Share on Facebook Tweet on Twitter Share on LinkedIn Share by email
Matthew Smith







I am an ecologist working in the Computational Science Lab at Microsoft Research in Cambridge.



I am a research scientist and I aim to develop useful new predictive models of ecological systems to make significant advances in our ability to predict their properties and dynamics in societally important areas. How can we predict how much the climate is going to be influenced by biological feedbacks and vice versa? What crops and crop planting strategies work best where and why, and for how long? When will an emerging disease arrive in an area and what can we do to prevent it arriving or minimise it's effects? These are the sorts of questions I believe we can provide informative and useful future predictions before they become the present using a new generation of data-constrained process based models. My belief is grounded in evidence - time and time again I have seen that by combining sufficiently (but not overly) detailed mathematical abstractions of what we believe is going on in a system (e.g how a plant grows) with some skilfully incorporated data constraints (e.g. through Bayesian parameter estimation) we can obtain predictive models that not only explain the data well, but perform well under extrapolation (as in, if we train to one environment and test on another, the model still works!). The ability to produce useful ecological forecasts is quickly becoming reality!



Like the rest of my group, I believe that in order to advance our ability to do ecological prediction and forecasting, beyond just pursuing academic results and into providing demonstrably useful information for societies, we need new models, new ways to make those models, improved understanding of the systems we're trying to predict, of how predictable they are in the first place (is ecosystem forecasting more like weather forecasting or predicting the phases of the moon?) and of how much information we need to be able to predict them successfully. In order to achieve these improvements we need to ALSO make fundamental advances in computational science: in how we formulate and constrain complex nonlinear stochastic models, in how we share models and data, in using high performance computing and masses of data, in how we make models and their predictions useful as tools and services. Microsoft Research is a brilliant place to investigate these problems because not only does working here enable us to make those computational breakthroughs to allow us to make the ecological breakthroughs, but those breakthroughs (both computational and ecological) can feedback to benefit the company... and of course, hopefully, to benefit human society.



My research is focussed on 3 main areas but I typically get excited about working on anything that involves making and taking models of ecological systems and investigating the extent to which they can be made to be useful:


  • I am  building the next generation of data-constrained terrestrial carbon models: how the terrestrial carbon cycle will respond to and influence climate change is one of the biggest scientific questions of our age - depending on how this plays out could be the difference between snowball and inferno Earth in centuries to come - we just don't know enough.
  • I am improving our ability to predict food and water availability to human societies worldwide through improving our ability to predict agricultural crop dynamics
  • I am facilitating multi-factorial, multi-space, multi-timescale cost benefit analyses of the consequences of different ecosystem use scenarios. E.g. GIVEN how much plants, animals, crops, ecosystem services, roads, protected areas, fire risk, carbon sequestration etc.. there is in a particular area THEN how do we identify at least better than present solutions to managing that area to increase benefits while minimising detrimental consequences. These are really hard, sometimes wicked, environmental problems for which decision makers really need help from ecologists, and their predictive models, to find solutions. 


  • Gian Marco Palamara, University of Zurich, (co-supervisor Owen Petchey) who is applying stochastic models and inference to investigate our ability to predict ecological dynamics in a range of systems, from experimental microcosms to complex food webs.
  • Jelte Mense, University of Edinburgh, (co-supervisor Paul Palmer) who is trying to predict the spatiotemporal dynamics of human populations in different scenarios, from riots to climate change induced conflict and migration
  • Johannes Meyerholt, University of Jena, (co-supervisor Soenke Zahele) who is undertaking a systematic analysis of how nitrogen cycling is modelled in global vegetation models
  • Ludovica Luisa Vissat, University of Edinburgh, (co-supervisor Jane Hillston) who is working on formal language support for ecological modelling  











    Other peer-reviewed publications

    M.Smith, A.White, J.A.Sherratt, S.Telfer, M.Begon, X.Lambin, (2008) Disease effects on reproduction can cause population cycles in seasonal environments. Journal of Animal Ecology, 77(2), 378-389 doi:10.1111/j.1365-2656.2007.01328.x.

    M.J. Smith, J.A.Sherratt (2007) The effects of unequal diffusion coefficients on periodic travelling wave properties in oscillatory reaction diffusion systems. Physica D, 236(2), 90-103, doi:10.1016/j.physd.2007.07.013

    Sherratt, J.A. & Smith M.J. (2008) REVIEW, Periodic Travelling Waves in Cyclic Populations: Field studies and reaction-diffusion models. doi: 10.1098/rsif.2007.1327 Proc. R. Soc. Interface, 5, 483-505.

    M.J. Smith, R.Sibly (2008) Identification of tradeoffs underlying the primary strategies of plants. Available online, Evolutionary Ecology Research , 10(1), 45-60.

    M.J. Smith, J.A.Sherratt, N.J.Armstrong, (2008) The effects of obstacle size on periodic travelling waves in oscillatory reaction-diffusion equations. Proceedings of the Royal Society of London - Series A, 464, 365-390. doi: 10.1098/rspa.2007.0198.

    M.J. Smith, A.White, J.A.Sherratt, X.Lambin, M.Begon, (2006) Delayed Density Dependent Season Length Alone can Lead to Rodent Population Cycles. American Naturalist 167(5), 695-704. doi: 10.1086/503119.

    C.Buckee, K.Koelle, M.J. Mustard (Smith), S.Gupta, (2004). The Effects of Host Contact Network Structure on Pathogen Diversity and Strain Structure. PNAS 101(29), 10839-44. doi:10.1073/pnas.0402000101

    M.J.Aitkenhead, M.J.Mustard (Smith), A.J.S.McDonald, (2004). Using neural networks to predict spatial structure in ecological systems. Ecol. Mod. 179(3), 393-403. doi:10.1016/j.ecolmodel.2004.05.008.

    M.J. Mustard (Smith), D.B.Standing, M.J.Aitkenhead, D.Robinson, A.J.S.McDonald (2003). The emergence of primary strategies in evolving virtual-plant populations. Evol. Ecol. Res. 5, 1067-81. Available online.

    Other Publications

    Smith, M.J., Brodie, C., Kowalczyk, J., Michnowicz, S. & McGough, H.N. (2006). CITES Orchid Checklist Volume 4.

    United Nations (2005). “Endangered Species” stamp series. Contributed text.

    Mustard (Smith), M.J. & Yuzbasioglu, S. (2005). Turkish Delights. Kew Magazine, Spring 2005

    McGough, H.N. Groves, M.G. Sajeva, M. Mustard (Smith), M.J. & Brodie, C (2004). CITES and Succulents, A User’s Guide. Lego Press

    McGough, H.N. Groves, M.G., Mustard (Smith), M.J.‡ & Brodie, C.(2004). CITES and Plants, A User’s Guide. Lego Press

    Williams, C. Davis, K. & Cheyne, P. (with the assistance of Mustard (Smith), M.J. & Brodie, C) (2003). The CBD, for Botanists.