By Rebecca Wheatley, 20th January 1:00 PM

Over the past five months, huge areas of Australia have been devastated by fire and drought. As of writing, over 10 million hectares of land have burned since September and it is estimated that well over a billion wild animals have perished (not including fish, frogs, bats or invertebrates). Animal hospitals are full to bursting with injured and displaced wildlife, and carrots and sweet potato have been dropped in some burned areas to get some food to animals that have nothing left. Provisioning for affected wildlife during this catastrophe is both positive and necessary.

But what about provisioning for wildlife outside of times of extreme crisis? Habitat loss is the number one threat to species in Australia and many animals are affected by this even under normal conditions, particularly in urban areas. Climate change means we are experiencing warmer average temperatures and lower annual rainfall more and more frequently. Under these stressful but no longer unusual conditions, many people (including yours truly) want to help wildlife in their local area. But how can we do this in a responsible way that doesn’t cause more harm than good? I’ve been doing some research on this to use in my own life, and I thought I’d share what I’ve learned.

The first thing I learned was there are multiple kinds of provisioning that are carried out in urban, rural, and wild areas, generally providing some combination of food, water, and shelter. The second thing I learned was that, no matter how innocuous the provisioning you do appears at first glance, there can be unintended negative consequences if it isn’t carried out thoughtfully and responsibly. I will discuss four broad types of provisioning (with a totally biased bent toward fauna rather than flora), their potential pitfalls and ways to negotiate them. However, it is important to remember that the biggest and most impactful things we can do to help our wildlife – animals, plants, fungi, all of it – is to support and engage in climate action and prevent further habitat clearing.

Dandenong Ranges National Park, Victoria. Image credit: Rebecca Wheatley


One of the best ways to support local wildlife is to provide and support natural habitats. This can be done by leaving pre-existing vegetation in place or by planting native gardens. For example, in a large garden and where it is safe to do so, large or dead old trees or fallen wood can be left in place to provide habitat for larger animals like marsupials and hollow-using birds, but also for smaller creatures like lizards and insects.

Many people also plant native gardens to support insects and larger animals. A diverse garden with a multi-layered structure is important for providing shelters, homes, and food for an array of different creatures. For instance, rather than open gardens with only showy, nectar-heavy flowering species like grevillea and eucalypts (which there is some evidence can result in communities dominated by aggressive honeyeaters like Noisy Miners), a variety of native plants that form some dense thickets can cater to other birds and smaller animals as well. Smaller, more subtle flowers can also be useful for encouraging native pollinators. Many online guides are available for helping to create a native garden that caters for a range of different species (e.g. this one from Birds In Backyards).

A native garden being enjoyed by Yellow-tailed Black Cockatoos in Tasmania. Image credit: Rebecca Wheatley


When it isn’t possible to have natural hollows for animals (like dead trees), an alternative is to put up structures like nest boxes and insect “hotels”. However, as with all the provisioning options I will discuss, some pre-thought is required.

While putting up nest boxes in urban areas seems like an activity that could only have positive impacts, long-term monitoring research has found that the installation of bat boxes in parks in Melbourne may have led to the dominance of a single bat species. In fact, nest box use is often dominated by particular species in urban areas (e.g. possums) – which can be either a good or a bad thing, depending on what you are trying to achieve.

Before you put up any boxes, it is necessary to think about the species you aim to encourage. Different types of nest box will allow access for different species, varying in size, entry hole dimensions, and entry platform type. Examples of some different types of nest boxes can be found on the Birds In Backyards website, which has a range of free nest box plans for some of the less common Australian urban birds. Once you’ve decided on a box, it’s important to choose a secure, quiet location to place it which will be out of reach of predators while still being accessible to the animals the box is designed for.

My own nest boxes have frequently been occupied by unintended species: ants. This isn’t necessarily a problem if the ants are native to the area and are not a pest. However, nest boxes can also be occupied by invasive species like Common Myna, Blackbird, Starlings, and European wasps. There are methods of reducing the likelihood that this can happen, including using an appropriately sized entrance hole, installing features like baffles at the front of the box, and checking occupancy regularly.

An insect hotel. Image credit: Rebecca Wheatley


Providing fresh, clean water for wildlife can save lives during heat waves and drought. However, there are a few important things to consider when putting out a water source.

First, you want to put out a water source that is suitable for the sort of animals you are provisioning. Pedestal or elevated baths can be useful for birds because it keeps them out of easy reach of cats. On the other hand, water bowls on the ground are more useful for species like Koala, echidnas, snakes, and larger lizards. Any water source should be placed in a sheltered location, out of the way of people and pets.

Second, water sources must be refilled regularly and kept clean. Parasites and diseases can be transmitted between animals in dirty water baths, so cleaning is essential. Mosquitos can breed in stagnant water, so water must be replaced at regular intervals.

Finally, small animals like skinks and insects can drown in shallow water. Adding a few sticks or rocks will allow them to crawl out if they fall in.

Yellow-faced Honeyeaters visiting a bird bath in New South Wales. Image credit: Peter Firminger


Feeding animals is by far the most controversial type of wildlife provisioning because it can have pretty awful consequences if done poorly. Most animals have not evolved to cope with human foods, and animals that are regularly fed human food can become obese, nutrient deficient, and sick. Feeding macropods such as wallabies and kangaroos can result in a condition called lumpy jaw which prevents feeding and eventually results in death. Similarly, birds fed an artificially high-carbohydrate, high-protein diet are more likely to develop deformities such as angel wing.

Even if animals are given appropriate foods, there are still serious potential negative consequences just in the act of feeding. Regularly fed animals can become dependent on artificial food sources and/or exhibit increased aggression towards other animals and humans. In extreme cases, habituated and aggressive wild animals are euthanised. Feeding animals can also result in more aggressive species becoming dominant in the community, as these species are more likely to gain access to artificial food sources and exclude shyer, less aggressive species.

A Bennet’s Wallaby munching down some natural forage (leaves) at a campground in Tasmania. This gives a good indication of the foods their digestive system has evolved to cope with. Image credit: Rebecca Wheatley

These various side-effects mean that feeding animals must be approached very carefully if it is to be done at all. The safest way to feed wild animals is by filling your garden with natural resources for them and letting them feed themselves. You can plant suitable native vegetation for granivores, frugivores, and nectivores, and use mulch and leaf litter to provide habitat for invertebrates and lizards (to support carnivorous birds, small mammals, reptiles, and frogs).

Recently a lot of research has come out on the pros and cons of feeding birds in Australia. I will not cover this information as it is specific to birds, but if you are keen to know more, some great resources from an expert are: The Birds at My Table: Why We Feed Wild Birds and Why It Matters and Feeding the Birds at Your Table: A Guide for Australia by Professor Darryl Jones.


