Automating Crop Damage Assessments with Unmanned Aerial Systems and Machine Learning

Agricultural pests are responsible for millions of dollars in crop losses and management costs every year. In order to implement optimal site-specific treatments and reduce control costs, new methods to accurately monitor and assess pest damage need to be investigated. In this paper we explore the combination of unmanned aerial vehicles (UAV), remote sensing and machine learning techniques as a promising methodology to address this challenge. The deployment of UAVs as a sensor platform is a rapidly growing field of study for biosecurity and precision agriculture applications. In this experiment, a data collection campaign is performed over a sorghum crop severely damaged by white grubs (Coleoptera: Scarabaeidae). The larvae of these scarab beetles feed on the roots of plants, which in turn impairs root exploration of the soil profile. In the field, crop health status could be classified according to three levels: bare soil where plants were decimated, transition zones of reduced plant density and healthy canopy areas. In this study, we describe the UAV platform deployed to collect high-resolution RGB imagery as well as the image processing pipeline implemented to create an orthoimage. An unsupervised machine learning approach is formulated in order to create a meaningful partition of the image into each of the crop levels. The aim of this approach is to simplify the image analysis step by minimizing user input requirements and avoiding the manual data labelling necessary in supervised learning approaches. The implemented algorithm is based on the K-means clustering algorithm. In order to control high-frequency components present in the feature space, a neighbourhood-oriented parameter is introduced by applying Gaussian convolution kernels prior to K-means clustering. The results show the algorithm delivers consistent decision boundaries that classify the field into three clusters, one for each crop health level as shown in Figure 1. The methodology presented in this paper represents a venue for further esearch towards automated crop damage assessments and biosecurity surveillance.

Fusarium wilt of cotton: Population diversity and implications for management

Fusarium wilt of cotton, caused by the fungus Fusarium oxysporum Schlechtend. f. sp. vasinfectum (Atk.) Snyd. & Hans, was first identified in 1892 in cotton growing in sandy acid soils in Alabama (8). Although the disease was soon discovered in other major cotton-producing areas, it did not become global until the end of the next century. After its original discovery, Fusarium wilt of cotton was reported in Egypt (1902) (30), India (1908) (60), Tanzania (1954) (110), California (1959) (33), Sudan (1960) (44), Israel (1970) (27), Brazil (1978) (5), China (1981) (17), and Australia (1993) (56). In addition to a worldwide distribution, Fusarium wilt occurs in all four of the domesticated cottons, Gossypium arboretum L., G. barbadense L., G. herbaceum L., and G. hirsutum L. (4,30). Disease losses in cotton are highly variable within a country or region. In severely infested fields planted with susceptible cultivars, yield losses can be high. In California, complete crop losses in individual fields have been observed (R. M. Davis, unpublished). Disease loss estimates prepared by the National Cotton Disease Council indicate losses of over 109,000 bales (227 kg or 500 lb) in the United States in 2004 (12).

Bromus catharticus striate mosaic virus: a new mastrevirus infecting Bromus catharticus from Australia.

Although monocotyledonous-plant-infecting mastreviruses (in the family Geminiviridae) are known to cause economically significant crop losses in certain areas of the world, in Australia, they pose no obvious threat to agriculture. Consequently, only a few Australian monocot-infecting mastreviruses have been described, and only two have had their genomes fully sequenced. Here, we present the third full-genome sequence of an Australian monocot-infecting mastrevirus from Bromus catharticus belonging to a distinct species, which we have tentatively named Bromus catharticus striate mosaic virus (BCSMV). Although the genome of this new virus shares only 57.7% sequence similarity with that of its nearest known relative, Digitaria didactyla striate mosaic virus (DDSMV; also from Australia), it has features typical of all other known mastrevirus genomes. Phylogenetic analysis showed that both the full genome and each of its probable expressed proteins group with the two other characterised Australian monocot-infecting mastreviruses. Besides the BCSMV genome sequence revealing that Australian monocot-infecting mastrevirus diversity rivals that seen in Africa, it has enabled us, for the first, to time detect evidence of recombination amongst the Australian viruses. Specifically, it appears that DDSMV possesses a short intergenic region sequence that has been recombinationally derived from either BCSMV or a close relative that has not yet been identified.

