Three conditions Globally


Protecting 50% of the world's ecosystems by 2050 (IUCN) is an important and audacious goal. In the US, the Biden administration recently committed to protecting 30% of lands and waters by 2030 (news). Central to meeting these goals of conserving biodiversity as well as transitioning to a green energy economy by reducing carbon emissions through sustainable development and other means, is to move from: goal to implementation.

What would a road map based on a generalized, simplified map of human development (see recent post) look like? How could it be used to guide conservation that better recognizes the socio-economic context in which conservation can add profound value to our lives?
Inspired by a map of three land use conditions, I decided to take on the (self-imposed) challenge. I generated a refined and more detailed map with three classes that spans  landscapes from urban areas to shared (or "working") lands to wildlands, using a recent map of the degree of human modification.


First thing you should ask is: how are the classes differentiated?

I distinguished wild lands (i.e. natural) based on a threshold that includes the vast majority (95%) of protected areas and parks that protect biodiversity (and therefore count towards the goal of meeting 30% by 2030 calculations). The distinction between shared or "working lands" (e.g., multi-use resource lands), and urban (including urban built-up lands and cropland/pasturelands) is based on a threshold from conservation science that marks where habitat loss and fragmentation rapidly increase at the edge of urbanized areas.
Why are these classes valuable? Because the kind of management action will need to fit the context. So, for wild lands, land managers would focus on "protecting what is natural, natural" and minimize any land activities that may fragment habitat. For shared lands, management would focus on providing a permeable landscape to allow wildlife to roam and allow adaptation to climate change to occur through connected ecosystems. For urbanized (developed) lands, management would focus on targeted protection of unique habitats and riparian lands, and/or restoration of seriously degraded lands.

Want to explore these maps further? Check these out.



Excited to be attending an in-person meeting again! I am presenting on the need for better global ecosystem mapping and leading an Ecological Connectivity networking session at the North American Congress on Conservation Biology 2022.

Multiple ways to calculate new images based on a conditional expression


A common task is to calculate a new image based on the values of an existing image (or multiple images). (Many other GIS programs call this a "raster calculator").

But, it turns out that doing this with more than one or two conditional expressions is a bit unwieldy in EE. For example, what if we wanted to identify vegetational "life zones" that are often characterized by elevational thresholds, defined as: alpine is greater than 11,500 feet upper montane forest is below alpine and greater than 8,000 feet lower montane forest is below upper montane, that is less than 8,000 feet I've worked out three ways this can be accomplished here in EE code -- moving from a simple to a slightly more involved approach -- but the benefits are vastly increasing the power and flexibility of your code.

Enjoy, and Happy Halloween!

Landscape patterns: landscape signature code


Here's the link to the Google Earth Engine code to demonstrate the landscape signature metric. See this post for a brief description of this approach.

Landscape signature is a simple, informative, and robust metric to analyze landscape patterns, particularly of "patches" of a discrete unit such as habitat patches. This metric builds on the GISFrag metric (Ripple et al. 1991) computes fragmentation by computing the average distance from the nearest patch and accounts for the configuration of patches. This metric assumes a binary landscape (e.g., habitat/non-habitat), and is computed by finding the straight-line distance away from patches (shown in white below). Smaller distance values from the edges of polygons indicate less fragmented landscapes and larger values indicate higher levels of fragmentation (see darker brown colors in second image below). A summary measure of GISFrag is typically calculated as the mean of the distance values.

This approach can be extended to a measure not just the configuration of patches within the matrix, but measure the size and shape of patches themselves (darker green areas are "cores").

And, developing a graph as a "signature" of the landscape assists greatly in a deeper and more robust understanding of the patterns. Both the shape and configuration of patches in a landscape can be examined by measuring how the proportion of a landscape that is occupied by patches changes when patches are enlarged (buffering out) and shrunk (buffering in) across a range of scales. This can be computed by calculating both the straight-line distance away from and into the patches. Notice that the proportion of patch “habitat” is shown by the frequency of pixels at the intersection of 0 with the x-axis. The shape of the curve with positive values indicates the configuration of the patches, while negative values reflect the number, shape, and size of the patches.

So, how do you calculate this in Google Earth Engine? How are results of this metric used? What are its particular sensitivities? More information and scripts are provided in this link provided in the Geographical Analysis using Google Earth Engine learning materials.

Ripple, W.J., G.A. Bradshaw, and T.A. Spies. 1991. Measuring forest landscape patterns in the Cascade Range of Oregon, USA. Biological Conservation 57: 73-88.