This paper presents an innovative method for tourist market segmentation-based on dominant movement patterns of tourists; that is, the travel sequences or patterns used by tourists most frequently. There were three steps to achieve this goal. In the first step, general log-linear models were adopted to identify the dominant movement patterns, while the second step was to discover the characteristics of the groups of tourists who travelled with these patterns. The Expectation–Maximisation algorithm was then used to partition tourist segments in terms of socio-demographic and travel behavioural variables. The third step was to select target markets based upon the earlier analysis. These methods were applied to a sample of tourists, over the period of a week, on Phillip Island, Victoria, Australia. A significant outcome of this research is that it will assist tourism organisations to identify tourism market segments and develop better tour packages and more efficient marketing strategies aligned to the characteristics of the tourists.

On-farm experimentation (OFE) is a farmer-centric process that can enhance the adoption of digital agriculture technologies and improve farm profitability and sustainability. Farmers work with consultants or researchers to design and implement experiments using their own machinery to test management practices at the field or farm scale. Analysis of data from OFE is challenging because of the large spatial variation influenced by spatial autocorrelation that is not due to the treatment being tested and is often much larger than treatment effects. In addition, the relationship between treatment and yield response may also vary spatially. We investigate the use of geographically weighted regression (GWR) for analysis of data from large on-farm experiments. GWR estimates local regressions, where data are weighted by distance from the site using a distance-decay kernel. It is a simple approach that can be easily explained to farmers and their agronomic advisors. We use simulated data to test the ability of GWR to separate yield variation due to treatment from any underlying spatial variation in yield that is not due to treatment; show that GWR kernel bandwidth can be based on experimental design to accurately separate the underlying spatial variability from treatment effects; and demonstrate a step-wise model selection approach to determine when the response to treatment is global across the experiment or locally varying. We demonstrate our recommended approach on two large-scale experiments conducted on farms in Western Australia to investigate grain yield response to potassium fertiliser. We discuss the implications of our results for routine practical application to OFE and conclude that GWR has potential for wide application in a semi-automated manner to analyse OFE data, improve farm decision-making, and enhance the adoption of digital technologies.

Satellite remote sensing offers a cost-effective means of generating long-term hindcasts of yield that can be used to understand how yield varies in time and space. This study investigated the use of remotely sensed phenology, climate data and machine learning for estimating yield at a resolution suitable for optimising crop management in fields. We used spatially weighted growth curve estimation to identify the timing of phenological events from sequences of Landsat NDVI and derive phenological and seasonal climate metrics. Using data from a 17,000 ha study area, we investigated the relationships between the metrics and yield over 17 years from 2003 to 2019. We compared six statistical and machine learning models for estimating yield: multiple linear regression, mixed effects models, generalised additive models, random forests, support vector regression using radial basis functions and deep learning neural networks. We used a 50-50 train-test split on paddock-years where 50% of paddock-year combinations were randomly selected and used to train each model and the remaining 50% of paddock-years were used to assess the model accuracy. Using only phenological metrics, accuracy was highest using a linear mixed model with a random effect that allowed the relationship between integrated NDVI and yield to vary by year $({\mathrm{R}}^2$ = 0.67, MAE = 0.25 t ha⁻¹, RMSE = 0.33 t ha⁻¹, NRMSE = 0.25). We quantified the improvements in accuracy when seasonal climate metrics were also used as predictors. We identified two optimal models using the combined phenological and seasonal climate metrics: support vector regression and deep learning models (${\mathrm{R}}^2$ = 0.68, MAE = 0.25 t ha⁻¹, RMSE = 0.32 t ha⁻¹, NRMSE = 0.25). While the linear mixed model using only phenological metrics performed similarly to the nonlinear models that are also seasonal climate metrics, the nonlinear models can be more easily generalised to estimate yield in years for which training data are unavailable. We conclude that long-term hindcasts of wheat yield in fields, at 30 m spatial resolution, can be produced using remotely sensed phenology from Landsat NDVI, climate data and machine learning.

This paper presents a method for predicting areas at risk from dryland salinity using information derived from multi-temporal Landsat Thematic Mapper satellite images combined with land form data derived from high-quality digital elevation models. The method aims to reproduce expert opinion about the future extent of salinity by using decision trees to determine the relationship between salinity risk and variables that describe various aspects of the landscape. Feature selection procedures are used to determine the optimal subset of variables for predicting risk areas. Preliminary studies were conducted in five subcatchments and the model was spatially extrapolated over 30 000 km² to produce maps of those areas expected to become saline under current management practices.