Deer impacts in New Zealand

Presented by David Coomes.

The goal is understanding the impact of introduced species.

History

• Introduction in 1860 of red deer from Scotland
• Another 7 other species (including moose!) released in the subsequent 50 years
• Rapid spread of red deer throughout most natural forests. Regional spread of other species.
• 1930s: Recognition that deer were seriously degrading tussock grasslands (both low altitude and alpine). Concern over erosion damage. Population reaches 8 million. First culling.
• 1970s: Highly profitable export industry for vension led to decimation of national deer herd, especially in exposed habitats where they could be picked off from helicopter
• 1990-2002: Less helicopter hunting. Department of Conservation controls deer in certain areas
• 2002: Wild venison banned from export, because bioaccumant toxins found in meat

Ecology

Flora are not evolved with defenses against ruminants.

Red deer feed on < 20% of wood species, and are very consistent in their diet selection. Mostly subcanopy trees of fertile sites

There are important differences between deer. White tail deer feed on tree ferns, sika deer feed on Nothofagus.

Deer eat first the palatable species, then less palatable species, then litter of palatable species (at least the current crop - unknown whether they eat older litter). This is an important distinction with N. American deer - NZ deer are folivores year-round, whereas N. American deer are winter woody browsers - what data there is suggests N. American deer eat very little leaf litter.

Today's forests

• Very very few palatable seedlings
• Some of the palatable species recruit epiphytically
• 20% loss of stems from 2-5 cm size classes over 20 years - however, mountain beech is unpalatable and can probably not be attributed to deer (but the other species can)

Will ecosystems "recover"?

Tussock grasslands - yes - there is already visible recovery.

Forests - more ambiguous. Deer effects probably interact with disturbance. In places where deer have been greatly reduced by hunting, the decimated palatable species don't really recover because the deer stop eating litter and less palatable species to concentrate on palatable species.

Discussion of this talk

Question: How do you explain decrease in stems for unpalatable species? There was a disturbance 50-60 years ago.

Seedlings do exist but they're small - the larger seedlings are missing. There's no canopy or sub-canopy replacement.

Question: Any evidence of satiation following disturbance or in high light? The study sites are in closed canopy - it would be an interesting question to study.

It was shown that after windthrow - debris offer protection for new seedlings. However, when removing competition for yellow birch, moose found and browsed the new seedlings that came up.

Incorporation of deer herbivory in SORTIE

Presented by Chris Tripler

In GMF, 50 cm tall saplings were planted both outside and inside exclosures - inside exclosures grew to 3 m, outside they didn't grow at all.

Herbivory and SORTIE

Deer herbivory is not explicitly in SORTIE - but the growth functions may include saplings that were probably browsed at some point.

Explicit approach to modeling:

• Track deer populations in model
• Spatial variation in browsing - difficult
• Temporal variation in deer population - difficult

Implicit approach to modeling - preferred - model effects only instead of the populations themselves. An example of implicit approach: separate growth into ambient and non-browsed saplings.

Simulation

In an ambient herbivory simulation, the site was dominated by hemlock with decline by red maple, sugar maple, and white ash.

In a deer-free simulation, the site exhibited depressed basal area. American beech dominated, and red maple, sugar maple, and white ash decline earlier than with ambient herbivory.

So modeling suggests that deer have a positive effect on diversity.

Considerations

Under low light, light is a stronger effect on mortality. Under high light, though, there is a strong increase in mortality in browsed saplings.

The species composition of a neighborhood influences which species are browsed - deer are preferentially browsing high-N plants (although they are probably looking for something besides N).

Proposed SORTIE approach

• Determine neighborhood composition around a focal sapling
• Calculate a neighborhood index
• Use index to look up the probability of herbivory
• Random coin flip determines whether herbivory actually occurs
• Run deer herbivory-induced mortality function
• Grow according to "browsed" or "not browsed" function, as appropriate

Discussion of this talk

Question: What about multiple browse incidents? Any sapling would be a candidate for browsing every timestep.

We are taking into account change in probability of browse for a relative spatial location of a palatable species next to unpalatable species, and a bunch of saplings of the same species. Deer can differentiate N consumer status at a fine spatial scale - we could add another term for local saplings to the focal sapling's probability.

May need to take into account amount of plant removed. What do people think? If you knew total consumption and how it breaks down by species, may be able to take it into account. Variation in offtake rates could be critical - 1 cm removed vs. 1 m removed has huge implications. There's a spatial variation aspect - do deer use a large territory? Range is about the size of a plot.

Initial census data shows browsed saplings grow faster - this effect confounds ability to parameterize. Browsed saplings put reserves into roots and then shoot up because they have lots of reserves? There's evidence that native herbivores "garden" the forest - eat fast-growing species on a regular route to keep a sustainable situation.

For both highly palatable and highly unpalatable species, your local neighborhood doesn't matter to your risk of browse. But in the middle, your neighborhood does matter. This has a huge implication for modeling - is risk of browse a simple fixed function independent of neighborhood or not?

Seed and seedling predation by rodents and possum in New Zealand forests

Presented by Deb Wilson.

Expected impacts on trees

Small mammals eat seeds, fruits, flowers, seedlings, and foliage - so recruitment is impacted, and possibly growth and mortality.