Controlling Pythium and associated pests in ginger

This project is to identify treatments that ginger growers can use to control two serious soil-borne pathogens that have emerged and threaten the viability of the ginger industry. Pythium myriotylum, responsible for a severe rhizome rot, is the more serious of the two. It was first identified by ginger growers in the 2007/08 growing season, with some producers reporting total crop losses in some blocks. Symphylids are wingless soil-inhabiting arthropods that feed on the ginger plant's root tips and impair the plants´ ability to absorb nutrients, seriously restricting plant growth and development. Damage caused by symphylids to ginger roots is also expected to facilitate entry of Pythium into the plant.

The global distribution of Banana bunchy top virus reveals little evidence for frequent recent, human-mediated long distance dispersal events

Banana bunchy top virus (BBTV; family Nanoviridae, genus Babuvirus) is a multi-component single-stranded DNA virus, which infects banana plants in many regions of the world, often resulting in large-scale crop losses. Weanalyzed 171 banana leaf samples from fourteen countries and recovered, cloned, and sequenced 855 complete BBTV components including ninety-four full genomes. Importantly, full genomes were determined from eight countries, where previously no full genomes were available (Samoa, Burundi, Republic of Congo, Democratic Republic of Congo, Egypt, Indonesia, the Philippines, and the USA [HI]). Accounting for recombination and genome component reassortment, we examined the geographic structuring of global BBTV populations to reveal that BBTV likely originated in Southeast Asia, that the current global hotspots of BBTV diversity are Southeast Asia/Far East and India, and that BBTV populations circulating elsewhere in the world have all potentially originated from infrequent introductions. Most importantly, we find that rather than the current global BBTV distribution being due to increases in human-mediated movements of bananas over the past few decades, it is more consistent with a pattern of infrequent introductions of the virus to different parts of the world over the past 1,000 years.

Natural host range, thrips and seed transmission of distinct Tobacco streak virus strains in Queensland, Australia

Diseases caused by Tobacco streak virus (TSV) have resulted in significant crop losses in sunflower and mung bean crops in Australia. Two genetically distinct strains from central Queensland, TSV-parthenium and TSV-crownbeard, have been previously described. They share only 81% total-genome nucleotide sequence identity and have distinct major alternative hosts, Parthenium hysterophorus (parthenium) and Verbesina encelioides (crownbeard). We developed and used strain-specific multiplex Polymerase chain reactions (PCRs) for the three RNA segments of TSV-parthenium and TSV-crownbeard to accurately characterise the strains naturally infecting 41 hosts species. Hosts included species from 11 plant families, including 12 species endemic to Australia. Results from field surveys and inoculation tests indicate that parthenium is a poor host of TSV-crownbeard. By contrast, crownbeard was both a natural host of, and experimentally infected by TSV-parthenium but this infection combination resulted in non-viable seed. These differences appear to be an effective biological barrier that largely restricts these two TSV strains to their respective major alternative hosts. TSV-crownbeard was seed transmitted from naturally infected crownbeard at a rate of between 5% and 50% and was closely associated with the geographical distribution of crownbeard in central Queensland. TSV-parthenium and TSV-crownbeard were also seed transmitted in experimentally infected ageratum (Ageratum houstonianum) at rates of up to 40% and 27%, respectively. The related subgroup 1 ilarvirus, Ageratum latent virus, was also seed transmitted at a rate of 18% in ageratum which is its major alternative host. Thrips species Frankliniella schultzei and Microcephalothrips abdominalis were commonly found in flowers of TSV-affected crops and nearby weed hosts. Both species readily transmitted TSV-parthenium and TSV-crownbeard. The results are discussed in terms of how two genetically and biologically distinct TSV strains have similar life cycle strategies in the same environment.

