Dr Ha Thanh Nguyen1, Dr Roger Lawes1, Dr. Randall Donohue1, Dr. Quanxi Shao1, Dr David Deery1
1CSIRO, Acton, Canberra, Australia
Biography:
I am a researcher in global environmental changes with domain expertise in vegetation dynamics (yield, carbon exchange, phenology, quality and health in response to climate extremes and variability). Across my career which spans the government, industrial and academic sectors, I have proficiency in developing and deploying analytics from datasets of heterogenous size, structure, spatiotemporal resolutions, instrument, and access configurations. I also maintain active engagement in co-designing and delivering decision-ready solutions to help stakeholders’ future proof against climate extremes and variability.
Abstract:
Australia is facing an increase in cold extremes associated with damaging spring frost events. Frost damage in crops has been defined using screen air temperatures from nearby weather stations, which may not represent the crop temperature below the screen height or the specific place of frost occurrence. Remotely sensed Land Surface Temperature (LST) approximates skin temperature over crops and can overcome the afore-mentioned spatial limitation. However, LST retrieval, and hence frost detection from LST, is challenging because current cloud screening algorithms have limited success with night-time, local cloud occurrences. We designed a set of statistical change-point detection methods for the integrated LST time series and air temperature data that could be used to distinguish clear (cloud-free) conditions, which are conducive to frosts, from cloudy conditions, so that true LST values can be reliably acquired at night. The statistical feature of choice, which is a moving hourly variance, reduced the RMSE in the original LST data by a factor of 2. Henceforth, clear/cloud night assessments were generated dynamically to adapt to observed changes in background and/or climatic conditions, facilitating a robust detection and analysis technique for agricultural frosts.