|  
Name:
Dr Mike Joy
Position:
Director
Qualifications:
BSc. (Massey); MSc. Hons. (Massey); PhD. (Massey)
Conatct:
M.K.Joy@massey.ac.nz
Personal profile:
A late starter in Academia, Mike commenced full time study
at Massey University Palmerston North in 1995 after 16 years
of working in a range of non academic jobs. Mike completed
a BSc (Ecology and Environmental Science) in 1997; an MSc
with first class honours (Ecology) in 1999, and completed
his PhD thesis in 2003. He was recently appointed as lecturer
in Environmental Science and Ecology at Massey University.
Interests include environmental issues in general especially
sustainable agriculture, alternative house construction, sustainable
energy use and the impact of globalisation on human and animal
diversity. Other interests put on hold for since academic
life began include sailing and boat and house building.
My research is centred on using fish in bioassessment using computer based predictive modelling. I try to combine reality with the virtual by ensuring that I do enough fieldwork to keep me in touch with reality and away from computers as often as possible.
Research:
Mike’s research involves predicting the spatial occurrence
of freshwater biota using physical and chemical habitat descriptors.
The ability to accurately model and predict the biology of
freshwater systems has many potential and realised uses including
bioassessment (the use of the biological components of stream-systems
as an indicator of their ‘health’), predicting
the invasion of exotic species, predicting the impacts of
environmental alteration etc. This ability to make accurate
predictions is imperative given the huge impacts on freshwater
ecosystems both in New Zealand and Globally. The advantage
of predictive ecology over many other types of analysis is
that by testing predictive models with real world known data
their accuracy can be tested and then and only then can the
relationship between the variables and biota be quantified.
 There
are a huge number of techniques available for predictive ecologists
and Mike has made use of a number of these from the traditional
statistical approaches such as discriminant analysis and logistic
regression to artificial neural networks (ANN) and Bayesian
Belief Networks (BBN). The latter approaches come under the
general description of artificial intelligence because they
are based on the architecture of mammalian brains for ANNs
or based human thought processes BBNs.
 An
example of the use of artificial neural networks is the prediction
of fish communities in the Wellington region New Zealand.
This model has been named ‘point click fish’ and
used by staff at the Wellington Regional Council. An example
of a fish map is shown below. The input variables for the
predictions came from the raw GIS variables used for the River
Environment Classification (REC) from NIWA (National Institute
of Water and Atmospheric Research). This map (below) shows
the predicted occurrence of the native redfin bully from the
model mapped onto the stream network using a Global Information
System (GIS). These maps can be made with the click of the
computer mouse and have links to ecological information associated
with the fish species selected. Any of the species can be
selected and mapped using any number of thresholds. These
predictive maps have been produced for a number of New Zealand
regions.
An example of the use of Bayesian Belief Networks is the
model shown here: The data is for Canterbury and the West
Coast regions of the South Island of New Zealand. The Nodes
along the bottom of the figure are the fish species being
predicted at a given site, and the interconnected nodes above
are the GIS variables being used top predict those fish. This
is an exciting new management tool because the numbers in
the GIS nodes can be altered and the predictions are updated
immediately. 
For example for a given site the catchment land use could
be changed from pastoral farming to exotic forest and the
resulting changes in fish communities observed. This enables
resource managers to assess landuse impacts before they happen.
Other predictive bioassessment models have been produced
for other New Zealand regions using invertebrates and fish.
These models are based on the RIVPACs and AUSRIVAS predictive
models used in the United Kingdom and Australia respectively
and are the first applications of predictive bioassessment
in New Zealand.
Another research field for Mike is untangling the relative
importance of biotic (i.e. predation and competition) and
abiotic (physical and chemical) in structuring freshwater
fish communities in New Zealand streams. This analysis involves
the use of null-models and randomisation processes to see
if fish communities are just random entities or have some
sort of structure.
Publications:
|