Gaseous Nitrogen Losses Following Soil Amendment with Biosolids under Controlled Conditions.

This paper quantifies gaseous N losses due to ammonia volatilisation and denitrification under controlled conditions at 30 degrees C and 75% to 150% of Field Capacity (FC). Biosolids were mixed with two contrasting soils from subtropical Australia at a rate designed to meet crop N requirements for irrigated cotton or maize (i.e., equivalent to 180 kg N ha(-1)). In the first experiment, aerobically (AE) and anaerobically (AN) digested biosolids were mixed into a heavy Vertosol soil and then incubated for 105 days. Ammonia volatilization over 72 days accounted for less than 4% of the applied NH4-N but 24% (AN) to 29% (AE) of the total applied biosolids' N was lost through denitrification in 105 days. In the second experiment AN biosolids with and without added polyacrimide polymer were mixed with either a heavy Vertosol or a lighter Red Ferrosol and then incubated for 98 days. The N loss was higher from the Vertosol with 16-29% of total N applied versus the Red Ferrosol with 7-10% of total N applied, while addition of polymer to the biosolids increased N loss from 7 to 10% and from 16 to 29% in the Red Ferrosol and Vertosol, respectively. A major product from the denitrification process was N-2 gas, accounting for >90% of the emitted N gases from both experiments. Our findings demonstrate that denitrification could be a major pathway of gaseous N losses under warm and moist conditions.

Better integration of crop-legume sequencing under limited water

It is proposed that over 4-5 years of study period, multiple collaborative sites will be established with on-farm cooperators to demonstrate better integration of crop-legume sequencing for improved root growth and functioning under limited water, leading to improved productivity and carbon sequestration, and reduced runoff and deep drainage losses.

Weed growth and crop yield loss in wheat as influenced by row spacing and weed emergence times

Reducing crop row spacing and delaying time of weed emergence may provide crops a competitive edge over weeds. Field experiments were conducted to evaluate the effects of crop row spacing (11, 15, and 23-cm) and weed emergence time (0, 20, 35, 45, 55, and 60 days after wheat emergence; DAWE) on Galium aparine and Lepidium sativum growth and wheat yield losses. Season-long weed-free and crop-free treatments were also established to compare wheat yield and weed growth, respectively. Row spacing and weed emergence time significantly affected the growth of both weed species and wheat grain yields. For both weed species, the maximum plant height, shoot biomass, and seed production were observed in the crop-free plots, and delayed emergence decreased these variables. In weed-crop competition plots, maximum weed growth was observed when weeds emerged simultaneously with the crop in rows spaced 23-cm apart. Less growth of both weed species was observed in narrow row spacing (11-cm) of wheat as compared with wider rows (15 and 23-cm). These weed species produced less than 5 seeds plant-1 in 11-cm wheat rows when they emerged at 60 DAWE. Presence of weeds in the crop especially at early stages was devastating for wheat yields. Therefore, maximum grain yield (4.91tha-1) was recorded in the weed-free treatment at 11-cm row spacing. Delay in time of weed emergence and narrow row spacing reduced weed growth and seed production and enhanced wheat grain yield, suggesting that these strategies could contribute to weed management in wheat.

