Australian Grain Research Articles

Molecular characterization of Brassica genebank germplasm confirms taxonomic identity and reveals low levels and source of taxonomic errors

AbstractCrop germplasm conserved in genebanks, are a fundamental resource of genetic diversity for crop improvement activities, underpinning future food security and sustainable agricultural practices. However, taxonomic errors in genebank germplasm (due to misclassification, contamination and poor data collation) restrict the effective use of this material for correct purpose. Earlier studies investigating species genetic diversity using genebank germplasm, have shown varying levels of taxonomic error within the Brassica species. In response to this reported taxonomic error of global collections, together with the availability of a multiplex PCR (MPCR) marker, targeting the specific chromosomes (A, B and C) of the six Brassica species in U's triangle, this study was undertaken to confirm the taxonomic identity of accessions within the Australian Grains Genebank's (AGG) long-term Brassica collection. A total of 5161 accessions were analysed with MPCR for taxonomic identification, of which, 4842 (93.8%) were confirmed to be consistent (correct) with their labelled taxonomy, while the remaining 319 (6.2%) were identified as taxonomically inconsistent (in-error). Through the evaluation of earlier regeneration and original seed of the error accessions with MPCR, we determined that 80.9% of the taxonomic errors were traced back to the original seed, while 19.1% of errors were the result of genebank seed regeneration handling practices. Results from this study directly enhance information of the AGG Brassica collection and shape directions for distribution, acquisition and regeneration practices within the AGG and potentially other global genebanks, which will facilitate in a more effective use of these valuable genetic resources by researchers and breeders.

Herbicide residues in Australian grain cropping soils at sowing and their relevance to crop growth

Herbicides are used extensively in Australian grain cropping systems. Despite occasional observations of herbicide-induced phytotoxicity, there is little information on the persistence and carryover of multiple herbicide classes in cropping soils and the risk to subsequent crops. Two soil surveys were conducted, in 2015 (n = 40) and 2016 (n = 42), across different Australian grain cropping fields prior to sowing of winter crops, and soil samples analysed for herbicide residues (16 analytes in 2015 and 22 analytes in 2016). Samples in 2015 were taken at two depths (0–10 cm and 10-30 cm), whilst samples in 2016 were taken in topsoil (0–10 cm) only, but from two discrete locations in each field. Our research in both years found at least one herbicide (or herbicide metabolite) residue at all sites, with a median of 6 analytes detected in 2015 and 7 analytes detected in 2016. The most frequently detected residues were glyphosate and its primary breakdown product aminomethylphosphonic acid (AMPA), in 87 and 100%, respectively, of topsoil (0–10 cm) samples in 2015, and 67 and 93% of samples in 2016. The median concentration of glyphosate in 2015 was 0.12 mg kg−1, while AMPA was 0.41 mg kg−1. In 2016, median concentrations of glyphosate and AMPA were 0.22 mg kg−1 and 0.31 mg kg−1. Residues of 2,4-dichlorophenoxyacetic acid, trifluralin and diflufenican were also detected in >40% of topsoil samples in both seasons, but with median concentrations of <0.05 mg kg−1. A literature review found limited availability of phytotoxicity thresholds for major grain crops exposed to soilborne herbicide residues. A risk assessment using available thresholds suggested that although up to 29% of fields contained trifluralin residues that could constrain cereal crop growth, and 24% of fields contained residues of phenoxy or sulfonylureas that could affect dicotyledonous crops, the majority of these fields when planted with tolerant crops would be unlikely to be affected by herbicide residues. More work is required to ascertain the spatial distribution, bioavailability and phytotoxicity of residues and residue mixtures to enable a more accurate agronomic risk assessment.

