After completing my BSc (geophysics) in 2001, I spent 15 years in industry focusing on groundwater, engineering and environmental projects. Working for initially a startup that became the largest near-surface geophysics business in Australia, I was fortunate to be involved with some of the latest developmnts in the field (4D electrical resitivity, passive seismic ,3D ground penetrating radar).

Over this time, I also started keeping honey bees and this led to starting a PhD (Applied Physics) on remote sensing to predict honey harvests. After receivng my doctorate in 2020, my research has involved applying my spatial and data science skills, along with machine learning and AI, to a range of ecological projects.

Data quality and analytical rigour in ecological science is a key interest of mine, having come from the relatively data-rich field of geophysics originally to ecology where often research is used to make significant environmental decisions off limited studies due to the time and cost of data acquisition.

The increased availability of remote sensing data, with computing power and AI, has improved some aspects of these issues in ecology but can also can lead to results being extrapolated further than is reasonable, but appear valid due to limitations in what remote sensing can actually achieve, and the beguiling ease at which modern analytical methods can generate models. Particularly when unconcsious bias comes into play. Altered baselines, 'significant' results based on insufficient samples and even simple data handling errors are unfortunately all too common.

My rule of thumb: Spend 3-5 times as long trying to prove the result you don't want. If you get a surprising result that validates your bias, make it 5 - 10 times as long! And, most importantly, don't under-estimate the value in trawling through raw data to validate quality and finding fundamental relationships before diving into modelling.