Inferential, non-parametric statistics to assess quality of probabilistic forecast systems

Many statistical forecast systems are available to interested users. In order to be useful for decision-making, these systems must be based on evidence of underlying mechanisms. Once causal connections between the mechanism and their statistical manifestation have been firmly established, the forecasts must also provide some quantitative evidence of `quality’. However, the quality of statistical climate forecast systems (forecast quality) is an ill-defined and frequently misunderstood property. Often, providers and users of such forecast systems are unclear about what ‘quality’ entails and how to measure it, leading to confusion and misinformation. Here we present a generic framework to quantify aspects of forecast quality using an inferential approach to calculate nominal significance levels (p-values) that can be obtained either by directly applying non-parametric statistical tests such as Kruskal-Wallis (KW) or Kolmogorov-Smirnov (KS) or by using Monte-Carlo methods (in the case of forecast skill scores). Once converted to p-values, these forecast quality measures provide a means to objectively evaluate and compare temporal and spatial patterns of forecast quality across datasets and forecast systems. Our analysis demonstrates the importance of providing p-values rather than adopting some arbitrarily chosen significance levels such as p < 0.05 or p < 0.01, which is still common practice. This is illustrated by applying non-parametric tests (such as KW and KS) and skill scoring methods (LEPS and RPSS) to the 5-phase Southern Oscillation Index classification system using historical rainfall data from Australia, The Republic of South Africa and India. The selection of quality measures is solely based on their common use and does not constitute endorsement. We found that non-parametric statistical tests can be adequate proxies for skill measures such as LEPS or RPSS. The framework can be implemented anywhere, regardless of dataset, forecast system or quality measure. Eventually such inferential evidence should be complimented by descriptive statistical methods in order to fully assist in operational risk management.

Improving management of summer weeds in dryland cropping systems with Cotton

In 2001 a scoping study (phase I) was commissioned to determine and prioritise the weed issues of cropping systems with dryland cotton. The main findings were that the weed flora was diverse, cropping systems complex, and weeds had a major financial and economical impact. Phase II 'Best weed management strategies for dryland cropping systems with cotton' focused on improved management of the key weeds, bladder ketmia, sowthistle, fleabane, barnyard grass and liverseed grass.In Phase III 'Improving management of summer weeds in dryland cropping systems with cotton', more information on the seed-bank dynamics of key weeds was gained in six pot and field studies. The studies found that these characteristics differed between species, and even climate in the case of bladder ketmia. Species such as sowthistle, fleabane and barnyard grass emerged predominately from the surface soil. Sweet summer grass was also in this category but also had a significant proportion emerging from 5 cm depth. Bladder ketmia in central Queensland emerged mainly from the top 2 cm, whereas in southern Queensland it emerged mainly from 5 cm. Liverseed grass had its highest emergence from 5 cm below the surface. In all cases the persistence of seed increased with increasing soil depth. Fleabane was also found to be sensitive to soil type with no seedlings emerging in the self-mulching black vertisol soil. A strategic tillage trial showed that burial of fleabane seed, using a disc or chisel plough, to a depth of greater than 2 cm can significantly reduce subsequent fleabane emergence. In contrast, tillage increased barnyard grass emergence and tended to decrease persistence. This research showed that weed management plans can not be blanketed across all weed species, rather they need to be targeted for each main weed species.This project has also resulted in an increased knowledge of how to manage fleabane from the eight experiments; one in wheat, two in sorghum, one in cotton and three in fallow on double knock. For summer crops, the best option is to apply a highly effective fallow treatment prior to sowing the crops. For winter crops, the strategy is the integration of competitive crops, residual herbicide followed by a knockdown to control survivors. This project explored further the usefulness of the double knock tactic for weed control and preventing seed set. Two field and one pot experiments have shown that this tactic was highly effective for fleabane control. Paraquat products provided good control when followed by glyphosate. When 2, 4-D was added in a tank mix with glyphosate and followed by paraquat products, 99-100% control was achieved in all cases. The ideal follow-up times for paraquat products after glyphosate were 5-7 days. The preferred follow-up times for 2, 4-D after glyphosate were on the same day and one day later. The pot trial, which compared a population from a cropping field with previous glyphosate exposure and a population from a non-cropping area with no previous glyphosate herbicide exposure, showed that the pervious herbicide exposure affected the response of fleabane to herbicidal control measures. The web-based brochure on managing fleabane has been updated.Knowledge on management of summer grasses and safe use of residual herbicides was derived from eight field and pot experiments. Residual grass and broadleaf weed control was excellent with atrazine pre-plant and at-planting treatments, provided rain was received within a short interval after application. Highly effective fallow treatments (cultivation and double knock), not only gave excellent grass control in the fallow, also gave very good control in the following cotton. In the five re-cropping experiments, there were no adverse impacts on cotton from atrazine, metolachlor, metsulfuron and chlorsulfuron residues following use in previous sorghum, wheat and fallows. However, imazapic residues did reduce cotton growth.The development of strategies to reduce the heavy reliance on glyphosate in our cropping systems, and therefore minimise the risk of glyphosate resistance development, was a key factor in the research undertaken. This work included identifying suitable tactics for summer grass control, such as double knock with glyphosate followed by paraquat and tillage. Research on fleabane also concentrated on minimising emergence through tillage, and applying the double knock tactic. Our studies have shown that these strategies can be used to prevent seed set with the goal of driving down the seed bank. Utilisation of the strategies will also reduce the reliance on glyphosate, and therefore reduce the risk of glyphosate resistance developing in our cropping systems.Information from this research, including ecological and management data were collected from an additional eight paddock monitoring sites, was also incorporated into the Weeds CRC seed bank model "Weed Seed Wizard", which will be able to predict the impact of different management options on weed populations in cotton and grain farming systems. Extensive communication activities were undertaken throughout this project to ensure adoption of the new strategies for improved weed management and reduced risk for glyphosate resistance.