Kernel Ridge Regression Hybrid Method for Wheat Yield Prediction with Satellite-Derived Predictors

Wheat dominates the Australian grain production market and accounts for 10–15% of the world’s 100 million tonnes annual global wheat trade. Accurate wheat yield prediction is critical to satisfying local consumption and increasing exports regionally and globally to meet human food security. This paper incorporates remote satellite-based information in a wheat-growing region in South Australia to estimate the yield by integrating the kernel ridge regression (KRR) method coupled with complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and the grey wolf optimisation (GWO). The hybrid model, ‘GWO-CEEMDAN-KRR,’ employing an initial pool of 23 different satellite-based predictors, is seen to outperform all the benchmark models and all the feature selection (ant colony, atom search, and particle swarm optimisation) methods that are implemented using a set of carefully screened satellite variables and a feature decomposition or CEEMDAN approach. A suite of statistical metrics and infographics comparing the predicted and measured yield shows a model prediction error that can be reduced by ~20% by employing the proposed GWO-CEEMDAN-KRR model. With the metrics verifying the accuracy of simulations, we also show that it is possible to optimise the wheat yield to achieve agricultural profits by quantifying and including the effects of satellite variables on potential yield. With further improvements in the proposed methodology, the GWO-CEEMDAN-KRR model can be adopted in agricultural yield simulation that requires remote sensing data to establish the relationships between crop health, yield, and other productivity features to support precision agriculture.

Environmental effect on temporal patterns in lentil seed quality development

AbstractTo maximize seed longevity, seeds should be harvested at optimal maturity, that is, when seeds have acquired maximum physiological quality before deterioration begins. The aim of this study was to map the variation in temporal patterns of lentil (Lens culinaris Medik.) seed quality development when grown across four regeneration environments, which differ in the level of temperature and humidity control throughout the growing season, at the Australian Grains Genebank. Seeds of two lentil accessions (76080 and 76072) were harvested at different stages throughout development, commencing at 21 d after 50% anthesis until a maximum of 130 d. At each harvest, physiological quality traits, including germinability (fresh and dried seeds) and seed longevity, were determined, as well as seed dry weight and moisture content. Seeds of both accessions, and in all environments, started to accumulate physiological quality early on in development but did not reach their maximum until 3–54 d after mass maturity. The temporal patterns of desiccation tolerance and storage longevity were highly influenced by the environmental conditions during the maturation drying phase, affecting both ‘when’ maximum quality was attained and for how long it was maintained, thereafter. Seeds did not show a typical developmental response, rather variation was observed in seed quality development both between and within accessions grown in the different environments. The poorest storage longevity was seen when seeds of both accessions were grown in the cooler, temperature-controlled glasshouse, and the maximum longevity was observed in the warmer, semi-protected environments of the green and the big igloo for accessions 76080 and 76072, respectively.

Improving early-season wheat yield forecasts driven by probabilistic seasonal climate forecasts

Seasonal climate forecasts (SCF) are evolving rapidly alongside improvements in climate modelling and downscaling research, and have great potential for weather-sensitive sectors, especially agriculture, by reducing weather-related risks and increasing productivity. Skilful yield forecasts at the beginning of, or before, a cropping season can provide farmers and other stakeholders in agribusiness with the necessary information for early planning and actions. Only a few yield forecast studies have a forecast lead time of four months or longer due to the problem complexity. To enable SCFs from Global Climate Models (GCMs) to be used for early-season yield forecasts, this paper uses a statistical downscaling technique, Extended Copula Post-Processing (ECPP) and the Schaake shuffle, to downscale four climate variables to generate weather-like daily data that are suitable for agricultural applications. Climate forecasts drive a process-based crop model APSIM (Agricultural Production Systems sIMulator) to simulate crop forecasts on 50 stations, well-distributed across the Australian grain zone. To focus on yield forecast skills attributable to SCF, we propose best practice management rules to predict water-limited winter wheat yield. Yield forecasts from ECPP have a significant improvement over quantile mapping downscaling and raw SCF from the Australian recent seasonal forecast model ACCESS-S1 in terms of bias, accuracy, reliability, and overall forecast skill. In addition, even at the beginning of a cropping season with a forecast lead time of four or more months, yield forecasts driven by ECPP illustrate higher skill than climatology, a benchmark for yield forecast. Early-season yield forecasts driven by SCFs provides a promising alternative to regression/machine-learning-based forecasts. Performance sensitivity and issues, and gaps on using skilful SCFs to help growers with their farming decision-making are discussed.