In addition to the increased stressors affecting our wildlife, it is widely accepted that humans are facing an “extinction of experience” of nature. Essentially, humans living in increasingly developed and urbanised environments rarely experience natural surroundings and phenomenon. The connection and relationship we develop with wildlife through various forms of provisioning can be valuable, not only for the animals, but for us as well. However, we need to be careful to ensure we aren’t harming the animals we are aiming to help, by doing our research and behaving in a thoughtful, considered manner.

All posts are personal reflections of the blog-post author and do not necessarily reflect the views of all other DEEP members.


By Luke Yates, 11th December 3:00 PM

The title of this blog post is a line that captured my attention while reading the recently published book, Range, by David Epstein. It’s a book about the merits of taking a generalist approach to problem solving which got me thinking about measures of efficiency and the importance of human creativity. Given increasing over-population, limited resources and a global ecosystem suffering unprecedented losses, what argument could there be for cultivating inefficiency?

The rise of automation

Clearly, it is imperative that we use our natural resources efficiently, and we are getting better at this. Through urbanisation, technological innovation, division of labour and large-scale industrialisation, we now supply goods and services to more people using less inputs per production unit than ever before. This does not mean we are consuming less resources overall; paradoxically, increased efficiency can lead to lower prices and an increase in demand. In any case, the developed capacity to produce goods efficiently is critical and at the leading edge of our efficiency gains is automation, the revolution of machine learning technologies that are disrupting traditional employment and driving towards zero, the marginal cost of goods and services.

The kind and the wicked

Is it inevitable that we lose our jobs to automation and live on a universal income? Well, humans are remarkably inefficient, and this gives us a real edge over automated processes. What do I mean by this? The usual metrics of efficiency assume that the desired output of a process is both known and measurable. This is true in so-called kind environments, where progress towards goals can be measured, and feedback can be given. The result is the creation of a learning environment where experience improves performance, and humans thrive in these settings. Indeed, we are so successful that our machines perform many tasks much better than us – automation is (indeed) a human achievement. 

However, the most challenging problems that humanity faces are not so kind. Problems such as climate change and global poverty have been dubbed as wicked or even super-wicked (yes, that’s very bad).  Wicked problems are characterised as: having no definitive formulation; having complex interdependencies; each unique; and each unsolvable – only improvable. How do we work on problems like this?

Now a better name for human inefficiency, might be creativity. Humans have diverse interests, competing goals, and take multiple levels of perspective.  We import ideas between our various endeavours, are intrinsically motivated, and self-directed. Unfortunately, these creative qualities are underutilised because we live in a tyranny of metrics which evaluate our performance (and underpin our wages) as if we lived in a kind world. But wicked problems cannot be addressed with machine-like performance. If we continue to allocate resources using business-as-usual metrics, we will only support those who can promise to perform in kind and narrow contexts. This leads to ever increasing specialisation, silos of thought and a reluctance to take risks.

It’s not that specialisation or expert systems are a bad thing. On the contrary, the coupling of specialist skills with broader strategic capacities is a strong combination. For example, in 1997, the grand chess master, Garry Kasparov, was beaten by an artificial intelligence (AI) system for the first time, and computers have dominated the game since this time. However, human players working together with AI, can consistently win against standalone artificial systems.

We have successfully incentivised specialisation for a long time, and we can celebrate the fruits of this labour. But we no longer need to focus on being machine-like ourselves; we have a more strategic and creative role to play. There is an immense landscape of possible strategies that we could apply to our wicked problems; the pressing efficiency question is: How do we search for them? 

Incentivising creativity

The question of how to foster a creative workplace is certainly not new. For decades, technology companies have sought to harness the creative potential of their employees by creating a culture of ‘fail and fail fast’. This acknowledges the nature of creative exploration; that most ideas are failures and the best way to find the good ones is to search, try, fail and search again.

It turns out that our intuition about how fun and satisfying it would be to work in a creative workplace can be quite different from the reality. Bosses need to be ruthless in killing off bad ideas and many of the good ones too. It takes a robust and resilient sense of self to tolerate repeated failure and the inevitable rejection of one’s own ideas.  The creation of a supportive work culture is vital. At times, even great ideas will need to be abandoned if we are to avoid the lure of the sunken cost fallacy

All this amounts to the sharing of risk. To encourage individuals or groups to efficiently explore new solutions, the risk of failure needs to be collectively absorbed. This could liberate individual creativity, but it is not obvious how to resolve the problem of personal accountability. What are the right metrics of performance? Many people object to the notion of a universal income because they (erroneously) believe that most other people are lazier than they are. Is it possible that our strong aversion to support a minority of so-called free-riders is thwarting our collective efforts to address our most pressing problems?

A lifetime of work spent exploring a new theory seems like a slow process, but millions of theories being explored simultaneously by a generation of people, would enable a rapid exploration of a vast and complex search space. Maybe our metrics of performance are so tyrannous that they suppress our most useful contributions and synergies.


I’ll leave you with a manifesto of sorts:

Let the machines become more specialised. May they perform mundane and technical tasks better than we ever have. May we work with these technologies as we (re)ignite our creative capacities. Let us innovate, decoupling our incomes from ineffective metrics, while we work to increasingly decouple our economic activity from natural ecosystems. The provision of services, to each other, is virtually unlimited, providing a means to meet our many human needs in a resource-efficient manner. To measure our true efficiency, we need to take the long view. Successes and breakthroughs have high variance yet pay good dividends to society once realised. To efficiently explore the vast landscape of possibility we need to cultivate and celebrate our small-scale inefficiencies – or rather, our natural curiosity and creative potential.

All images are sourced from and licensed for free commercial use.

All posts are personal reflections of the blog-post author and do not necessarily reflect the views of all other DEEP members.


By Tessa Smith, 7th October 11:30 AM

Like all good netizens, I have at least once gotten into an argument with strangers on the internet.  The argument in question was about whether it was ‘ethical to purposely drive a species (in this case, a mosquito) to extinction’. I argued that it was not ethical, on the basis that the species has the same inherent right to survive as all other species, including the charismatic ones. The other parties (for there were many) argued that I was a privileged idiot with no empathy for the millions of people worldwide affected by the diseases. They probably had a point, but ethics aside, it got me thinking about 1) what the problem with the mosquito was, 2) what was being developed to supress it, and 3) how likely it was that the species could become extinct as a result. Thanks to Emily Flies for her helpful suggestions to the piece.