Evaluating legume species as alternative trap crops to chickpea for management of Helicoverpa spp. (Lepidoptera: Noctuidae) in central Queensland cotton cropping systems

Mounting levels of insecticide resistance within Australian Helicoverpa spp. populations have resulted in the adoption of non-chemical IPM control practices such as trap cropping with chickpea, Cicer arietinum (L.). However, a new leaf blight disease affecting chickpea in Australia has the potential to limit its use as a trap crop. Therefore this paper evaluates the potential of a variety of winter-active legume crops for use as an alternative spring trap crop to chickpea as part of an effort to improve the area-wide management strategy for Helicoverpa spp. in central Queensland’s cotton production region. The densities of Helicoverpa eggs and larvae were compared over three seasons on replicated plantings of chickpea, Cicer arietinum (L.), field pea Pisum sativum (L), vetch, Vicia sativa (L.) and faba bean, Vicia faba (L.). Of these treatments, field pea was found to harbour the highest densities of eggs. A partial life table study of the fate of eggs oviposited on field pea and chickpea suggested that large proportions of the eggs laid on field pea suffered mortality due to dislodgment from the plants after oviposition. Plantings of field pea as a replacement trap crop for chickpea under commercial conditions confirmed the high level of attractiveness of this crop to ovipositing moths. The use of field pea as a trap crop as part of an areawide management programme for Helicoverpa spp. is discussed.

A survey of management and economic impact of weeds in dryland cotton cropping systems of subtropical Australia

In dryland cotton cropping systems, the main weeds and effectiveness of management practices were identified, and the economic impact of weeds was estimated using information collected in a postal and a field survey of Southern Queensland and northern New South Wales. Forty-eight completed questionnaires were returned, and 32 paddocks were monitored in early and late summer for weed species and density. The main problem weeds were bladder ketmia (Hibiscus trionum), common sowthistle (Sonchus oleraceus), barnyard grasses (Echinochloa spp.), liverseed grass (Urochloa panicoides) and black bindweed (Fallopia convolvulus), but the relative importance of these differed with crops, fallows and crop rotations. The weed flora was diverse with 54 genera identified in the field survey. Control of weed growth in rotational crops and fallows depended largely on herbicides, particularly glyphosate in fallow and atrazine in sorghum, although effective control was not consistently achieved. Weed control in dryland cotton involved numerous combinations of selective herbicides, several non-selective herbicides, inter-row cultivation and some manual chipping. Despite this, residual weeds were found at 38-59% of initial densities in about 3-quarters of the survey paddocks. The on-farm financial costs of weeds ranged from $148 to 224/ha.year depending on the rotation, resulting in an estimated annual economic cost of $19.6 million. The approach of managing weed populations across the whole cropping system needs wider adoption to reduce the weed pressure in dryland cotton and the economic impact of weeds in the long term. Strategies that optimise herbicide performance and minimise return of weed seed to the soil are needed. Data from the surveys provide direction for research to improve weed management in this cropping system. The economic framework provides a valuable measure of evaluating likely future returns from technologies or weed management improvements.

