Post provided by Tom Harwood, Randall Donohue, Simon Ferrier, Tim McVicar, Graeme Newell, Matt White and Kristen Williams
It’s very hard to make sensible choices without sensible information. When it comes to actions around changing land use and its ecological impact though, this is often what we are forced to do. If we want to reduce the impact of human activities on natural ecosystems, we need to know how much change has already occurred and how altered an ecosystem might be from its “natural” state.
Working out which parts of the landscape have been changed and mapping the absence of natural vegetation is an achievable (though onerous) task. However, moving beyond this binary view of the world is a huge challenge. Pretty much all habitat has been modified by human influences to some extent – by, for example, wood extraction, the introduction of invasive species or livestock grazing. This means that a lot of the apparently native habitat is no longer capable of supporting its full complement of native biodiversity.
Assessing the Condition of Habitat for Biodiversity
Whilst the term “condition” can have many meanings – from the silkiness of hair, to the state of a second hand car – we use “condition” to describe the capacity of an area of land to support its full complement of native biodiversity.
At a local scale, people can directly survey the landscape and use consistent field measurement protocols to provide an assessment of habitat condition for biodiversity. But when you move to larger areas, such as continental Australia, this rapidly becomes too expensive. Unfortunately, this is the very scale at which information is most important for policy making. We need a way to build on the value of field data.
Working at a Larger Scale
CSIRO has partnered with the Victorian Government Arthur Rylah Institute, and more recently the Australian Government Department of the Environment and Energy, to come up with a practical solution to the problems facing both research scientists and policy makers. Within CSIRO, we have drawn on expertise in large scale fine resolution modelling of biodiversity responses, analysis of the dynamics of remotely sensed (satellite-derived) vegetation and water information, and statistics to develop a method allowing a continental view of condition for Australia.
The Challenges of Using Remotely Sensed Data
At first sight it may seem quite an easy task to map habitat condition from remotely sensed data. After all, satellites are increasingly sophisticated and carry a vast array of sensor types. However, habitat condition is not something you can see directly. In the same way that a photo of a second hand car may tell you that it looks nice, but not whether it will run, remote sensing will measure how the land surface affects electromagnetic radiation but it will not tell you about the surface’s naturalness. For example, if half the trees in a dense forest are removed, satellite information will tell you that it looks like healthy open woodland. Only by knowing or inferring its original forest state can we conclude that the habitat condition has been altered.
So surely we could just model what the habitat should look like and see how different it looks from space, right? Unfortunately, for any set of environmental conditions we find that community composition varies continuously. Sometimes there may be distinct habitats such as savanna and rainforest coexisting. Sometimes cyclical succession will drive a mosaic of vegetation types. Sometimes a natural fire or flood event may cause temporary natural disturbance. If you are not careful, you may end up attributing these natural processes to a change in habitat condition.
Another challenge is figuring out how to use satellites to ‘see’ vegetation naturalness. Satellite remote sensing can’t measure naturalness directly. It can measure various characteristics of vegetation that we know are important indicators of naturalness though – such as forest canopy foliage cover, the seasonal dynamics in grass foliage or the amount of litter covering the ground. As such, remote sensing only provides proxy measures of naturalness. It does, however, provide them in great detail across entire continents and at very low cost.
HCAS: The Habitat Condition Assessment System
We developed the Habitat Condition Assessment System (HCAS) as an approach which tackles many of these issues. We treat ecological communities as continuous in space and naturally dynamic; varying within and between years. In a nutshell, our approach uses a host of both remotely sensed vegetation naturalness metrics and enviro-climatic metrics alongside best available, field-based condition estimates.
For a given target location, we compare the enviro-climatic and naturalness metrics to those of a similar reference location where the vegetation condition is known. Using a new modelling technique, we estimate what the remote sensed naturalness metric values would be at the target location if that location were to have the same condition as the reference site. The difference between the estimated and observed metric values provides an indication of the relative condition of the target site. While the subtleties of the approach are quite convoluted the upshot is that if supplied with adequate field or inferred condition data, HCAS can provide fine scale condition estimates over large areas.
HCAS and Decision Making in Australia
At CSIRO, we are now working with the Australian Government Department of the Environment and Energy to collate field and improved environmental data and refine the HCAS approach. We aim to create a national scale, 250m resolution product which is fit for the purpose of national decision making.
HCAS is not a replacement for local field surveys, rather it adds value to these by extrapolating out expert assessments of condition across the landscape. The product has the potential to act as an “alert” system by identifying places where more information is needed locally, and helping to inform national scale assessments about where biodiversity habitat is likely to be relatively intact, degraded or removed.
A future challenge will be to upscale HCAS from a continental to a global scale. Given the huge variety of ecosystems and variation in field data availability and quality, this will be a significant task – but it’s one that we’re looking forward to.
BTW… if you’re wondering which car actually has the working engine, you’d need to check ‘under the bonnet’ – which is exactly what HCAS helps us to do for our remote sensing information.
To find out more about HCAS, read our Methods in Ecology and Evolution article ‘Habitat Condition Assessment System: a new way to assess the condition of natural habitats for terrestrial biodiversity across whole regions using remote sensing data’.
This article is part of our National Tree Week Virtual Issue. All articles in this Virtual Issue are freely available for a limited time.
I would add beyond the inclusion of field study and verification a need to better understand the vertical relationship of the forest. We tend to treat it in a flattened plan view. Remote sensing will increase this tendency to flatten the dynamics of the layers within a forest. The use of possible research from the US government and Smithsonian NEON open source forestry research network of data observation towers would help to alleviate the flattening effect and flush out more scientifically field study and verification. The forest has both a horizontal and vertical structure to its ecological dynamics.
The hope is that field assessments of condition take structure into account (i.e. the assessors look in all directions!) The real challenge is in the remote sensing component. Whilst some (e.g. LIDAR) sensors can describe vertical structure, spatial coverage, temporal frequency and repeats tend to be a limitation at present, In principle, however, HCAS can incorporate all these data sources.