Effect of sugarcane cropping systems on herbicide losses in surface runoff

Herbicide runoff from cropping fields has been identified as a threat to the Great Barrier Reef ecosystem. A field investigation was carried out to monitor the changes in runoff water quality resulting from four different sugarcane cropping systems that included different herbicides and contrasting tillage and trash management practices. These include (i) Conventional - Tillage (beds and inter-rows) with residual herbicides used; (ii) Improved - only the beds were tilled (zonal) with reduced residual herbicides used; (iii) Aspirational - minimum tillage (one pass of a single tine ripper before planting) with trash mulch, no residual herbicides and a legume intercrop after cane establishment; and (iv) New Farming System (NFS) - minimum tillage as in Aspirational practice with a grain legume rotation and a combination of residual and knockdown herbicides. Results suggest soil and trash management had a larger effect on the herbicide losses in runoff than the physico-chemical properties of herbicides. Improved practices with 30% lower atrazine application rates than used in conventional systems produced reduced runoff volumes by 40% and atrazine loss by 62%. There were a 2-fold variation in atrazine and >10-fold variation in metribuzin loads in runoff water between reduced tillage systems differing in soil disturbance and surface residue cover from the previous rotation crops, despite the same herbicide application rates. The elevated risk of offsite losses from herbicides was illustrated by the high concentrations of diuron (14mugL-1) recorded in runoff that occurred >2.5months after herbicide application in a 1st ratoon crop. A cropping system employing less persistent non-selective herbicides and an inter-row soybean mulch resulted in no residual herbicide contamination in runoff water, but recorded 12.3% lower yield compared to Conventional practice. These findings reveal a trade-off between achieving good water quality with minimal herbicide contamination and maintaining farm profitability with good weed control.

A review of weed management in wheat using crop competition

Wheat occupies a principal place in the diet of humans globally, contributing more to our daily calorie and protein intake than any other crop. For this reason, preventing weed induced yield losses in wheat has high significance for world food sustainability. Herbicides and tillage play an important role in weed control, but their use has often unacceptable consequences for humans and the wider environment. Additionally, the range of herbicides effective on key weeds is dwindling due to the evolution of herbicide resistance. Elevating crop competitiveness against weeds, through a combination of wheat breeding and innovative planting design (planting density, row spacing and orientation), has strong potential to reduce weed-induced yield losses in wheat. The last decade of research has provided a solid foundation for the breeding of weed suppressive wheat cultivars, and continued research in this area should be a focus for the future. In the interim, there is cause for optimism that weeds can be effectively suppressed using existing wheat varieties, through careful cultivar selection and choice of planting design. Further research is required to define the nature of relationships between cultivar traits and competitive planting strategies, across diverse weed flora in multiple countries, sites and seasons. Investment in such innovation promises to produce benefits, not only in terms of sustained wheat yields, but also in terms of human and ecosystem health, through ameliorating chemical and sediment contamination, soil degradation, and CO2 pollution.

Milk Production Responses to Oversowing a Portion of Tropical Grass Pasture with Summer Crop Legumes on Sub-tropical Dairy Farms

Northern Australian dairy farms have a large area of tropical dryland grass pasture available for use as summer pastures. Late summer-autumn in sub-tropical Australia is traditionally a difficult period in which to produce milk because of the decline in both quality and quantity of tropical grasses (Ehrlich et al. 1994). Options to improve autumn feed on dairy farms include introducing forage crops and conservation, increasing concentrate feeding and introducing legumes. Perennial tropical legumes have not been successful at this time of year because of their inability to sustain stocking rates above one cow/ha. This experiment, conducted on farms, was designed to test if annual crop legumes could be successfully oversown into tropical grass areas using minimal till methods to measure the subsequent impact on milk production on farms. Previous experiments using annual legumes in plots at Mutdapilly Research Station had demonstrated yields up to 10 t/ha can be achieved using annual tropical legumes with protein levels as high as 20% in the whole legume plant. Animal production for a consuming world : proceedings of 9th Congress of the Asian-Australasian Association of Animal Production Societies [AAAP] and 23rd Biennial Conference of the Australian Society of Animal Production [ASAP] and 17th Annual Symposium of the University of Sydney, Dairy Research Foundation, [DRF]. 2-7 July 2000, Sydney, Australia.