A systems evaluation of high-input management using fortified molasses for beef production in Australia's dry tropics

The potential of beef producers to profitably produce 500-kg steers at 2.5 years of age in northern Australia's dry tropics to meet specifications of high-value markets, using a high-input management (HIM) system was examined. HIM included targeted high levels of fortified molasses supplementation, short seasonal mating and the use of growth promotants. Using herds of 300-400 females plus steer progeny at three sites, HIM was compared at a business level to prevailing best-practice, strategic low-input management (SLIM) in which there is a relatively low usage of energy concentrates to supplement pasture intake. The data presented for each breeding-age cohort within management system at each site includes: annual pregnancy rates (range: 14-99%), time of conception, mortalities (range: 0-10%), progeny losses between confirmed pregnancy and weaning (range: 0-29%), and weaning rates (range: 14-92%) over the 2-year observation. Annual changes in weight and relative net worth were calculated for all breeding and non-breeding cohorts. Reasons for outcomes are discussed. Compared with SLIM herds, both weaning weights and annual growth were >= 30 kg higher, enabling 86-100% of HIM steers to exceed 500 kg at 2.5 years of age. Very few contemporary SLIM steers reached this target. HIM was most profitably applied to steers. Where HIM was able to achieve high pregnancy rates in yearlings, its application was recommended in females. Well managed, appropriate HIM systems increased profits by around $15/adult equivalent at prevailing beef and supplement prices. However, a 20% supplement price rise without a commensurate increase in values for young slaughter steers would generally eliminate this advantage. This study demonstrated the complexity of pro. table application of research outcomes to commercial business, even when component research suggests that specific strategies may increase growth and reproductive efficiency and/or be more pro. table. Because of the higher level of management required, higher costs and returns, and higher susceptibility to market changes and disease, HIM systems should only be applied after SLIM systems are well developed. To increase profitability, any strategy must ultimately either increase steer growth and sale values and/or enable a shift to high pregnancy rates in yearling heifers.

Row spacing and planting density effects on the growth and yield of sugarcane. 1. Responses in fumigated and non-fumigated soil.

It has been reported that high-density planting of sugarcane can improve cane and sugar yield through promoting rapid canopy closure and increasing radiation interception earlier in crop growth. It is widely known that the control of adverse soil biota through fumigation (removes soil biological constraints and improves soil health) can improve cane and sugar yield. Whether the responses to high-density planting and improved soil health are additive or interactive has important implications for the sugarcane production system. Field experiments established at Bundaberg and Mackay, Queensland, Australia, involved all combinations of 2-row spacings (0.5 and 1.5 m), two planting densities (27 000 and 81 000 two-eyed setts/ha), and two soil fumigation treatments (fumigated and non-fumigated). The Bundaberg experiment had two cultivars (Q124, Q155), was fully irrigated, and harvested 15 months after planting. The Mackay experiment had one cultivar (Q117), was grown under rainfed conditions, and harvested 10 months after planting. High-density planting (81 000 setts/ha in 0.5-m rows) did not produce any more cane or sugar yield at harvest than low-density planting (27 000 setts/ha in 1.5-m rows) regardless of location, crop duration (15 v. 10 months), water supply (irrigated v. rainfed), or soil health (fumigated v. non-fumigated). Conversely, soil fumigation generally increased cane and sugar yields regardless of site, row spacing, and planting density. In the Bundaberg experiment there was a large fumigation x cultivar x density interaction (P<0.01). Cultivar Q155 responded positively to higher planting density in non-fumigated soil but not in fumigated soil, while Q124 showed a negative response to higher planting density in non-fumigated soil but no response in fumigated soil. In the Mackay experiment, Q117 showed a non-significant trend of increasing yield in response to increasing planting density in non-fumigated soil, similar to the Q155 response in non-fumigated soil at Bundaberg. The similarity in yield across the range of row spacings and planting densities within experiments was largely due to compensation between stalk number and stalk weight, particularly when fumigation was used to address soil health. Further, the different cultivars (Q124 and Q155 at Bundaberg and Q117 at Mackay) exhibited differing physiological responses to the fumigation, row spacing, and planting density treatments. These included the rate of tiller initiation and subsequent loss, changes in stalk weight, and propensity to lodging. These responses suggest that there may be potential for selecting cultivars suited to different planting configurations.