1. The Problem

According to the World Health Organisation, mosquito-borne diseases cause several million deaths and hundreds of millions of infections each year. Mosquitoes of the genus Aedes, Anopheles and Culex are the primary culprits for transmission of infections (e.g. malaria, Ross River virus). In these genera, the female mosquito feeds on blood from a host, using the protein and iron in the blood to grow her eggs. By feeding on several hosts (from the same or different species), they may transmit the blood-borne diseases if present.


Figure 1: Aedes aegypti mosquito, E. A. Goeldi (1905).

Aedes aegypti (Figure 1), which is a vector for multiple diseases and lives in urban areas, thereby threatening over half of the world’s population. They are one of the most important species for transmitting disease for two main reasons: 1) they feed during the day (from sunrise to after sunset) when people are outdoors and active, giving them more opportunities to infect people, and 2) they lay their eggs in small containers of water around urban areas, making it very hard to target control efforts to breeding sites. While not the only insect vector for the multiple diseases it can spread, A. aegpyti is by far the most common and of the most concern. Some of the diseases transmitted by Aedes aegypti are:

  • Yellow fever is endemic in Africa and South America where its primary reservoir is monkeys. The disease can cause death in a small percentage of infected people and there is a vaccine available. During 2017-2018 there were deadly outbreaks of the disease in Brazil and Nigeria.
  • The Zika virus has been recorded in over 86 countries in the Americas, Africa, Asia and the Pacific. While causing few symptoms from infection in most people, it can cause complications of pregnancy and malformation in babies. Those infected also suffer an increase risk of neuropathic conditions. There is no vaccine currently available for Zika.
  • Chikungunya is present in Africa, Asia, the Americas and Europe. It causes fever and joint pain in affected individuals and there is currently no vaccine for the disease.
  • Dengue is most prevalent in Asia, the Americas and Africa, with almost half of the world’s population at risk (Figure 2). The disease may have been underreported in Africa due to a range of factors including presence of similar illnesses (11). The disease disproportionately affects people in developing countries (especially indigenous people, people of colour and immigrants) (10).There is currently no vaccine for Dengue, and the vaccine discovery process proving difficult. Aedes mosquitos are not a vector for malaria.


Figure 2: World Health Organization Estimated Deaths from Dengue per million persons, 2012. Yellow=0, Red=9 (CC BY-SA 4.0).

Here, I focus on the research aimed at suppressing the ability of this species to transmit disease.

The spread of Aedes aegypti away from its native home in Africa and to a large area of the world, especially tropical and subtropical areas is also a major concern. The distribution of the mosquito in some areas of South America has been attributed to ship dispersal during the slave trade in the 1600’s (13) and more recently the trade in used car tyres (WHO Dengue Control). 

Controlling the spread and reproduction of these mosquitoes is critical for reducing disease transmission. A variety of historical control methods have been used against Aedes mosquitos, some of which have caused widespread environmental problems:

a) The mosquito was nearly eradicated in the Americas in the 1960’s using DDT (the chemical Dichlorodiphenyltrichloroethane), but re-established itself there when the project was discontinued (7) due to strong health and environmental concerns.

b) Biological control of the mosquito with fish or predatory copepods has had limited local success in controlling mosquito numbers and has caused ecological problems when the exotic species escape and consume indigenous fauna. The mosquito fish (Gambusia holbrooki) was listed as one of the IUCN’s 100 most harmful invasive exotic species after being released around the world to eat mosquito larvae.

The challenge of combatting the spread of mosquitoes is also influenced by other factors. The cost of effective management programs using insecticides is ongoing, and often a burden for poorer governments. Many mosquito species are evolving insecticide resistance, making insecticides increasingly ineffective. With climate change there is a predicted increase in A. aegypti’s global distribution (Figure 3) as it establishes in areas that were previously too cold to inhabit (3).


Figure 3: Summary of the modelled global distribution of Aedes aegypti under both current (dark blue) and future (dark orange) climatic conditions in 2050 showing stability of predictions at present and into the future with RCP (representative concentration pathway) 4.5 (3).

Traditional methods of controlling Aedes and its disease spread through chemical and biological control are increasingly inneffective and unsustainable, necessitating the use of new methods for control (4). 

2. New options in the control of mosquito-borne disease

Wolbachia bacteria

Wolbachia is a genera of bacteria that occur naturally in 60% of all insect and nematode species, including some mosquitos. The bacterium modifies the sperm of the host insect, only allowing eggs infected with Wolbachia to develop normally (8). The presence of the bacteria within the mosquito has also been found to reduce the ability of the mosquito to transmit Dengue virus to humans.  


Figure 4: The Percentage of mosquitos with Wolbachia, Townsville, Australia, 2015-2016 (6).

The introduction of the Wolbachia bacteria into A.aegypti mosquitos has been successful in local trials in northern Australia since 2011 ( Once introduced, Wolbachia is self sustaining within the local mosquito population (Figure 4). Modelling by Dorigatti, McCormack (4) predicted that infection with Wolbachia reduces a mosquito’s ability to transmit dengue by 40% .

Genetic modification and gene drive

Another method for reducing infection rates is the release of genetically modified Aedes into the wild population. Unlike normal inheritance where an altered gene is not always spread, with gene drive inheritance, the altered gene is always inherited; male mosquitos carry a dominant lethal gene, that is passed on when they mate with females and kills 96% of their progeny (12). Modified Aedes aegypti mosquitos have so far been released into Burkina Faso and Brazil where they have been able to suppress local populations (4). This process requires the release of many modified mosquitos over several months (9).

Non-gene drive genetically modified mosquitos released in Brazil were found to have passed portions of their DNA into local populations, forming viable hybrid individuals that bred to pre-release numbers within months (2). The establishment of a genetic monitoring program (2) was recommended as an important part of the development process for any new genetic tool to ensure success.

3. How likely is it that the mosquito will become extinct as a result?

In a laboratory population, gene drive was successfully used to eliminate a population of Anopheles gambiae mosquitoes (a species that spreads Malaria)(5). The likelihood of gene drive and Wolbachia to create a complete extinction of the Aedes aegypti species over their entire range is low, but possible; models suggests that non-random mating will likely prevent gene-drives from causing extinction (1). However, this may be investigated by future research teams.  

All posts are personal reflections of the blog-post author and do not necessarily reflect the views of all other DEEP members.