Harvest weed seed control: impact on weed management in Australian grain production systems and potential role in global cropping systems

The introduction of harvest weed seed control (HWSC) techniques and associated machinery has enabled the routine use of an alternative weed control technology at a novel weed control timing in global grain cropping fields. Driven by the significant threat of widespread populations of annual ryegrass (Lolium rigidum) with multiple-herbicide resistance, in the 1990s Australian growers and researchers developed techniques to target, at grain harvest, the seed production of annual ryegrass and other important weed species. The HWSC approach to weed management is now routinely used by a majority of Australian grain producers as an integral component of effective weed control programs. Here we detail the development and introduction of current HWSC systems and describe their efficacy in Australian grain production systems. The use of HWSC has likely contributed to lower annual ryegrass population densities and thus mitigated the impacts of herbicide resistance as well as slowing further evolution of resistance. In addition, low weed densities enable the introduction of site-specific weed control technologies and the opportunity to target specific in-crop weeds with non-selective alternative weed control techniques. With an awareness of the evolutionary potential of weed species to adapt to all forms of weed control, there is an understanding that HWSC treatments need to be judiciously used in grain cropping systems to ensure their ongoing efficacy. The successful use of Australian developed HWSC systems has attracted global interest and there is now a considerable international research effort aimed at introducing this alternative weed control approach and timing into the world’s major cropping systems.

Genetic Diversity among Maize (Zea mays L) Genotypes Based on Fodder Yield and Quality Parameters

Maize (Zea mays) is cultivated all over the world including Pakistan for fodder and grain. Genetic diversity and environmental effects greatly affect the fodder yield and quality in maize. Therefore, it is imperative to assess the genetic diversity among 31 maize genotypes for fodder and quality related traits. Genotypes were collected from the Australian Grain Gene Bank and grown under field conditions with three replications following randomized complete block design. The morphological traits such as plant height (cm), leaf area (cm2 ), leaf-stem ratio (wt. basis), stem girth (cm), leaves plant-1 , days to 50% silking, days to 50% silking, leaf moisture (%), dry fodder yield plant-1 (g), green fodder yield plant-1 (g) and quality traits such as crude protein (%), ether extract (%), ash content (%), crude fiber (%) and nitrogen-free extract (%) were recorded. Analysis of variance, biplot analysis and genotypic and phenotypic correlation were performed. Analysis of variance revealed significant differences with days to 50% silking, days to 50% tasseling, plant height, stem girth, leaf area, leaves plant-1, moisture percentage, green and dry fodder yield, crude protein, crude fiber, ether extract, ash content and nitrogen free extract and non-significant differences with leaf stem ratio. At phenotypic and genotypic level, dry fodder yield plant-1 , plant height, stem diameter, leaf moisture %, No. of leaves, days to 50% silking, crude fiber and ether extract revealed significant correlation with fodder yield plant-1 . Biplot analysis based on PCA for different quantitative parameters showed that first two principal components i.e F1, F2 are contributing 23.29% and 14.53 % to the total variations respectively. Based on all the results the best genotypes were DTMA-271, DTMA-15, DTMA-281 and DTMA-295 could be used in breeding programs.

The extent of herbicide resistance in

Context Annual ryegrass (Lolium rigidum) is a major weed of crop production in southern Australia that readily develops resistance to herbicides. Resistance increases both yield losses and control costs associated with this species. Aims This study aimed to gauge the extent and distribution of resistance to herbicides in L. rigidum across south-eastern Australian grain production systems by collecting seed from randomly selected fields. Methods A total of 1441 weed populations were collected through random surveys conducted over 5 years across 13 agricultural regions of four states with these samples then tested for resistance to eight herbicides from six modes of action. Key results Resistance to diclofop-methyl and sulfometuron-methyl was most common, being present in 64% and 63% of populations respectively. Glyphosate resistance was present in 4% of populations collected. Only 15% of populations collected were susceptible to all herbicides tested. Large differences in resistance occurred between the 13 regions surveyed with resistance to diclofop-methyl ranging from 15% to 86% of populations and sulfometuron-methyl from 12% to 96%. Resistance to post-emergent herbicides tended to be higher than pre-emergent herbicides. Multiple resistance was common with 60% of populations collected having resistance to two or more herbicide modes of action. Conclusions There were significant differences in the extent of multiple resistance in L. rigidum populations collected from individual regions suggesting that the rates of resistance evolution have differed between regions. Implications The high incidence of herbicide resistance in L. rigidum populations randomly collected from south-eastern Australian cropping fields highlights the need for the adoption of additional weed control practices to mitigate the impact of this species on grain production systems.

