9 arcsecond gridded HCAS 2.1 (2001-2018) base model estimation of habitat condition for terrestrial biodiversity, 18-year trend and 2010-2015 epoch change for continental Australia

Brief description

This data collection comprises, for continental Australia, 9-arcsecond gridded HCAS version 2.1 (2001-2018) base model (v2.1) estimation of habitat condition for terrestrial biodiversity, an 18-year condition trend derived from annual epochs, two 10-year epochs ending 2010 and 2015 respectively, the epoch change for 2010-2015, and a series of ancillary datasets to support their use.
This HCAS version 2.1 (v2.1) product suite comprises 19 datasets of which 17 are 9 arcsecond raster datasets covering continental Australia. All raster datasets are in GeoTIFF format (*.tif) at 0.0025° resolution (approx. 250 m grids), GDA 1994 datum.
The HCAS v2.1 base model dataset describes natural habitat condition for native biodiversity applicable to the assessment period, 2001 to 2018, with scores ranging from 0.0 (habitat completely removed) to 1.0 (habitat in reference condition), being the ‘best’ estimate possible using the current HCAS implementation.
The over-arching framework for the method was published in an internationally peer-reviewed scientific journal by Harwood et al. (2016) and has since been incrementally improved through two trial implementations (HCAS 2.0 beta – Williams et al. 2020, and this version HCAS 2.1 – Williams et al. 2021.).
The 19 datasets in this collection:
A. HCAS results – the base model and two 10-yr epochs (3 datasets):
1. HCAS v2.1 (2001-2018) base model. Developed using 18 years of remote sensing data to provide the HCAS ‘best estimate’ of habitat condition for that period.
2. HCAS v2.1 (2001-2010) epoch. HCAS ‘derived estimate’ of habitat condition for that 10-year period.
3. HCAS v2.1 (2006-2015) epoch. HCAS ‘derived estimate’ of habitat condition for that 10-year period.
B. Trend and epoch condition change (6 datasets):
1. HCAS v2.1 (2001-2018) condition trend, standard error and masked significant trend for the 18-year assessment period.
2. HCAS v2.1 (E2010-E2015) change in habitat condition between the derived estimates of habitat condition for the two 10-year epochs, standard error of the difference in their interannual means, and masked significant change in habitat condition between epochs.
C. Partial measures of uncertainty for aspects of the base modelling framework (4 datasets):
1. HCAS v2.1 (2001-2018) Standard error of the interannual mean over 18 years of annual fluctuations in estimated departures of remote sensing characteristics of a test site of interest from those of environmentally similar reference sites.
2. HCAS v2.1 (2001-2018) Distance in kilometres from a test site to any nearest reference site.
3. HCAS v2.1 (2001-2018) Predicted environmental distance from a test site to any most environmentally similar reference site within 200km.
4. HCAS v2.1 (2001-2018) A relative index of certainty that integrates how ‘nearby’ and how relevant were the 20 reference sites that were used as benchmarks to estimate each pixel’s condition.
D. Masks to support appropriate use and interpretation of HCAS products (3 datasets):
1. HCAS v2.1 mask used in developing the product suite
2. HCAS v2.1 mask plus lakes mask – additional surface water areas from consideration when using the product suite or applying area-based summary statistics
3. HCAS v2.1 mask plus lakes plus plantations mask – to additionally exclude surface water areas and plantations from consideration when using the product suite or applying area-based summary statistics
E. Reference sites used in developing HCAS v2.1 product suite (3 datasets):
1. HCAS v2.1 inferred reference condition sites (method as detailed in Williams et al. 2020 for HCAS v2.1)
2. HCAS v2.1 reference condition sites – training sample
3. HCAS v2.1 reference condition sites – benchmark sample

Lineage

The HCAS v2.1 base model product suite was developed using a combination of:
a) time series remote sensing metrics, derived from the CSIRO MODIS fractional cover dataset and MODIS persistent and annual vegetation cover; updated to Collection 6;
b) Gridded environmental layers, including climate from the 9s climatology for continental Australia 1976-2005, soil and landscape data from the Soil and Landscape Grid of Australia aggregated to 9s; and
c) Spatial inference training data representing reference condition sites (locations in natural areas that are the least modified examples of their type) used to build the model of natural vegetation remote sensing signal, sampled to improve representation across Australia’s varied environments.

The HCAS v2.1 base model was updated to address (as far as possible) several known limitations and methodological enhancements identified in developing HCAS v2.0 (beta). The enhanced approach, as HCAS v2.1, builds upon the work detailed in the earlier, HCAS v2.0 technical report: Williams KJ, Donohue RJ, Harwood TD, Lehmann EA, Lyon P, Dickson F, Ware C, Richards AE, Gallant J, Storey RJL, Pinner L, Ozolins M, Austin J, White M, McVicar TR and Ferrier S (2020) Habitat Condition Assessment System: developing HCAS version 2.0 (beta). A revised method for mapping habitat condition across Australia. Publication number EP21001. CSIRO Land and Water, Canberra, Australia. DOI: https://doi.org/10.25919/85f4-1k65.

These enhancements include:
- extension of assessment period to 18 years, 2001-2018
- development of all remote sensing variables using MODIS collection 6 (formerly collection 5)
- stratified random sampling of spatial inference reference condition training data (c.100,000 9s grids of a much larger sample)
- landscape water distribution derived from a MODIS remote sensing algorithm
- coastal registration issues addressed (especially western-aspects)
- snow lie issues addressed (montane seasonal variation)
- updated MODIS collection 5 to collection 6 remote sensing variables
- refined 2D histogram (probability of being in reference condition) used in generating the uncalibrated condition algorithm
- a non-linear calibration of habitat condition in the range 0-1

Additional products include an 18-year condition trend analysis derived from eighteen 1-year epochs, and two 10-year epochs (2001-2010 and 2006-2015) that were used to derive condition change between 2010 and 2015. Supplementary products include partial measures of uncertainty for specific aspects of the base modelling framework, the complete set of inferred reference sites and the sample locations used as training and benchmark data respectively. A product brief is also provided with this collection that describes the datasets.

A comprehensive review of the remaining limitations and opportunities for further enhancing HCAS is provided in the technical report accompanying this collection: Williams KJ, Harwood TD, Lehmann EA, Ware C, Lyon P, Bakar S, Pinner L, Schmidt RK, Mokany K, Niel TGV, Richards AE, Dickson F, McVicar TR and Ferrier S (2021) Habitat Condition Assessment System (HCAS version 2.1): Enhanced method for mapping habitat condition and change across Australia. Technical report EP2021-1200. CSIRO, Canberra, Australia.