  1. Bull J. J., Remien C. H., Krone S. M. Gene-drive-mediated extinction is thwarted by population structure and evolution of sib mating. Evolution, Medicine, and Public Health. 2019; 2019(1): 66-81.
  2. Evans B. R., Kotsakiozi P., Costa-da-Silva A. L., Ioshino R. S., Garziera L., Pedrosa M. C., Malavasi A., Virginio J. F., Capurro M. L., Powell J. R. Transgenic Aedes aegypti Mosquitoes Transfer Genes into a Natural Population. Scientific Reports. 2019; 9(1): 13047.
  3. Kamal M., Kenawy M. A., Rady M. H., Khaled A. S., Samy A. M. Mapping the global potential distributions of two arboviral vectors Aedes aegypti and Ae. albopictus under changing climate. PloS one. 2019; 13(12): e0210122.
  4. Dorigatti I., McCormack C., Nedjati-Gilani G., Ferguson N. M. Using Wolbachia for Dengue Control: Insights from Modelling. Trends in parasitology. 2018; 34(2): 102-113.
  5. Kyrou K., Hammond A. M., Galizi R., Kranjc N., Burt A., Beaghton A. K., Nolan T., Crisanti A. A CRISPR–Cas9 gene drive targeting doublesex causes complete population suppression in caged Anopheles gambiae mosquitoes. Nature Biotechnology. 2018; 36: 1062.
  6. O’Neill S., Ryan P., Turley A., Wilson G., Retzki K., Iturbe-Ormaetxe I., Dong Y., Kenny N., Paton C., Ritchie S., Brown-Kenyon J., Stanford D., Wittmeier N., Anders K., Simmons C. Scaled deployment of Wolbachia to protect the community from Aedes transmitted arboviruses [version 1; peer review: 1 approved, 1 approved with reservations]. Gates Open Research. 2018; 2(36).
  7. Hotez P. J. Zika in the United States of America and a Fateful 1969 Decision. PLOS Neglected Tropical Diseases. 2016; 10(5): e0004765.
  8. Jiggins F. M. Open questions: how does Wolbachia do what it does? BMC Biology. 2016; 14(1): 92-92.
  9. Carvalho D. O., McKemey A. R., Garziera L., Lacroix R., Donnelly C. A., Alphey L., Malavasi A., Capurro M. L. Suppression of a Field Population of Aedes aegypti in Brazil by Sustained Release of Transgenic Male Mosquitoes. PLoS neglected tropical diseases. 2015; 9(7): e0003864-e0003864.
  10. Hunter P. Tropical diseases and the poor: Neglected tropical diseases are a public health problem for developing and developed countries alike. EMBO reports. 2014; 15(4): 347-350.
  11. Bhatt S., Gething P. W., Brady O. J., Messina J. P., Farlow A. W., Moyes C. L., Drake J. M., Brownstein J. S., Hoen A. G., Sankoh O., Myers M. F., George D. B., Jaenisch T., Wint G. R. W., Simmons C. P., Scott T. W., Farrar J. J., Hay S. I. The global distribution and burden of dengue. Nature. 2013; 496: 504.
  12. Phuc H. K., Andreasen M. H., Burton R. S., Vass C., Epton M. J., Pape G., Fu G., Condon K. C., Scaife S., Donnelly C. A., Coleman P. G., White-Cooper H., Alphey L. Late-acting dominant lethal genetic systems and mosquito control. BMC Biology. 2007; 5(1): 11.
  13. Mousson L., Dauga C., Garrigues T., Schaffner F., Vazeille M., Failloux A.-B. Phylogeography of Aedes (Stegomyia) aegypti (L.) and Aedes (Stegomyia) albopictus (Skuse) (Diptera: Culicidae) based on mitochondrial DNA variations. Genetical Research. 2005; 86(1): 1-11.


By Kasirat Turfi Kasfi, 26th August 11:30 AM

When I sat down to brainstorm a topic to write for the DEEP blog, I thought what would be that one thing that interest me and the rest of the DEEP members? Data, right! We all work with data, we want to find meanings and patterns in our data. We want to be able to make inference, make decisions based on or make predictions from our data. We do that by using statistical learning methods.

I believe that most of us are familiar with one or more statistical learning models. And there seems to be quite a few of them out there! It can be difficult to decide which one to use! Or knowing which one is the best! There is no best method that fits all kinds of data perfectly, and no one method “to rule them all” (only if Tolkien invented the models!). I would therefore like to explain intuitively the bias-variance analysis that helps us to understand if a model is going to capture the true pattern in the “seen” data and will also generalize well to “unseen” data, thus help us in selecting the best model for the given data.

Before diving into what “bias and variance trade-off” means, let’s get a little background on “statistical learning”. Any supervised statistical learning model will try to find a hypothesis function ħ (i.e. mathematical representation) that approximates the relationship between the predictors (independent variables) and the response (dependent variable). The measure of how well the hypothesis function ħ is fitting the given data can be found by getting the error E between the output of the hypothesis function ħ and the output of the target function ƒ that describes the data. It is called an error because it represents the gap between the hypothesis function and the target function. The smaller the value of the error E, better the learning, meaning that the hypothesis has approximated the target function ƒ well. Therefore, there are two objectives: a) finding a good approximation of target function f, and b) the approximation holding for out-of-sample data. A more complex (meaning bigger) hypothesis set has a better chance of approximating the target function, because it is more likely to hold the target function in the set, but it becomes increasingly harder to find that needle in the haystack! On the other hand, if the hypothesis set is simpler (smaller) then it may not hold the target function in it, but luckily if it does hold the function then it is easier to find. In order to find the best candidate function in the hypothesis set, the hypothesis set must be navigated through the means of the sample data provided, which is the only resource in finding one hypothesis over the other.

Bear with me, I will soon get to an explanation of what I am talking about with an illustration! Just two more paragraphs to go!

Now, getting back to bias and variance, these two entities are inherent properties of a learning model. Mathematically, when the error term E is decomposed, we get bias and variance [2]. Simply put, the trade-off is between approximation and generalization, between bias and variance. The total error term E measures how far the hypothesis function ħ learned from the given data is from the target function ƒ. Of the decomposed entities, bias is a measure of approximation ability, it measures how far the best approximation is from the target function ƒ and the variance is a measure of how far the hypothesis function ħ learned from a dataset is from the best possible candidate function that could be obtained from the hypothesis set H. The hypothesis set H chosen is dependent on the data that is provided, so a different set of data will give a different hypothesis set to choose from. (This dependency is very important in the bias-variance analysis.)

The trade-off is that if bias goes up, then the variance goes down, or if the bias goes down the variance goes up.  If the hypothesis set H gets bigger, the bias gets smaller, getting it closer to f, but then there is a greater variety to choose from for the function which increases the variance. Below is the graph [1] of error E, and the relationship between bias and variance, this relationship is independent of the data, and holds for any statistical learning model. Model complexity is equivalent to hypothesis set size, meaning it holds functions with greater complexity.