Explainable Machine Learning to Map the Impact of Weather and Soil on Wheat Yield and Revenue Across the Eastern Australian Grain Belt

Explainable Machine Learning to Map the Impact of Weather and Soil on Wheat Yield and Revenue Across the Eastern Australian Grain Belt

Mechanical scarification technique breaks seed coat-mediated dormancy in wild oat (Avena fatua)

AbstractWild oat is a herbicide resistance-prone global weed species that causes significant economic losses in dryland and horticultural agriculture. As a result, there has been a significant research effort to control this species. A major impediment to this research is the seed coat-mediated dormancy of wild oat, which requires a labor-intensive incision or puncturing of the seed coat to initiate seed germination. This study defines the most efficient settings of a mechanical thresher to overcome wild oat seed dormancy and then validates these settings using multiple populations collected from the Western Australian grain belt. We also compare the effects of rapid mechanical scarification and known germination stimulus tactics such as scarification with sulfuric acid (H2SO4), partial endosperm removal, sandpaper scarification of the seed coat, and immersion in sodium nitroprusside (NO donor SNP) solution on wild oat seedling growth rate. Threshing treatment of 1,500 rpm for 5 s provides equivalent germination compared with manually puncturing individual wild oat seeds, with no difference in seedling relative growth rate. The mechanical scarification of seeds using the thresher resulted in greater germination (66%) than H2SO4 scarification (0%), partial endosperm removal (10%), sandpaper seed coat scarification (25%), and exposure to NO donor SNP (34%). This study demonstrates that the physical dormancy of wild oat can be rapidly overcome using a commercially available mechanical thresher.

Mesotrione: a new preemergence herbicide option for wild radish (Raphanus raphanistrum) control in wheat

AbstractWild radish is the most problematic broadleaf weed in Australian grain production. The propensity of wild radish to evolve resistance to herbicides has led to high frequencies of multiple herbicide–resistant populations present in these grain production regions. The objective of this study was to evaluate the potential of mesotrione to selectively control wild radish in wheat. The initial dose response pot trials determined that at the highest mesotrione rate of 50 g ha−1 applied preemergence (PRE) was 30% more effective than when applied postemergence (POST) on wild radish. This same rate of mesotrione applied POST resulted in a 30% reduction in wheat biomass compared to 0% for the PRE application. Subsequent mesotrione PRE dose response trials identified a wheat selective rate range of &gt;100 and &lt;300 g ai ha−1 that provided greater than 85% wild radish control with less than 15% reduction in wheat growth. Field evaluations confirmed the efficacy of mesotrione at 100 to 150 g ai ha−1 in reducing wild radish populations by greater than 85% following PRE application and incorporation by wheat planting. Additionally, these field trials demonstrated the opportunity for season-long control of wild radish when mesotrione applied PRE was followed by bromoxynil applied POST. The sequential PRE application of mesotrione, a herbicide that inhibits p-hydroxyphenylpyruvate dioxygenase, followed by POST application of bromoxynil, a herbicide that inhibits photosystem II, has the potential to provide 100% wild radish control with no effect on wheat growth.

The Remarkable Journey of a Weed: Biology and Management of Annual Ryegrass (Lolium rigidum) in Conservation Cropping Systems of Australia.

