Trish Nowak

Past research

Bachelor of Science (Hons) in Physical Geography 1^st, Kingston University London

Thesis title: Quantifying magnitude of electrostatic forces on pollen grains in wind pollinated Plantago lanceolata. (93%)

Masters of Science in Oceanography (Physical & Biological) with Merits, University of Southampton

Thesis title: Meteorological controls on primary production in stratifying shelf seas of the global oceans. (70%)

Awards

1^st place, Guy Robinson Dissertation Prize for highest departmental mark, 2018

1^st place, Pete Guest Memorial Prize for highest achievements in Ecology and Conservation, 2018

Current Project

PhD in Environmental Intelligence: AI and Data Science for Sustainable Futures (UKRI CDT studentship)

Thesis title: Wandering Dusts. Using Convolutional Neural Networks to predict atmospheric dust transport, and disentangle curious reasons for its impact on the environment and climate.

My current project focuses on utilising AI methods to gain better understanding of environmental processes. Advanced AI methods offer possibility to predict the transport of atmospheric particles based on observational data only. Such methods offer significant improvement in time and accuracy of predictions as compared to traditional, physics-based models. In particular, my research focuses on building a convolutional neural network model (CNN) capable of predicting the transport of Saharan mineral dust 24h in advance. The second part of my research plans to utilise this model to analyse the impact of Saharan dust on biogeochemical cycles and primary productivity of the central Atlantic Ocean. In the third part the CNN model will be utalised to unvail the biogeochemical cycles and productivity of the Amazon rainforest. For example, the Amazon rainforest's soils are found to be nutrient poor, with no clear sources of essential micro- and macronutrients (phosphorous and iron in particular). Current theory of Saharan mineral dust enriching Amazon rainforest soils is fairly well known. However, annual quantity estimates of transported dust (and thus the nutrients) yield large uncertainties and vary between the studies. The power of AI may provide us with reduced uncertainties, better estimates, and timely predictions. CNN's ability to analyse dozens of parameters over decades of data sets may unveil new or previously overlooked causalities. The possibility of solving a long-standing scientific question is what inspires and drives me throughout my research.

Supervisors

Dr Benno Simmons - Research Fellow and Lecturer in Ecological Data Science

Dr Stefan Siegert - Lecturer in Advanced Statistical Modelling and Applied Data Science

Prof Andy Augousti - Professor of Applied Physics and Instrumentation, Kingston University London