Finally, if readers are still with me, here is an example with illustration!

Let me explain this trade-off using an example. Imagine that we have to find a target function ƒ. In real-life the target function is what we find by learning, but here for the purpose of illustration let’s assume we know the target function. Assume the target function ƒ is a sinusoid (displayed with orange line in the following figures for the rest of the article). Our objective is to find the best approximation of the target function ƒ given some data points. Also assume that we are using two hypothesis sets namely H0 and H1, where H0 is the constant model, and H1 is the linear model. Again, assume for illustration purpose that we only have these two hypothesis sets, to keep things simple! Both hypothesis sets will give an approximate function and we compare which one is better using the bias-variance analysis.

We start off with approximation, before doing any learning. The H0 hypothesis set should only give constants, and the H1 hypothesis set only gives lines. In the two graphs below, the light grey shaded regions represent all the possible functions (constants and lines) the H0 and H1 model can generate from the range of data points that are available. The shapes of the grey shaded region are a result of the model complexity and the available data points. The olive lines are the mean of each of the hypothesis sets representing the best of that hypothesis set.

The linear model will choose the function that will get most of the data points it possibly can. And the constant model will be better off choosing zero, as the error will be squared. As expected, we can see from the figures below, (the shaded area showing the errors), that clearly the linear model is the winner, it is a better approximate as the error is the smaller of the two, in fact for the constant model, all of it is an error!

Now, let’s look at the generalizing ability of the two models. Let’s say we have only two example data points (yes, we are stingy!) that we will use to approximate the sinusoid using the H0 and H1 hypothesis sets. The first figure below shows the points. The second shows the points with the target function. The third shows the points fitted with approximate functions from the hypothesis sets that best fits the data points provided.

image 4

Let’s bring back the target function and see how much error we get for the constant and the line. As per expectation the error is smallest for the linear model.

The constant and the line we have here are dependent on the data set, if we had another two points then the approximation of the constant and the line would be different. From a learning perspective, which model is generalizing well? Until now we have been looking at a subset of the dataset that defined the target function (the sinusoid). Now if we look at more datapoints from the population set, we are in for a surprise!

For all the unseen data points stretching infinitely before and after the datapoints we were working with, the constant model at least makes the right predictions periodically, but for the linear model it is a complete disaster! The variance error in the constant model stays constant, whereas the variance error for the linear model keeps increasing with more datapoints. In conclusion, the bias-variance analysis helped us to figure out, that for this particular target function (a sinusoid), given a set of data points and choices between a constant and a linear model we will be better off choosing the constant model that will not be the best approximation of the target but will be the best generalized model. In this case we traded off approximation (bias) for generalization (variance).

All posts are personal reflections of the blog-post author and do not necessarily reflect the views of all other DEEP members.

Reference & Source:

[1] The figure taken from

[2] James, G., Witten, D., Hastie, T. and Tibshirani, R., 2013. An introduction to statistical learning (Vol. 112, p. 18). New York: springer.

[3] All graphs plotted using Matplotlib.Pyplot



By Cristian Montalvo Mancheno, 15th March 11:00 AM

Land-use change is one of the major drivers of biodiversity loss (1). Setting aside protected areas (PAs) has been the main strategy to confront this, but despite the significant increase in PAs over the last 30 years, biodiversity continues to decline globally (2, 3). At the same time, our understanding of PA effectiveness—and the ways we assess it—has evolved. To date, most assessments have only considered impacts at the local scale, but in today’s interconnected world, we must also think about how PAs contribute to net gains or losses in global biodiversity.

The first generation of PA impact assessments simply compared the amount of forest cover, or the number and composition of species, within and outside PA boundaries (e.g., 4, 5), typically finding that biodiversity was higher inside. However, these studies’ findings were swiftly contested due to the inherent location bias of PAs towards areas of low value for agriculture and other human uses. As a result, many areas under legal protection today would have remained undisturbed even in the absence of legal protection. This means that the expansion of PAs has not necessarily led to much additional conservation in terms of representing global biodiversity, and halting land-use change (6).

Furthermore, the simple inside-outside comparisons have often failed to account for land-use dynamics. When a PA is established, it might reduce destructive human activities inside, but if these activities simply move outside, the net effect locally might be reduced or even reversed. This phenomenon is generally known as spillover, that is, the local displacement of land uses to other areas.


These concerns motivated the development of a second generation of studies, in which the non-random allocation of PAs and/or the local spillover effect have explicitly been accounted for (e.g., 7, 8). These studies showed that the effectiveness of PAs not only is lower than those found in inside-outside comparisons, but also varies greatly. Nevertheless, even this last generation of PA impact assessments may be insufficient, because they fail to account for long-distance interactions between land systems, referred as teleconnections (9).


Such teleconnections, for example, may arise when PAs restrict access to resources—limiting the quantity of supply—which shifts the market equilibrium and causes price adjustments and/or the entry of new suppliers across an entire region or the world. This is especially the case in today’s interconnected world where flows of resources, people and capital across large distances affect land-use changes at any spatial scale (1, 9). In fact, teleconnections are more common than many of us would expect.

To illustrate how teleconnections might occur, let us look at the case of the collapse of the Soviet Union in 1991. This shock-like event led to the abandonment of large areas of farmland and war, with large impacts on the environment and biodiversity of many ex-Soviet countries (e.g., 10, 11, 12). In 2016, Schierhorn and colleagues (13) went beyond local impact assessments to illustrate how Russia’s transition from a state-owned to a market-oriented economy caused a long-lasting trade with Brazil, with Brazilian farmers being able to absorb the falling output from Russia due not only to large production potential of beef in the Amazon and Cerrado regions, but also to other enabling factors, such as technological advances in beef production and changes in beef trade flows at global scale. As a result, the gains in forest carbon and biodiversity in one part of the world was in part offset by losses elsewhere.

Teleconnections also play an important role in the effectiveness of policies and interventions aimed at halting biodiversity loss. For example, Ingalls and colleagues (14) investigated the displacement of deforestation under the Reduced Emissions from Deforestation and Forest Degradation (REDD+) framework. They found that the shift from net deforestation to net reforestation in Vietnam occurred through trade of forest-risk commodities[1] with Cambodia and Laos, which was facilitated by Vietnamese companies’ large-scale land acquisitions in these two neighbouring countries. Again, teleconnections meant that one country’s gain was, to some degree, another country’s loss.