Annual ryegrass (Lolium rigidum Gaud.), traditionally utilised as a pasture species, has become the most problematic and difficult-to-control weed across grain production regions in Australia. Annual ryegrass has been favoured by the adoption of conservation tillage systems due to its genetic diversity, prolific seed production, widespread dispersal, flexible germination requirements and competitive growth habit. The widespread evolution of herbicide resistance in annual ryegrass has made its management within these systems extremely difficult. The negative impacts of this weed on grain production systems result in annual revenue losses exceeding $93 million (AUD) for Australian grain growers. No single method of management provides effective and enduring control hence the need of integrated weed management programs is widely accepted and practiced in Australian cropping. Although annual ryegrass is an extensively researched weed, a comprehensive review of the biology and management of this weed in conservation cropping systems has not been conducted. This review presents an up-to-date account of knowledge on the biology, ecology and management of annual ryegrass in an Australian context. This comprehensive account provides pragmatic information for further research and suitable management of annual ryegrass.

Strip spraying delays pyrethroid resistance in the redlegged earth mite, Halotydeus destructor: a novel refuge strategy.

Pesticide resistance has seen control options for the redlegged earth mite (RLEM), Halotydeus destructor, dwindle for Australian grain farmers. The recent discovery of high recessiveness for pyrethroid resistance in RLEM provided an opportunity to examine the feasibility of a refuge strategy to slow the evolution of resistance. Unlike lepidopterous pests in Bt crops, where refuge strategies are routinely practiced, RLEM is a slow-moving pest, which will impact the design of susceptible refuges. Firstly, we confirmed the pyrethroid resistant allele is recessive to the susceptible (wildtype) allele (in terms of resistance) across spatially separated Australian populations. Secondly, we demonstrated that a small, localized resistant mite population can revert to susceptibility at field relevant scales and conditions. Next, we used a simulation modelling approach to design a practical refuge strategy to maintain susceptibility to pyrethroids in populations with a low incidence of resistance. Certain configurations (e.g. a pesticide strip width of 50 m and refuge spacing of 10m) maintained low levels of resistance across a 10-year time horizon, with lower mite abundance and minimal yield loss. A larger refuge proportion did not always delay resistance, and, under certain conditions, increased resistance frequency. Strip spraying to maintain refuges can be readily incorporated into RLEM management programs where sprayer widths in commercial cropping contexts are typically between 20-40 m. A refuge approach to RLEM management that uses strip spraying may enhance long term control options in the absence of new chemical registrations but will now require field validation. © 2021 Society of Chemical Industry.

Impact of Fungicide Application and Host Genotype on Susceptibility of Brassica napus to Sclerotinia Stem Rot across the South-Western Australian Grain Belt: A Genotype × Environment × Management Study

Sclerotinia stem rot (SSR), caused by the necrotroph Sclerotinia sclerotiorum Lib. (de Bary), is a major disease of canola in Australia, greatly reducing yields in high infection years. This study investigated genotype by environment by management interactions at 25 sites across the south-west Australian grainbelt from 2017 to 2020. Up to 10 canola varieties were grown each year with +/− fungicide application at 30% flowering. Disease incidence was low, with less than 20% infection recorded across most sites. Most variation in yield occurred between sites, rather than by management or variety, due to the environmental differences between the sites. Petal assays were found to be a poor indicator of later disease severity, suggesting the winter growing season in south-west Australia does not have reliable conducive conditions for disease development following petal drop in canola. The Additive Main Effects and Multiplicative Interaction model (AMMI) indicated that the open-pollinated varieties were broadly adapted and stable when fungicide was applied but became unstable with no fungicide, indicating SSR has a significant impact on yield when disease incidence is higher. This study highlights that further research is necessary to determine disease thresholds that lead to significant yield loss.

Detect, Consolidate, Delineate: Scalable Mapping of Field Boundaries Using Satellite Images