Weed management in wide-row cropping systems: a review of current practices and risks for Australian farming systems.

Growing agricultural crops in wide row spacings has been widely adopted to conserve water, to control pests and diseases, and to minimise problems associated with sowing into stubble. The development of herbicide resistance combined with the advent of precision agriculture has resulted in a further reason for wide row spacings to be adopted: weed control. Increased row spacing enables two different methods of weed control to be implemented with non-selective chemical and physical control methods utilised in the wide inter-row zone, with or without selective chemicals used on the on-row only. However, continual application of herbicides and tillage on the inter-row zone brings risks of herbicide resistance, species shifts and/or changes in species dominance, crop damage, increased costs, yield losses, and more expensive weed management technology.

Predicting risk and benefit a priori in biological control of invasive plant species: a systems modelling approach

While the method using specialist herbivores in managing invasive plants (classical biological control) is regarded as relatively safe and cost-effective in comparison to other methods of management, the rarity of strict monophagy among insect herbivores illustrates that, like any management option, biological control is not risk-free. The challenge for classical biological control is therefore to predict risks and benefits a priori. In this study we develop a simulation model that may aid in this process. We use this model to predict the risks and benefits of introducing the chrysomelid beetle Charidotis auroguttata to manage the invasive liana Macfadyena unguis-cati in Australia. Preliminary host-specificity testing of this herbivore indicated that there was limited feeding on a non-target plant, although the non-target was only able to sustain some transitions of the life cycle of the herbivore. The model includes herbivore, target and non-target life history and incorporates spillover dynamics of populations of this herbivore from the target to the non-target under a variety of scenarios. Data from studies of this herbivore in the native range and under quarantine were used to parameterize the model and predict the relative risks and benefits of this herbivore when the target and non-target plants co-occur. Key model outputs include population dynamics on target (apparent benefit) and non-target (apparent risk) and fitness consequences to the target (actual benefit) and non-target plant (actual risk) of herbivore damage. The model predicted that risk to the non-target became unacceptable (i.e. significant negative effects on fitness) when the ratio of target to non-target in a given patch ranged from 1:1 to 3:2. By comparing the current known distribution of the non-target and the predicted distribution of the target we were able to identify regions in Australia where the agent may be pose an unacceptable risk. By considering risk and benefit simultaneously, we highlight how such a simulation modelling approach can assist scientists and regulators in making more objective decisions a priori, on the value of releasing specialist herbivores as biological control agents.

Non-destructive prediction of 'Hass' avocado dry matter via FT-NIR spectroscopy.

BACKGROUND: The inability to consistently guarantee internal quality of horticulture produce is of major importance to the primary producer, marketers and ultimately the consumer. Currently, commercial avocado maturity estimation is based on the destructive assessment of percentage dry matter (%DM), and sometimes percentage oil, both of which are highly correlated with maturity. In this study the utility of Fourier transform (FT) near-infrared spectroscopy (NIRS) was investigated for the first time as a non-invasive technique for estimating %DM of whole intact 'Hass' avocado fruit. Partial least squares regression models were developed from the diffuse reflectance spectra to predict %DM, taking into account effects of intra-seasonal variation and orchard conditions. RESULTS: It was found that combining three harvests (early, mid and late) from a single farm in the major production district of central Queensland yielded a predictive model for %DM with a coefficient of determination for the validation set of 0.76 and a root mean square error of prediction of 1.53% for DM in the range 19.4-34.2%. CONCLUSION: The results of the study indicate the potential of FT-NIRS in diffuse reflectance mode to non-invasively predict %DM of whole 'Hass' avocado fruit. When the FT-NIRS system was assessed on whole avocados, the results compared favourably against data from other NIRS systems identified in the literature that have been used in research applications on avocados.

