Improving Australia’s digital soil map
To improve yield forecasting and digitally-enabled crop management across Australia’s cropping regions through accurate and cost effective soil property characterisation
The challenges with Australia’s soil data
Better soil information is a must for Australia to capture the full potential of emerging digital agriculture technologies. Australia’s broadacre agriculture, rangelands and forests are all limited by the supply of water and nitrogen for plant growth, and so forecasts of productivity require sound information about how much water is stored and how much nitrogen the soil can release.
Direct measurement of soil properties and crop roots is slow and expensive, and soil agencies in Australia have been constrained to sample the landscape much less intensively than in comparable countries. Measurements of the soil’s water-holding capacity are particularly sparse. To compensate for the relative lack of direct measurements, Australia’s soil science community has developed strong expertise in inferring soil information, culminating in the release of the Soil and Landscape Grid of Australia (SLGA).
We used the SLGA to support improved modelling of crops, to inform management decision making and to enable accurate forecasting of yields. We know, though, that for some purposes it is accurate enough (for in-season decision making at the paddock scale, for example).
Improving Australia’s digital soil map
New in-field and remotely-sensed data streams provide an opportunity for a new approach to predicting functional soil properties that can complement the spatial interpolations used to construct the SLGA. In Digiscape, we combined these new data with robust spatial and process-based inverse modelling approaches to enable prediction of functional soil properties. These predictions are used to better constrain simulation modelling with quantified uncertainties.
Our efforts so far focused on prediction of the Plant Available Water Capacity (PAWC) of soils, that is the capacity of soils to store water for crops to use. We developed a process-based inverse modelling approach to predict soil PAWC using the agricultural systems modelling and simulations tool, APSIM. The inverse modelling approach enabled prediction of PAWC at paddock and sub-paddock scales using crop yield, biomass or potentially remotely sensed vegetation indices, together with management information and climate data.
With the inverse modelling approach we’ve successfully predicted soil PAWC in case studies in Western Australia with actual wheat yields over 3-5 years, with an error of less than 25mm. We also predicted in-paddock variations in PAWC using yield maps at a case study site in Victoria, which better reflects within-field soil physicochemical variations than the AWC derived from direct soil measurements.
The approach also showed the potential to predict PAWC using leaf area index and green cover dynamics, implying possible use of remote sensing vegetation dynamics.
In addition, we have completed a spatial modelling approach using machine learning for predicting soil hydraulic properties across space, which incorporates the contributions of bioclimatic variables, topographical indices and remote sensing vegetation indices. The approach enabled us to generate spatial maps of saturation water content (SAT), drained upper limit (DUL), water content at 15 bar suction (LL15), and available water capacity (AWC) of soil down to 2 metres at 90 m spatial resolution across the agricultural regions of Australia.
This project developed and demonstrated cost-effective approaches for predictions of soil PAWC from easily measurable plant properties, together with on-site information and readily available spatial data of climate and landscape. The approach has shown potential to predict soil PAWC using remotely sensed vegetation indices that are readily available at spatial scales. It highlights the importance of crop monitoring data and on-site management information, which together can potentially enable future prediction of soil functional properties at fine spatial scales.
Gladish DW, He D, Wang EL (2021) Pattern analysis of Australia soil profiles for plant available water capacity. Geoderma 391, 114977. doi:10.1016/j.geoderma.2021.114977
He D, Oliver Y, Wang EL (2021) Predicting plant available water holding capacity of soils from crop yield. Plant and Soil 459, 315-328. doi:10.1007/s11104-020-04757-0
He D, Wang EL (2019a) On the relation between soil water holding capacity and dryland crop productivity. Geoderma 353, 11-24. doi:10.1016/j.geoderma.2019.06.022
He D, Wang EL (2019b). The potential of using LAI time series to predict plant available water capacity (PAWC) of soils. In Pratley J (ed.), ‘Cells to Satellites: Proceedings of the 19th Australian Society of Agronomy Conference.’ Wagga Wagga, NSW, 25-29 August 2019. Available at: www.agronomyaustraliaproceedings.org/images/sampledata/2019/2019ASA_He_Di_215.pdf
He D, Wang EL, OliverY (2019). Combining process-based modelling and remotely sensed vegetation dynamics to predict soil plant available water capacity. Proceedings of the International Society for Ecological Modelling Global Conference 2019, Salzburg, Austria, 1-5 October 2019.
Malone BP, Luo ZK, He D, Viscarra Rossel RA, Wang EL (2020) Bioclimatic variables as important spatial predictors of soil hydraulic properties across Australia’s agricultural region. Geoderma Regional 23, e00344. doi:10.1016/j.geodrs.2020.e00344
Somarathna PDSN, Searle R (2019) Mapping available soil water capacity with sparse data – an inverse Bayesian approach. p. 93 in ‘Pedometrics 2019’, Guelph, Canada, 2-6 June 2019. Available at: http://pedometrics.org/wp-content/uploads/2020/06/Abstract- Book_Pedometrics-2019.pdf
Somarathna PDSN, Searle R, Gladish DW (2021). Mapping available soil water capacity in New South Wales, Australia using sparse data – an inverse Bayesian approach. Geoderma Regional 25, e00396. doi:10.1016/j.geodrs.2021.e00396
Wang EL, Smith CJ, Macdonald BCT, Hunt JR, Xing HT, Denmead OT, Zeglin S, Zhao ZG, Isaac P (2018) Making sense of cosmic-ray soil moisture measurements and eddy covariance data with regard to crop water use and field water balance. Agricultural Water Management 204, 271- 280. doi:10.1016/j.agwat.2018.04.017
Wang EL, He D (2019) Connecting process-based modelling with remote sensing and in-situ measurements to better monitor and predict the dynamics of soil and plant properties. Proceedings of the International Society for Ecological Modelling Global Conference 2019, Salzburg, Austria, 1-5 October 2019.
Wang EL, He D, Luo ZK (2019) Predicting soil water holding capacity from climate and crop yield. In: In Pratley J (ed.), ‘Cells to Satellites: Proceedings of the 19th Australian Society of Agronomy Conference.’ Wagga Wagga, NSW, 25-29 August 2019. Available at: http://agronomyaustraliaproceedings.org/images/sampledata/2019/2019ASA_Wang_Enli_2 18.pdf
Wang EL, He D, Zhao ZG, Smith CJ, Macdonald BCT (2020) Using a systems modeling approach to improve soil management and soil quality. Frontiers in Agricultural Science and Engineering 7, 289-295. doi:10.15302/J-FASE-2020337
Dr Enli Wang
- Enli is a Chief Research Scientist and project leader. His recent research focuses on integration of genomic data with physiologically-based crop modelling to enable cross-scale prediction of crop performance, and simulation of carbon and nutrient cycling in soil-plant systems.
- Di is a Research Scientist at CSIRO, based in Canberra. She is the key member in the Digiscape’s work on improving Australia’s digital soil map and developed the inverse modelling approach to predict the soil PAWC.
Dr Dan Gladish
- Dan is a Research Scientist at CSIRO, based in Brisbane. He is the key member in the Digiscape’s work on improving Australia’s digital soil map. Dan conducted the global sensitivity analysis of process-based models and pattern analysis of Australia soil profiles.
- Sanji Pallegedara was the postdoctoral research fellow in the Digiscape. She played a key role in the development of the spatial modelling approach for predicting soil hydraulic properties and soil PAWC.