Digital agriculture services can greatly assist growers to monitor their fields and optimize their use throughout the growing season. Thus, knowing the exact location of fields and their boundaries is a prerequisite. Unlike property boundaries, which are recorded in local council or title records, field boundaries are not historically recorded. As a result, digital services currently ask their users to manually draw their field, which is time-consuming and creates disincentives. Here, we present a generalized method, hereafter referred to as DECODE (DEtect, COnsolidate, and DElinetate), that automatically extracts accurate field boundary data from satellite imagery using deep learning based on spatial, spectral, and temporal cues. We introduce a new convolutional neural network (FracTAL ResUNet) as well as two uncertainty metrics to characterize the confidence of the field detection and field delineation processes. We finally propose a new methodology to compare and summarize field-based accuracy metrics. To demonstrate the performance and scalability of our method, we extracted fields across the Australian grains zone with a pixel-based accuracy of 0.87 and a field-based accuracy of up to 0.88 depending on the metric. We also trained a model on data from South Africa instead of Australia and found it transferred well to unseen Australian landscapes. We conclude that the accuracy, scalability and transferability of DECODE shows that large-scale field boundary extraction based on deep learning has reached operational maturity. This opens the door to new agricultural services that provide routine, near-real time field-based analytics.

Using agricultural metadata: a novel investigation of trends in sowing date in on-farm research trials using the Online Farm Trials database

Background: A growing ability to collect data, together with the development and adoption of the FAIR guiding principles, has increased the amount of data available in many disciplines. This has given rise to an urgent need for robust metadata. Within the Australian grains industry, data from thousands of on-farm research trials (Trial Projects) have been made available via the Online Farm Trials (OFT) website. OFT Trial Project metadata were developed as filters to refine front-end database searches, but could also be used as a dataset to investigate trends in metadata elements. Australian grains crops are being sown earlier, but whether on-farm research trials reflect this change is currently unknown. Methods: We investigated whether OFT Trial Project metadata could be used to detect trends in sowing dates of on-farm crop research trials across Australia, testing the hypothesis that research trials are being sown earlier in line with local farming practices. The investigation included 15 autumn-sown, winter crop species listed in the database, with trial records from 1993 to 2019. Results: Our analyses showed that (i) OFT Trial Project metadata can be used as a dataset to detect trends in sowing date; and (ii) cropping research trials are being sown earlier in Victoria and Western Australia, but no trend exists within the other states. Discussion/Conclusion: Our findings show that OFT Trial Project metadata can be used to detect trends in crop sowing date, suggesting that metadata could also be used to detect trends in other elements such as harvest date. Because OFT is a national database of research trials, further assessment of metadata may uncover important agronomic, cultural or economic trends within or across the Australian cropping regions. New information could then be used to lead practice change and increase productivity within the Australian grains industry.

Using agricultural metadata: a novel investigation of trends in sowing date in on-farm research trials using the Online Farm Trials database

Background: A growing ability to collect data, together with the development and adoption of the FAIR guiding principles, has increased the amount of data available in many disciplines. This has given rise to an urgent need for robust metadata. Within the Australian grains industry, data from thousands of on-farm research trials (Trial Projects) have been made available via the Online Farm Trials (OFT) website. OFT Trial Project metadata were developed as filters to refine front-end database searches, but could also be used as a dataset to investigate trends in metadata elements. Australian grains crops are being sown earlier, but whether on-farm research trials reflect this change is currently unknown. Methods: We investigated whether OFT Trial Project metadata could be used to detect trends in sowing dates of on-farm crop research trials across Australia, testing the hypothesis that research trials are being sown earlier in line with local farming practices. The investigation included 15 autumn-sown, winter crop species listed in the database, with trial records from 1993 to 2019. Results: Our analyses showed that (i) OFT Trial Project metadata can be used as a dataset to detect trends in sowing date; and (ii) cropping research trials are being sown earlier in Victoria and Western Australia, but no trend exists within the other states. Discussion/Conclusion: Our findings show that OFT Trial Project metadata can be used to detect trends in crop sowing date, suggesting that metadata could also be used to detect trends in other elements such as harvest date. Because OFT is a national database of research trials, further assessment of metadata may uncover important agronomic, cultural or economic trends within or across the Australian cropping regions. New information could then be used to lead practice change and increase productivity within the Australian grains industry.

