}]] [--overwrite]

optional arguments:
  -h, --help            show this help message and exit
  -d TASKDATE, --taskdate TASKDATE
  -j [{viirs,coefficients,quantiles}], --job [{viirs,coefficients,quantiles}]
                        Calculation to run with HIIPowerPreprocessTask.
  --overwrite           overwrite existing outputs instead of incrementing

License

Copyright (C) 2022 Wildlife Conservation Society The files in this repository are part of the task framework for calculating Human Impact Index and Species Conservation Landscapes (https://github.com/SpeciesConservationLandscapes⁠) and are released under the GPL license: https://www.gnu.org/licenses/#GPL⁠ See LICENSE for details.