Phenology of Carpophilus spp. (Coleoptera: Nitidulidae)in stone fruit orchards as determined by pheromone trapping: Implications for prediction of crop damage

Traps baited with synthetic aggregation pheromone and fermenting bread dough were used to monitor seasonal incidence and abundance of the ripening fruit pests, Carpophilus hemipterus (L.), C. mutilatus Erichson and C. davidsoni Dobson in stone fruit orchards in the Leeton district of southern New South Wales during five seasons (1991-96). Adult beetles were trapped from September-May, but abundance varied considerably between years with the amount of rainfall in December-January having a major influence on population size and damage potential during the canning peach harvest (late February-March). Below average rainfall in December-January was associated with mean trap catches of < 10 beetles/trap/week in low dose pheromone traps during the harvest period in 1991/92 and 1993/94 and no reported damage to ripening fruit. Rainfall in December-January 1992/93 was more than double the average and mean trap catches ranged from 8-27 beetles/week during the harvest period with substantial damage to the peach crop. December-January rainfall was also above average in 1994/95 and 1995/96 and means of 50-300 beetles/trap/week were recorded in high dose pheromone traps during harvest periods. Carpophilus spp. caused economic damage to peach crops in both seasons. These data indicate that it may be possible to predict the likelihood of Carpophilus beetle damage to ripening stone fruit in inland areas of southern Australia, by routine pheromone-based monitoring of beetle populations and summer temperatures and rainfall.

Assessing the relationship between shire winter crop yield and seasonal variability of the MODIS NDVI and EVI images

Australian researchers have been developing robust yield estimation models, based mainly on the crop growth response to water availability during the crop season. However, knowledge of spatial distribution of yields within and across the production regions can be improved by the use of remote sensing techniques. Images of Moderate Resolution Imaging Spectroradiometer (MODIS) vegetation indices, available since 1999, have the potential to contribute to crop yield estimation. The objective of this study was to analyse the relationship between winter crop yields and the spectral information available in MODIS vegetation index images at the shire level. The study was carried out in the Jondaryan and Pittsworth shires, Queensland , Australia . Five years (2000 to 2004) of 250m resolution, 16-day composite of MODIS Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) images were used during the winter crop season (April to November). Seasonal variability of the profiles of the vegetation index images for each crop season using different regions of interest (cropping mask) were displayed and analysed. Correlation analysis between wheat and barley yield data and MODIS image values were also conducted. The results showed high seasonal variability in the NDVI and EVI profiles, and the EVI values were consistently lower than those of the NDVI. The highest image values were observed in 2003 (in contrast to 2004), and were associated with rainfall amount and distribution. The seasonal variability of the profiles was similar in both shires, with minimum values in June and maximum values at the end of August. NDVI and EVI images showed sensitivity to seasonal variability of the vegetation and exhibited good association (e.g. r = 0.84, r = 0.77) with winter crop yields.

Modelling Leaf Production and Crop Development in Maize (Zea mays L.) After Tassel Initiation Under Diverse Conditions of Temperature and Photoperiod

Prediction of the initiation, appearance and emergence of leaves is critically important to the success of simulation models of crop canopy development and some aspects of crop ontogeny. Data on leaf number and crop ontogeny were collected on five cultivars of maize differing widely in maturity and genetic background grown under natural and extended photoperiods, and planted on seven sowing dates from October 1993 to March 1994 at Gatton, South-east Queensland. The same temperature coefficients were established for crop ontogeny before silking, and the rates of leaf initiation, leaf tip appearance and full leaf expansion, the base, optimum and maximum temperatures for each being 8, 34 and 40 degrees C. After silking, the base temperature for ontogeny was 0 degrees C, but the optimum and maximum temperatures remained unchanged. The rates of leaf initiation, appearance of leaf tips and full leaf expansion varied in a relatively narrow range across sowing times and photoperiod treatments, with average values of 0.040 leaves (degrees Cd)-1, 0.021 leaves (degrees Cd)-1, and 0.019 leaves (degrees Cd)-1, respectively. The relationships developed in this study provided satisfactory predictions of leaf number and crop ontogeny (tassel initiation to silking, emergence to silking and silking to physiological maturity) when assessed using independent data from Gatton (South eastern Queensland), Katherine and Douglas Daly (Northern Territory), Walkamin (North Queensland) and Kununurra (Western Australia).