While the displacement of land uses over large distances in the Mekong region has diminished the net impact of REDD+ at regional and global scales, teleconnections can also result in net gains for conservation. For example, displacement of agricultural and wood-derived products through trade, in particular to the United States, contributed to Costa Rica’s forest transition. But as Jadin and colleagues (15) showed, such teleconnection had an overall positive impact on the global environment due to striking differences in production and management practices between these two countries and the overall higher biological diversity of Costa Rica compared to the agricultural regions of the United States.

Although challenging, net impact assessments of PAs on the environment across countries, ecoregions and biomes are urgently needed to understand the trade-offs of a future expansion of PAs. This is increasingly relevant today due to the growing interest of conservation scientists and practitioners in setting aside half of the Earth for biodiversity (16, 17). Could such vast expansion bring about perverse outcomes? A recent study showed that the trade-offs between giving back half of the Earth to nature and maintaining food security are large and strongly dependent on the strategy selected and the scales of analyses (18). Therefore, describing coupling of land systems beyond local socio-ecological contexts will be an important next step in in our efforts to effectively conserve global biodiversity within national and regional networks of PAs.

[1] Forest-risk commodities are products that commonly impact forests through their conversion to other land uses or by their degradation, such as timber, semi-processed wood products, mining and agriculture (14).


I would like to thank Linus Blomqvist (Director of Conservation and Food & Agriculture at Breakthrough) for helping me frame and structure this post around my interests in PAs and the teleconnection concept, as well as for his comments and suggestions throughout the development of this blog-post. Also, I am grateful to Carley Fuller (Post-graduate student in the D.E.E.P. Research Group, University of Tasmania) for clarifying the different terminology used in the conservation literature about the spillover effect, and for sharing with me interesting articles around these two types of research projects.


All posts are personal reflections of the blog-post author and do not necessarily reflect the views of all other DEEP members.


  1. Lambin EF, Meyfroidt P. Global land use change, economic globalization, and the looming land scarcity. Proceedings of the National Academy of Sciences. 108(9): p. 3465–3472.
  2. Butchart SHM, Walpole M, Collen B, van Strien A, Scharlemann JPW, Almond REA, et al. Global biodiversity: indicators of recent declines. 2010. 328(5982): p. 1164–1168.
  3. Rodrigues ASL, Brooks TM, Butchart SHM, Chanson J, Cox N, Hoffmann M, et al. Spatially explicit trends in the global conservation status of vertebrates. PLoS ONE. 9(11): p. e113934.
  4. Greve M, Chown SL, van Rensburg BJ, Dallimer M, Gaston KJ. The ecological effectiveness of protected areas: a case study for South African birds. Animal Conservation. 14(3): p. 295–305.
  5. Southworth J, Nagendra H, Carlson LA, Tucker C. Assessing the impact of Celaque National Park on forest fragmentation in western Honduras. Applied Geography. 24(4): p. 303–322.
  6. Blomqvist L, Nordhaus T, Shellenberger M. Nature unbound: Decoupling for conservation. 2015. Available from:
  7. Andam KS, Ferraro PJ, Pfaff A, Sanchez-Azofeifa GA, Robalino JA. Measuring the effectiveness of protected area networks in reducing deforestation. Proceedings of the National Academy of Sciences. 105(42): p. 16089–16094.
  8. Brandt JS, Butsic V, Schwab B, Kuemmerle T, Radeloff VC. The relative effectiveness of protected areas, a logging ban, and sacred areas for old-growth forest protection in southwest China. Biological Conservation. 181: p. 1–8.
  9. Friis C, Nielsen JØ, Otero I, Haberl H, Niewöhner J, Hostert P. From teleconnection to telecoupling: taking stock of an emerging framework in land system science. Journal of Land Use Science. 11(2): p. 131–153.
  10. Baumann M, Radeloff V, Avedian V, Kuemmerle T. Land-use change in the Caucasus during and after the Nagorno-Karabakh conflict. Regional Environmental Change. 15(8): p. 1703–1716.
  11. Hostert P, Kuemmerle T, Prishchepov A, Sieber A, Lambin EF, Radeloff VC. Rapid land use change after socio-economic disturbances: the collapse of the Soviet Union versus Chernobyl. Environmental Research Letters. 6: p. 045201.
  12. Sieber A, Kuemmerle T, Prishchepov AV, Wendland KJ, Baumann M, Radeloff VC, et al. Landsat-based mapping of post-Soviet land-use change to assess the effectiveness of the Oksky and Mordovsky protected areas in European Russia. Remote Sensing of Environment. 133: p. 38–51.
  13. Schierhorn F, Meyfroidt P, Kastner T, Kuemmerle T, Prishchepov AV, Müller D. The dynamics of beef trade between Brazil and Russia and their environmental implications. Global Food Security. 11: p. 84–92.
  14. Ingalls ML, Meyfroidt P, To PX, Kenney-Lazar M, Epprecht M. The transboundary displacement of deforestation under REDD+: problematic intersections between the trade of forest-risk commodities and land grabbing in the Mekong region. Global Environmental Change. 50: p. 255–67.
  15. Jadin I, Meyfroidt P, Lambin EF. International trade, and land use intensification and spatial reorganization explain Costa Rica’s forest transition. Environmental Research Letters. 11(3): p. 035005.
  16. Dinerstein E, Olson D, Joshi A, Vynne C, Burgess ND, Wikramanayake E, et al. An Ecoregion-Based Approach to Protecting Half the Terrestrial Realm. 2017. 67(6): p. 534–45.
  17. Wilson EO. Half-earth: our planet’s fight for life. Liveright, NY: WW Norton & Company. 2016.
  18. Mehrabi Z, Ellis EC, Ramankutty N. The challenge of feeding the world while conserving half the planet. Nature Sustainability. 1(8): p. 409–12.




By Stefania Ondei, 8th January 1:30 pm

Bushfires have been ravaging Tasmania for more than a month. This is not an isolated event, but rather the latest episode in a series of intense fires that have threated the unique biodiversity of the island over the past years. In the summer of 2012-2013, record hot temperatures and dry conditions led to a six-month-long fire season in the south-east of Tasmania; this peaked in January 2013 with the fires on the Forestier and Tasman Peninsula, where several towns were evacuated. 100 homes were lost in the town of Dunalley.

It was unprecedented, the longest fire season ever recorded in Tasmania.

In January 2016, another streak of bushfires hit the island. While this time properties were spared, fires burned over 72,000 hectares in central-western Tasmania, 20,000 of which located in the World Heritage Area. This caused severe damage to the unique sub-alpine and alpine habitats where species such as Pencil Pines, King Billy Pines, and cushion plants are found. The iconic Pencil Pines and King Billy pines (Athrotaxis cupressoides and A. selaginoides) are slow growing trees that can live over 1,000 years, but are restricted to fire-sheltered, high rainfall locations, as they lack those traits typically associated with fire-resistance, such as resprouting ability, thick bark, and aerial or soil seedbanks [1, 2]. The peat on which these plants grow, which can take millennia to form, is also prone to fire damage when dry, releasing great amounts of carbon in the atmosphere [3].

