Satellite data is central to so much of modern life, including much of our environmental monitoring programs. From satellites, you can get data about vegetation, water and so much more. In fact, the instruments on board three different satellites – the MODIS, GRACE, and SMOS missions – can give you specific data about water on the surface, water in the uppermost few centimetres of soil, and total water column respectively.

If you want to try and figure out how soil moisture affects plant growth around the world, you can take those three streams of satellite data, feed them into a state-of-the-art water model and run the whole thing on a powerful supercomputer. That’s what researchers from the Fenner School of Environment and Society’s Centre for Water and Landscape Dynamics at The Australian National University have done, using NCI Australia’s integrated supercomputer and data analysis facilities.

Dr Siyuan Tian, the lead researcher of this study, says, “Combining satellite observations with a hydrological model in this way has not been done before, and would have been impossible without the easy access to NCI’s computing and data services. We are actually using today’s water data to help us understand plant growth in the coming months.”

In the past, environmental models were limited by the amount of data they could effectively handle. Now, the challenge is combining disparate datasets in a way that a model can easily understand says Dr Luigi Renzullo, a co-author on the study.

“Another important outcome of the research is the development of mathematical methods, so-called data assimilation methods, that combine the satellite data into the water model in a way that maximise the respective strengths of each data source. This results in improved model predictions that are beyond the accuracy provide by any one source. Access to the NCI computing resources was key to the success of this work, allowing us to experiment with many different formulations of data assimilation methods.”

From there, the model produced a remarkably accurate global forecast of vegetation response to available water, up to three months into the future. For arid landscapes in particular, water availability is the biggest and most important variable to understand. This new capability produces much more detailed environmental data for farmers and land managers to use for important seasonal decision-making.

While the model currently runs at 25-kilometre resolution globally, the plan is to scale it down to a useful land management scale of around 25-metre resolution for Australia. At that level, the data and computational needs of the model increase massively, but the improved detail provides directly useable information straight to the farmers who need it. Especially during a period of drought, knowing how much water is available to your crops and what their growth response might be can help inform a multitude of decisions around planting and harvesting.

The entire research paper is published in Nature Communications. You can read it here:

For more information about the hydrological model used in this research, click here to visit the collaboration’s website:

This research highlight was originally published in the 2018-2019 NCI Annual Report.