Brushtail possums - arboreal folivore, also forage on ground

• May cause adult mortality
• Eat seeds and fruits
• Kill seedlings
• Possum and seedling abundance inversely proportional

Rats

• Eat seeds, fruits, flowers, and other plant parts
• Suppress recruitment of seedlings on islands without ungulates or possums
• Missing size classes in some island forests

Mice

• Eat seeds and green plant parts
• Can eat a very large number of seeds
• Unknown impact on seedling recruitment

Are there beneficial effects (possible seed dispersal)?

• Possums and rats (but not mice) excrete some viable seeds and may act as seed dispersers
• But overall effect is probably negative because they eat eggs and nestlings of native birds which act as seed dispersers

Study

Exclosure study - exclusion of possums and rats have significant effect on number of seedlings (effect different at different sites - didn't catch this well). Effects consistent across tree/shrub species, including highly unpalatable species.

Challenges

• Birds probably excluded too
• As well as deer at one site
• Both of above could create a false "possum effect"
• Leaf litter and light reduced inside exclosures even though it was tipped inside
• Animals might perch on top of exclosures and drop seed inside
• Forest floor patchiness (site location challenge)

Solutions

• Fencing out deer
• Open-topped exclosures
• Very small mesh to exclude house mice
• Careful plot selection
• High replication
• Long-term studies
• Experimentally sow or offer known numbers of seeds or seedlings

Experiment at Waitutu

• Offer seeds in Petri dishes at rodent trap locations
• Four species per trap location
• Checked after 2, 4, 7 days
• Predation = <= 2 seeds left

Results

There were differences between seed species - rimu and beech were preferred by mice. The differences were consistent between sites.

All sites had abundant mice, some rats, possums most abundant at one site.

Next steps

• Test for neighborhood effects based on mapped canopy trees
• Repeat experiment at lower mouse density

Putting these effects into SORTIE

We need to be able to specify level of control of possums and rodents in order to predict forest outcomes. This affects recruitment, growth, and mortality. There's also a need for a spatial component.

We need to be able to model masting.

• How small mammals affect recruitment of masting species
• Also the effect on recruitment of other species
• Need different length timestep to adequately capture masting

Discussion of this talk

Question: Is there any suggestion that the mice are learning where the food is? The rain covers are easy to spot. Mice forage at night by smell, but yes, once they learn where food is they probably come back there.

Question: Do you capture mice who come to eat seeds? No, there are separate trapping studies to determine population density.

Results showed not much difference between medium terraces and alluvials.

They tried to put seeds directly on the ground but they're too small - they disappear - thus the dishes

General herbivory discussion

How to model rodents?

• It's expected that predation has significant effect only when seeds are in short supply - alternative prey species
• There's an indication that the first fruit to fall is more likely to survive since the mouse population is low at that point
• Just getting the relative offtake rates right might be as critical as anything

Over the long term - is the seed production rate constant? If on average there is no variation, why consider masting?

• There could be other effects unrelated to predator satiation that could be important - it's an effect that should be considered
• There are clear cohorts in the woods that are related to masting events - so it's an important effect
• If the rate of predation is influenced by who you're near (oaks can't get toehold in maple stand because their seeds are eaten preferentially) - that's got big implications

Why don't we just simulate the cohorts directly instead of the original seed production? It depends on what question you're trying to answer - if you're interested in predator satiation you'll want the seeds.

What about seed production every year, and recruitment evaluated every five years? From a computing standpoint, has no benefit.

Having a finer temporal resolution is important - if you want to understand interactions between seed predators and plants, you need it - and you need it from a management perspective as well.

How about looking at rodents as a risk factor - treat it as an externality? What about a binary pattern - no mice vs. bunch of mice?

Can you leave out the mice and include them into the mortality rate? Probably okay to do it that way, but you need to easily be able to change the mice risk factor. If you can comfortably predict animal density, we can do it that way. And you can. You could create an embedded separate model in a behavior that runs through a process and reports back.

How to model deer? What about treating deer as an externality vs. driven by the forest? Can't predict deer density but might be able to predict deer use.

Should we be just concentrating on seed survival? Lots of things can affect it. And the same for seedling/sapling survival - there's more factors than just predators. Do we want to get into the individual processes or just think more globally? It depends on whether you're interested in spatial variability. It's important enough so that some form of modeling predation will end up in the model.

In NZ - answering the question - "How low do I have to get population densities to get effect X?" is always the goal. We can answer "some" vs. "none" but not threshold/gradient-type questions. Are there ways to go about this? It's very hard to get data on population densities. So we'll probably look at functional responses of deer but not numerical responses.

We can use the exclosure studies of "no deer" / "all deer" and then move along the scale between them to find the threshold point of deer effect for a desired management goal (instead of deer population) - would that be useful? It's certainly better information than we have now.

Is level of browsedness binary or on a scale? Since deer behavior is so hard to predict - it's maybe not worth getting into the complexity.

Would it be possible to plug in the results of another model which predicts animal populations? Maybe a better way is just vary deer initial conditions to find an optimal level that correctly replicates observations. And that approach might work for insect attack as well.

How uniform are the effects of browsing? You would expect the same level of explained variance in browsed vs. unbrowsed datasets - if you see a difference you can characterize the effect on biomass.