1 pencil pines

Pencil Pines and cushion plants at the Walls of Jerusalem National Park. (Credit: Stefania Ondei)

The 2016 fires reached several of those fire-sensitive habitats in western Tasmania and on the Central Highlands, where the burnt areas might take decades or centuries to recover, if at all.

The damage caused was, once again, described as unprecedented.



Alpine vegetation at lake McKenzie (Tasmanian Central highlands) before (top) and immediately after (bottom) the 2016 fires (Credit:; Dan Broun).

This summer Tasmania is facing a new fire emergency, with hundreds of professional and volunteer firefighters constantly battling the over 50 fires that have been burning across the state, reddening the Tasmanian skies and covering half of the island in a blanket of smoke.

The unique Tasmanian fire sensitive vegetation is at risk again, particularly in the Central Highlands and in the south-west, where the largest remaining forest of King Billy pines is located. To date, the Tasmanian Fire Service estimated that nearly 200,000 hectares have burnt, and, despite the relief brought by some needed rain, the fire season might be far from being over.

The extent of these fires is unprecedented.


The frequency at which high-intensity fires have been recently occurring in Tasmania suggests that we will continue to use the word ‘unprecedented’ for quite some time. At the very least, bushfires of similar extent and intensity will likely become the norm rather than the exception. But what is driving this change? To start a fire, three key elements need to be present: fuel, oxygen, and an ignition source. The air provides plenty of oxygen. As for fuel, in Tasmania an increase in average temperatures, as well as natural and anthropogenic variations of the westerly winds of the Southern Annular Mode, led to drier summers [4, 5], facilitating the accumulation of high amount of flammable plant material. The ignition source is provided predominantly by lightning; rarely recorded before 2000, fires started by dry storms have become more frequent in the Tasmanian Wilderness World Heritage Area, and the area burnt by those fires has also increased [6].

How will fire-sensitive vegetation respond to these new fire regimes? A study of the King Billy pine population that was damaged by an intense fire of the Central Plateau in 1961 showed little to no regeneration of the trees in the burnt area, nor presence of new seedlings [7]. This was likely exacerbated by the post-fire establishment of alpine shrubs in the area, which provide high amounts of fuel load under dry condition, potentially increasing fire frequency and determining a further decline of King Billy pines [8]. The vegetation burnt by the 2016 fires also showed scarce signs of regeneration one year after the event. There is no reason to assume that this gloomy scenario could not happen elsewhere in Tasmania. It is likely only a matter of time, before the increasingly dry conditions and the more frequent, intense, and extensive fires will reach the many – if not all – subalpine and alpine vegetation communities, shrinking beyond recovery the fire refugia in which these iconic species are forced.


Alpine vegetation at lake McKenzie one year after the 2016 fire (Credit:

Only a combination of short- and long-term actions will help preventing this announced ecological tragedy. For instance, every year in Tasmania a substantial amount of prescribed burning is conducted to reduce fuel load, to limit fire spread and intensity in both inhabited and natural areas. However, it is challenging to establish where to burn and what pattern to follow. Different fire behaviour models agree that the most effective approach would likely be similar to the burning practices traditionally conducted by Aboriginal people, with a large overall burnt area made by a high number of small patches, which would create a fine-scale fuel mosaic [9, 10]. Sadly, such approach would be too expensive for the limited resources available. More financially realistic treatments were also modelled, but displayed limited effectiveness [9]. Local efforts, while useful to prevent and fundamental to stop ongoing fires, are thus unlikely to be able to completely avert the increasingly catastrophic fires driven by a climate that is changing even faster than predicted. In an ideal world, strong immediate measures would be taken at a global scale to limit climate change effects. In this world, however, this is yet to happen. In the meanwhile, we thank the rain.




  1. Prior, L.D., B.J. French, and D.M.J.S. Bowman, Effect of experimental fire on seedlings of Australian and Gondwanan trees species from a Tasmanian montane vegetation mosaic. Australian Journal of Botany, 2018. 66(7): p. 511-517.
  2. Worth, J.R.P., et al., Gondwanan conifer clones imperilled by bushfire. Scientific Reports, 2016. 6: p. 33930.
  3. Yu, Z., et al., Global peatland dynamics since the Last Glacial Maximum. Geophysical Research Letters, 2010. 37(13).
  4. Bureau of Meteorology. Trend in mean annual temperature – 1970-2018. 2019; Available from:
  5. Mariani, M. and M.-S. Fletcher, The Southern Annular Mode determines interannual and centennial-scale fire activity in temperate southwest Tasmania, Australia. Geophysical Research Letters, 2016. 43(4): p. 1702-1709.
  6. Styger, J., J. Marsden-Smedley, and J. Kirkpatrick, Changes in Lightning Fire Incidence in the Tasmanian Wilderness World Heritage Area, 1980–2016. Fire, 2018. 1(3).
  7. Holz, A., et al., Effects of high-severity fire drove the population collapse of the subalpine Tasmanian endemic conifer Athrotaxis cupressoides. Global Change Biology, 2015. 21(1): p. 445-458.
  8. Holz, A., et al. High severity fires, positive fire feedbacks and alternative stable states in Athrotaxis rainforest ecosystems in western Tasmania. in AGU Fall Meeting Abstracts. 2016.
  9. Furlaud, J.M., G.J. Williamson, and D.M.J.S. Bowman, Simulating the effectiveness of prescribed burning at altering wildfire behaviour in Tasmania, Australia. International Journal of Wildland Fire, 2018. 27(1): p. 15-28.
  10. King, K.J., et al., The relative importance of fine-scale fuel mosaics on reducing fire risk in south-west Tasmania, Australia. International Journal of Wildland Fire, 2008. 17(3): p. 421-430.



All posts are personal reflections of the blog-post author and do not necessarily reflect the views of all other DEEP members



By Vishesh Leon Diengdoh 14th December at 11:30AM

When I had camped nearby Port Arthur in the Tasman Peninsula, this November, I was woken up at five in the morning by eight or nine Yellow Wattlebirds who kept calling in a sequence and repeating this for at least half an hour. From within my tent, I counted eight, based on the loudness and direction of the call, the ninth call was too far to make out. The same thing happened like clockwork the next morning as well. Its call has been described as a person vomiting and I once heard someone say it sounded like a deranged chicken. Those are accurate descriptions and if you have not heard what a Yellow Wattlebird sounds like you should give it a listen. I’ve noticed that a lot of birds in Australia have a screaming like calls (they sound amazing but not at 5 in the morning) and this leads me to read a book by Tim Low called Where song began. One of the things I learned from the book was that birds scream for nectar present in the flowers. The amount of nectar available to birds in Australia is enough to make nectar feeding birds defend it with loud aggressive calls to assert possession.


