Local-scale spatial modelling for interpolating climatic temperature variables to predict agricultural plant suitability

Abstract

Assessment of local spatial climatic variability is important in the planning of planting locations for horticultural crops. This study investigated three regression-based calibration methods (i.e. traditional versus two optimized methods) to relate short-term 12-month data series from 170 temperature loggers and 4 weather station sites with data series from nearby long-term Australian Bureau of Meteorology climate stations. The techniques trialled to interpolate climatic temperature variables, such as frost risk, growing degree days (GDDs) and chill hours, were regression kriging (RK), regression trees (RTs) and random forests (RFs). All three calibration methods produced accurate results, with the RK-based calibration method delivering the most accurate validation measures: coefficients of determination (R 2 ) of 0.92, 0.97 and 0.95 and root-mean-square errors of 1.30, 0.80 and 1.31°C, for daily minimum, daily maximum and hourly temperatures, respectively. Compared with the traditional method of calibration using direct linear regression between short-term and long-term stations, the RK-based calibration method improved R 2 and reduced root-mean-square error (RMSE) by at least 5 % and 0.47°C for daily minimum temperature, 1 % and 0.23°C for daily maximum temperature and 3 % and 0.33°C for hourly temperature. Spatial modelling indicated insignificant differences between the interpolation methods, with the RK technique tending to be the slightly better method due to the high degree of spatial autocorrelation between logger sites.

Spatial analysis of frost risk to determine viticulture suitability in Tasmania, Australia

“…Ninety-three stations were analysed and temperature data prior to 2014 were considered (i.e. They were programed to record air temperature ( C) at 1.2 m above ground level every 30 min and were strategically deployed around the state from June 2011 using a stratified sampling approach based on clustering of topographic variables, as per Webb et al (2016). Ideally, it is preferable to provide 30-year climatic ranges, which are generally accepted for determining long-term climate trends (Déqué 2007); however, due to the irregularity of existing longterm BoM recordings prior to 1994, it was not possible to produce accurate long-term climate data that best represented temperature in the prior decade.…”

“…The calibration method outlined by Webb et al (2016) was employed to calibrate long-term (BoM) daily temperature recordings to the locations of the short-term sites. The method involved forming a series of daily BoM grids, which were generated from the sites possessing long-term daily minimum temperature data (refer Figure 1) via a regression kriging interpolation technique (Odeh et al 1995).…”

“…The regression kriging method, as advocated by Webb et al (2016), was suitable for producing the BoM grids because the number of long-term (BoM) stations satisfied the requirement of at least 50 sites (57 for this study), which were located at a range of distances from each other to sufficiently model spatial autocorrelation (Webster and Oliver 2007). The covariates used for the spatial modelling included the 3-arc sec digital elevation model (DEM; Gallant et al 2011), the System for Automated Geoscientific Analyses (SAGA) wetness index Antoni c 2009, Grabs et al 2009), a distance to sea index (derived using the Euclidean Distance Spatial Analyst tool in ESRI 2013 ArcGIS Desktop: Release 10.2: Environmental Systems Research Institute, Redlands, CA, USA) and latitude/longitude grids.…”

“…Frost mapping is best done at the mesoclimate scale due to the strong influence of topographic affects. Webb et al (2016) showed that using a dense network of short-term temperature sensors could accurately model frost risk for winegrapes at high resolution (30 m) over a longer period of 20 years. A high resolution study by Madelin and Beltrando (2005) demonstrated how it was possible to accurately interpolate minimum temperature values for nights of clear skies with cool wind conditions to produce a frost hazard map for Champagne vineyards with a spatial resolution of 50 m. The mapping used topographic parameters, such as elevation, slope and aspect, to create regression models linked to temperature.…”

“…For Pinot Noir-the dominant cultivar for vineyards across the state-budburst typically occurs in late September/early October (Smart and Wells 2014). The study adopts a multivariate spatial modelling approach prescribed by Webb et al (2016) to apply frost risk predictions in October across the state of Tasmania by interpolating from 636 short-term temperature logger records, dispersed across the state from 2012, in addition to 57 long-term Australian Bureau of Meteorology (BoM) climate stations. The study adopts a multivariate spatial modelling approach prescribed by Webb et al (2016) to apply frost risk predictions in October across the state of Tasmania by interpolating from 636 short-term temperature logger records, dispersed across the state from 2012, in addition to 57 long-term Australian Bureau of Meteorology (BoM) climate stations.…”