Most ecosystems capable of supporting elephants in Asia were lost since 1700s

Anthropogenic land-use change is a known driver of habitat loss for species globally. We use Asian elephants as a flagship surrogate species from whose perspective to trace the historical trajectory of these declines and find that they accelerated since the 1700s, concurrent with colonialism.
Published in Earth & Environment
Most ecosystems capable of supporting elephants in Asia were lost since 1700s
Like

How did we get here?

I was staring at a map by Olivier (1978) that purported to show the post-Pleistocene distribution of Asian elephants, on which I had overlaid the currently known distribution of elephants.  Taken at face value, these outlines suggested a precipitous decline - this species, Elephas maximus, had once occurred across the entire breadth and latitudes of Asia, yet today they remained in just a few fragmented pockets of habitat. It was often stated in the media and popular literature that elephants had lost as much as 90% of their habitat, but I could find no source for this assertion other than the two superimposed maps I was looking at. But surely, I thought, this couldn't be correct, since elephants could not have lived in the deserts and mountains contained within large parts of the continent. Yet, given that Asian elephants are considered a single species, and those on the mainland are likely to have been connected some time in the past, there did appear to be a serious loss of habitat at some point. How did we get here, when and where did this occur? Was there any way we could identify the areas more accurately?

Past (pale orange) and current (purple) distribution of Asian elephants, together with sampling locations (red dots).
Suspected past (pale orange) and current (purple) distribution of Asian elephants, together with sampling locations (red dots). Image from Figure 1 of de Silva et al. 2023.

Reconstructing history

So I turned to an approach known as 'ecological niche modelling.' Simply put, we try to use known locations where a species occurs to identify where else it could occur by looking for similar environmental conditions but for which we do not have occurrence data. Often this approach is used to predict the possible distributions of species in space, hence the term 'species distribution model' is also used. But the approach can also be applied to other points in time, most typically to predict what areas might be suitable for a species under different future climatic conditions. I thought, why not do the same thing, except looking backwards? Although climatic conditions are of course important, for terrestrial species it is land-use and land cover (LULCC) that most directly affects where they can live. The difficulty is that historical data on environmental conditions are difficult to come by, and species are often highly specialized in terms of the conditions that suit them.

As a species, Asian elephants were relatively simple to deal with because they are actually capable of occupying a wide range of habitat types - from dry seasonal grasslands, to moist evergreen rainforests. So elephants could be used as a proxy, or surrogate, for these diverse ecosystem types. They can also potentially move over large distances so the environmental variables did not have to be of particularly high resolution. I was in luck, because a group of researchers based at the University of Maryland had released newly released their Land-Use Harmonization Dataset with global reconstructions of land-use going all the way back to the 9th century! My co-authors and I compiled data from various sources on where wild elephants occurred near the turn of this millennium, which I then narrowed down so as not to over-represent particular habitat types or ecosystems, as well as remove landscape types (e.g. agricultural) where elephants were likely to conflict with people. We then related these locations to land-use variables for the year 2000.

As a sanity check, we also constructed an entirely separate model based on entirely different finer-scale datasets that included variables other than land-use, such as human and livestock population densities. We tried various techniques of doing so at first, but ultimately settled on using Maximum Entropy (MaxEnt), a machine-learning algorithm, based on its visibly better performance. Comparing the results based on the LUH variables and the other variables, there was overwhelming agreement (over 80%) between the two models, and if anything the LUH was more conservative as it predicted less areas as being suitable for elephants overall.

We then ran the MaxEnt model on LUH variables going back all the way to the year 850, up to the year 2015. Based on existing literature and the history of colonialism in Asia, we expected major changes to take place around the 1700s, which is in fact what we saw across most of the range but especially in South Asia and mainland China. To simplify things, we converted the resulting maps into binary suitable/unsuitable classification categories based on a thresholding procedure, which then allowed us to calculate the total extent of available habitat in each year as well as various metrics of fragmentation.

Good news and bad news

The results showed that over 60% of the areas suitable for elephants had been lost, much of it since the 1700s, amounting to over 3 million km2. This is both good news and bad news - on the one hand, the extent of loss might not be as great in magnitude as the previous crude estimates suggested. On the other, it is probably a far more accurate picture, and given that elephants were probably never found throughout all these landscapes to begin with, the possible scale of decline is staggering. India and China, which had the largest land areas with suitable habitat, each lost more than 80% of their habitat.

Curiously, a large hotspot of habitat in what is now central Thailand appeared to remain relatively intact despite the losses taking place elsewhere. Dramatically, this virtually disappears within the 20th century. Looking at the precise timing, it seems to coincide with the adoption of both high-intensity logging as well as the spread of industrial agriculture. On the islands of Borneo and Sumatra, where there are now substantial areas that have been converted to plantations, the areas of suitability re-configured and shifted. Looking at areas that today contain elephant range, our results indicate that landscapes within 100km of the extant range all across Asia could have been considered suitable for elephants back in the 1700s. By 2015 less than half of the known elephant range could be considered suitable.

Change in availability of suitable habitat for elephants over 315 years
Loss (orange) and gain (green) in suitable habitat for Asian elephants. Black outlines show the known range of the species as of the past two decades. The inset circle shows that 100% of the area within 100km of this known range could have been considered suitable for elephants in the 1700s, but only 48.6% of the current range was classified as suitable by the year 2015.

It’s important to note here that these suitable areas represent landscapes that could potentially support elephants, not where they actually are or were. we cannot say for sure whether elephants were in fact found throughout the predicted areas - ecological niche models predict the so-called fundamental niche of a species, but there are many reasons why a species may not occur where it could (e.g. dispersal constraints, overhunting) or conversely, why it occurs in places that don’t seem appropriate (e.g. time lags in population responses to ecological change). In the study we discuss the particular issues and challenges for elephants on a country-by-country basis that may affect their actual distributions.

Three things are clear, however.  First, the mismatch between where elephants are found today and the landscapes that can actually be considered suitable, is driven not only by recent human activities, but very long-term processes including both colonial influences and the industrialization and intensification of agriculture.

Second, this mismatch likely drives so-called human elephant conflicts throughout the range as elephants are long-lived and demographic impacts take a long time to manifest. They are moreover quite mobile, therefore elephant populations are likely try to move to better habitat, which can challenge human populations that now occupy these landscapes. In a separate study, we looked at how elephants use a protected area in South Asia and found that a whopping 3/4 of the population were likely to be non-resident. Therefore protected areas, which are generally small in Asia, are unlikely to ever fully support elephant populations. Other types of landscapes are crucial (more about that here).

This brings us to the third point. While it is tempting to think of the landscapes in which elephants were and are found as being “natural” or “wilderness” areas, they are and perhaps always have been influenced by people to a certain degree. Asian elephants in fact thrive in landscapes with some degree of disturbance as opposed to untouched primary forests; historically, such regimes would have been encouraged and facilitated by the traditional land management practices of indigenous, agrarian, and pastoral communities. Therefore, elephant ecosystems, perhaps fundamentally, were also human ecosystems. In thinking about the future management of these landscapes, it is therefore imperative to understand exactly what how important the human component was and how these regimes can be maintained given the constraints we face today.

Further reading:

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Subscribe to the Topic

Earth and Environmental Sciences
Physical Sciences > Earth and Environmental Sciences

Related Collections

With collections, you can get published faster and increase your visibility.

Retinal imaging and diagnostics

This Collection invites works providing insight into using novel or existing retinal imaging technologies in clinical applications or presents new or adaptive forms of these techniques.

Publishing Model: Open Access

Deadline: Apr 30, 2024