Figure: Yellow Wattlebird, Anthochaera paradoxa is Australia’s largest honeyeater and endemic to Tasmania and King Island (Image source:

Honeyeaters are nectarivorous birds of the Meliphagidae family which are distributed from eastern Indonesia to Palau, Hawaiian Islands, New Zealand and Australia. It is one of the largest family in the region with more than 170 species, with nearly 70 species occurring in Australia. Some honeyeaters are brightly coloured with striking patterns, but most are covered in shades of green, brown or grey. A common feature is a naked area on the head and presence of conspicuous clump of feathers which are often yellow or white. Males and females are nearly identical in appearance with a few exceptions (Longmore, 1991).

Honeyeaters have brush-tip tongues with numerous bristles that are long and fine allowing them to collect nectar across large surfaces and tiny fissures on tree branches. These features of the honeyeater’s tongue are different from other nectarivorous birds such as sunbirds and hummingbirds (Paton and Collins, 1989). In addition to nectar, they also consume fruits, berries, insects, manna and lerp (Barker and Vestjens, 1990).


Figure: Gross morphology and appearance in transverse sections of the brush-tipped tongue of the Spiny-cheeked Honeyeater (Acanthagenys rufogulari) (Image source: Paton and Collins, 1989)

Honeyeaters are conspicuous, highly active and aggressive both among themselves and with other species (Longmore, 1991). Although honeyeaters exhibit aggressive behaviour, a level of coexistence can exist resulting in a community of diverse honeyeaters. According to Ford (1979), larger species can easily defend a nectar source (interference competition) while smaller species are more efficient at feeding (exploitation completion). These two types of competitions create a balance to maintain species diversity in an area where nectar abundance varies spatially and temporally.

In Australia, Honeyeaters are among the major flower feeders. The genera most frequently visited are Eucalyptus, Callistemon, Banksia, Grevillea, Adenanthos, Epacris, Astroloma, Amyema, Correa, Xanthorrhoea, Anigozanthos and Eremophila. The genus Eucalyptus, with 74 species is visited by 83 species of birds making it one of the most important genus. The plant-bird relation in Australia exhibits a generalist relationship with birds visiting a range of flowers. This is different from the specific relationship between hummingbirds and plants in tropical America. There is no definitive indication as to why numerous and dominant plant genera in Australia would be pollinated by birds. Compared to insects, birds are active throughout the year and are more reliable when the flowering seasons and climate are erratic. Birds travel further distances than insects improving chances of outcrossing in plants (Ford et al., 1979).

According to the projections by Sekercioglu et al. (2004), 6-14% of all bird’s species will be extinct by 2100 and this will result in the decline of important ecosystem processes such as seed dispersal and pollination. Nectar and fruit-eating birds are expected to have a higher than average extinction which will, in turn, affect populations and communities. This would have significant importance in Australia and other oceanic regions where pollinating birds are higher than other parts of the world.

The Australian Environment Protection and Biodiversity Conservation Act 1999 (EPBC Act) and the International Union for Conservation of Nature (IUCN) red list have designated the Painted Honeyeater, Black-eared Miner and Regent Honeyeater as vulnerable, endangered and critically endangered respectively. The Helmeted Honeyeater has however been listed as critically endangered by the EPBC. The threats to these species include – agriculture & aquaculture, biological resource use, climate change, invasive non-native/alien species, disease, natural system modifications and residential & commercial development.


Figure: Regent Honeyeater, Anthochaera phrygia (Image source:

The state of the four Honeyeaters listed above represent the current situation in Australia. If we assume the projections of Sekercioglu et al. (2004) to take place, it would have serious consequences for the ecosystem. The aim of my PhD is to assess the influence of land-use and land-cover, climate change and other factors on the distribution of different pollinating birds (i.e. honeyeaters: Black-headed Honeyeater, Eastern Spinebill, Crescent Honeyeater, Little Wattlebird, New Holland Honeyeater, Noisy Miner, Strong-billed Honeyeater, Tawny-crowned Honeyeater, Yellow-throated Honeyeater and Yellow Wattlebird) in Tasmania. I’m focusing on land-use and land-cover and climate change since they are considered as the main drivers of global change (Hansen et al., 2001); with land-use change expected to have the largest impact on terrestrial ecosystems and biodiversity in the future, followed by climate change (Sala et al., 2000).


Figure: Crescent Honeyeater (Phylidonyris pyrrhopterus) at Fortescue Bay. It’s a bit blurry as I only had a macro lens with me.



Barker, R. D. & Vestjens, W. J. M. 1990. The food of Australian birds 2. Passerines, CSIRO PUBLISHING.

Ford, H. A. 1979. Interspecific competition in Australian honeyeaters—depletion of common resources. Australian Journal of Ecology, 4, 145-164.

Ford, H. A., Paton, D. C. & Forde, N. 1979. Birds as Pollinators of Australian Plants. New Zealand Journal of Botany, 17, 509-519.

Hansen, A. J., Neilson, R. P., Dale, V. H., Flather, C. H., Iverson, L. R., Currie, D. J., Shafer, S., Cook, R. & Bartlein, P. J. 2001. Global change in forests: responses of species, communities, and biomes: interactions between climate change and land use are projected to cause large shifts in biodiversity. AIBS Bulletin, 51, 765-779.[0765:GCIFRO]2.0.CO;2

Longmore, W. 1991. Honeyeaters & their allies of Australia, Collins/Angus & Robertson.

Paton, D. & Collins, B. 1989. Bills and tongues of nectar‐feeding birds: A review of morphology, function and performance, with intercontinental comparisons. Australian Journal of Ecology, 14, 473-506.

Sala, O. E., Chapin, F. S., Armesto, J. J., Berlow, E., Bloomfield, J., Dirzo, R., Huber-Sanwald, E., Huenneke, L. F., Jackson, R. B. & Kinzig, A. 2000. Global biodiversity scenarios for the year 2100. Science, 287, 1770-1774.

Sekercioglu, C. H., Daily, G. C. & Ehrlich, P. R. 2004. Ecosystem consequences of bird declines. Proceedings of the National Academy of Sciences, 101, 18042-7.


All posts are personal reflections of the blog-post author and do not necessarily reflect the views of all other DEEP members