Integrated Cotton Farming Systems for Central Queensland

This project has delivered outcomes that address major agronomic and crop protection issues closely linked to the profitability and sustainability of cotton production enterprises in CQ. From an agronomic perspective, the CQ environment was always though to support economically viable cotton production in a wide sowing window from the middle of September to early January prior to this research. The ideal positioning of Bollgard II varieties in the CQ planting window was, therefore, critical to the future of the local cotton industry because growers needed baseline information to determine how best to take advantage of the higher yield potential offered by the Bt cotton technology, optimise irrigation water use and fibre characteristics. The project’s outputs include a number of key agronomic findings. Over three growing seasons, Bollgard II crop planted in the traditional sowing window from the middle of September to the end of October consistently produced the highest yields. The project delivers a clear and quantitative assessment of the impacts of planting outside the traditional cropping window - a yield penalty of between 1-4 bales/ha for November and December planted cotton. Whilst yield penalties associated with December-planted crops are clearly linked to declining heat units in the second half of the crop and a cool finish, those associated with November-planted cotton are not consistent with the theoretical yield potential for this sowing date. Further research to understand and minimize the physiological constraints on November-planted cotton would give CQ cotton growers far greater flexibility to develop mixed/double/rotation cropping farming systems that are relevant to the rapidly evolving nature of Agricultural production in Australia. The equivalence of cultivar types with clearly distinguishable, genetically based growth habits, demonstrated in this project, gives growers important information for making varietal choices. The entomological outcomes of this project represent strategic and tactical tools that are highly relevant to the viability and profitability of the cotton industry in Australia. The future of the cotton industry is inextricably linked to the survival and efficacy of GM cotton. Research done in the Callide irrigation area demonstrates the unquestionable potential for development of alternative and highly effective resistance management strategies for Bollgard II using novel technologies and strategies based on products such as Magnet®. Magnet® and similar technologies will be increasingly important in strategies to preserve the shelf life and efficacy of current and future generations of GM technology. However, more research will be required to address logistical and operational issues related to these new technologies before they can be fully exploited in commercial production systems. From an economic perspective, SLW is the sleeping giant in terms of insect nemeses of cotton, particularly from the standpoint of climate change and an increasingly warmer production environment. An effective sampling and management strategy for SLW which has been delivered by this project will go a long way towards minimising production costs in an environment characterised by rapidly rising input costs. SLW has the potential to permanently debilitate the national cotton industry by influencing market sentiment and quality perceptions. Field validation of the SLW population sampling models and management options in the Dawson irrigation area cotton and southern Queensland during 2006-07 documents the robustness of the entomological research outcomes achieved through this project.

Sensory Systems in Sawfishes. 2. The Lateral Line.

The lateral line system allows elasmobranchs to detect hydrodynamic movements in their close surroundings. We examined the distribution of pit organs and lateral line canals in 4 species of sawfish (Anoxypristis cuspidata, Pristis microdon, P. clavata and P. zijsron). Pit organs could only be located in A. cuspidata, which possesses elongated pits that are lined by dermal denticles. In all 4 pristid species, the lateral line canals are well developed and were separated into regions of pored and non-pored canals. In all species the tubules that extend from pored canals form extensive networks. In A. cuspidata, P. microdon and P. clavata, the lateral line canals on both the dorsal and ventral surfaces of the rostrum possess extensively branched and pored tubules. Based on this morphological observation, we hypothesized that these 3 species do not use their rostrum to search in the substrate for prey as previously assumed. Other batoids that possess lateral line canals adapted to perceive stimuli produced by infaunal prey possess non-pored lateral line canals, which also prevent the intrusion of substrate particles. However, this hypothesis remains to be tested behaviourally in pristids. Lateral line canals located between the mouth and the nostrils are non-pored in all 4 species of sawfish. Thus this region is hypothesized to perceive stimuli caused by direct contact with prey before ingestion. Lateral line canals that contain neuromasts are longest in P. microdon, but canals containing neuromasts along the rostrum are longest in A. cuspidata.