Uncertainty analytics toolbox

Uncertainty symbolTo help land managers make better decisions through novel analytics that assist in quantifying, visualising and communicating uncertainty

‘Uncertainty’ can be described as a situation involving imperfect or unknown information. By definition, making decisions in agriculture and land management about what to do (such as when to plant a crop) involves risk, because some things (such as next month’s rainfall) aren’t known. Land managers make their ‘best guesses’ or use predictions as they look to the future but these will be imperfect, unknown or incomplete.

VizumApp output

With our expertise in core statistics, machine learning, software engineering and agriculture, Digiscape’s Uncertainty project developed a suite of analytics workflows, or a ‘toolbox’, that quantifies, visualises and communicates uncertainty for agricultural problems for decision makers to better understand risk and make more informed decisions.

Named Tykhe after the the Greek goddess of chance, fate and fortune, the toolbox tells us how outputs from models can be judged, where models break down, where more monitoring data needs to be collected or where expert information is needed.

While the quantification of uncertainty in some modelling domains is not new, the development of a toolbox of analytics that incorporates workflows with multiple sources of input at different spatial and temporal resolutions was extremely novel – not to mention challenging.

What’s in our Uncertainty toolbox?

Tools we developed include: 

  • Vizumap: the R package of visualisation approaches for exploring predictions and their uncertainties in spatial data
  • VizumApp: based off the Vizumap R package. You can download the package for running the app locally here (also in the top right hand corner of the app).
  • ST-EDA: space-time explorations of data to assist in identifying important features and space-time interactions
  • Methods for designing experiments that can be used to simulate more structured simulations from complex models to investigate (i) sensitivities in the model, and (ii) emulation.

The tools are a collection of R packages and are open source.

How can I find out more?

Lucchesi LR, Kuhnert PM, Wikle CK (2021) Vizumap: an R package for visualising uncertainty in spatial data. Journal of Open Source Software 6, 2409. https://doi.org/10.21105/joss.02409

Quigley MC, Bennetts LG, Kuhnert PM, Durance P, Lindsay MD, Pembleton KG, Roberts ME, White CJ (2019) The provision and utility of earth science to decision-makers: synthesis and key findings.  Minerva 39, 349-367. https://dx.doi.org/10.1007/s10669-019-09737-z

Quigley MC, Bennetts LG, Durance P, Kuhnert PM, Lindsay MD, Pembleton KG, Roberts ME, White CJ (2019) The provision and utility of earth science to decision-makers: earth science case studies.  Environment Systems and Decisions 39, 307-348. https://dx.doi.org/10.1007/s10669-019-09728-0

Kuhnert PM, Pagendam DE, Bartley R, Gladish DW, Lewis SE, Bainbridge (2018) Making management decisions in the face of uncertainty: a case study using the Burdekin catchment in the Great Barrier Reef. Marine and Fresh Water Research 69, 1187-1200. https://doi.org/10.1071/mf17237

Dr Petra Kuhnert

  • Petra is a research statistician with Data 61, CSIRO's digital powerhouse. She led Digiscape's Uncertainty project.

Lydia Lucchesi

  • Lydia is a PhD Candidate in Computer Science at the Australian National University. Her current research focuses on the visualisation of data quality.

Sam Nelson

  • Sam is a research technician in the Biosecurity Risk team in CSIRO’s Data61. He started as a cadet through the Indigenous Cadetship program in 2017. Sam's work predominantly revolves around visualising data, in the form of web applications.