Life stages of the non‐native Ommatoiulus moreleti (Lucas, 1860) (Julida, Julidae) in Australian small grain systems

The Portuguese millipede, Ommatoiulus moreleti (Lucas, 1860), is increasingly a pest of grains crops in Australian agricultural ecosystems. With the rapid uptake of minimum‐tillage practices, habitat quality has been enhanced for several species of soil dwelling invertebrates, in particular O. moreleti. To understand the population dynamics of O. moreleti in Australian grains crops, populations were sampled at multiple sites for more than 2 years using pitfall traps and cardboard roll traps. Specimens were dissected to investigate reproductive status and developmental stage. Millipede trapping rates varied between sites and across the year, and tended to be lower when soil water content (m3/m3) was high. Both sexes were active year round, however females were relatively more abundant when ground temperatures were higher. Males in a copulatory state and females with mature eggs were collected year round although females lacking mature eggs were more common when ground temperatures were higher and the soil was drier, and female egg load was higher under cooler conditions. Females at a stadium 9 or later carried mature eggs and copulatory males tended to be from stadium 8 or later. Different developmental stages could be recovered at all times of the year. These findings indicate persistent populations of O. moreleti across fields throughout the year, with a key breeding period in autumn but likely opportunistic breeding all year. Control of millipedes will need to focus on increasing the impact of biological agents, decreasing suitable habitat for millipede breeding and intensive control at susceptible crop periods.

Minimum temperature mapping augments Australian grain farmers’ knowledge of frost

Economic losses due to crop damage caused by frost in Australia are estimated to be many hundreds of millions of dollars. In broadacre cropping, pre-sowing management to alleviate frost risk, and timely post-frost decisions rely on locally relevant information about the extent and severity of frosts. Yet temperature information currently available to grain farmers often has limited local relevance due to distance from meteorological stations. The availability of minimum temperature (Tmin) maps at an appropriate scale would facilitate improved farmer decision-making in response to frost risk and frost events.This study deployed temperature loggers and utilised Multivariate Adaptive Regression Splines (MARS) modelling to develop maps of Tmin at farm scale (30 × 30 m grid). We use terrain derived variables to generate nightly Tmin maps across a whole farm, based on data from a single on-farm weather station. Based on cross-validation, only elevation and elevation standard deviation were found to be useful predictors. Validation of the model against different years and locations resulted in good predictive RMSE values in the range 0.72 to 1.61°C. Classification accuracy scores (F1) for prediction of temperatures being above or below a 2°C threshold ranged from 83% to 96%.A priority of this work was to understand how the minimum temperature mapping complements farmers’ understanding of frost, and how it would contribute to improved management of frost in their cropping systems. Evaluation of the maps by farmers showed general agreement that the maps complemented local knowledge, with the main interest in the maps being as a guide to know where to start looking for frost damage.We have developed a method to generate Tmin maps based on a data from a single on-site temperature logger combined with terrain data and demonstrated that these maps are appropriately accurate and at a scale that is relevant to farmer management actions.

Machine learning produces higher prediction accuracy than the Jarvis-type model of climatic control on stomatal conductance in a dryland wheat agro-ecosystem

We compared Support Vector Machine (SVM) and Random Forest (RF) machine learning approaches with the widely used Jarvis-type phenomenological model for predicting stomatal conductance (gs) in wheat (Triticum aestivum L.) using historical measurements collected in the Australian Grains Free-Air CO2 Enrichment (AGFACE) facility. The machine learning-based methods produced greater accuracy than the Jarvis-type model in predicting gs from leaf age, atmospheric [CO2], photosynthetically active radiation, vapour pressure deficit, temperature, time of day, and soil water availability (i.e. phenological and environmental variables determining gs). The R2 was 0.76 for the Jarvis-type but 0.92 for SVM and 0.97 for RF machine learning-based models, with a calculated RMSE of 0.292 mol m−2 s−1 in the Jarvis-type compared to 0.129 mol m−2 s−1 in SVM and 0.081 mol m−2 s−1 in RF. The machine learning models, however, needed large datasets for training to achieve statistical significance, and do not offer the same opportunity to provide physiological insights through a statistically testable hypothesis. These results show that using the machine-learning based methods can achieve high prediction accuracy of gs that is especially important when incorporated into larger models, but their ability to extrapolate beyond observed data ranges will need to be assessed before they could be considered in place of the physical model.