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.

Simulation of the supply capability of sheep on the Mitchell grass downs of north west Queensland

Sheep in western Queensland have been predominantly reared for wool. When wool prices became depressed interest in the sheep meat industry, increased. For north west Queensland producers, opportunities may exist to participate in live sheep and meat export to Asia. A simulation model was developed to determine whether this sheep producing area has the capability to provide sufficient numbers of sheep under variable climatic conditions while sustaining the land resources. Maximum capacity for sustainability of resources (as described by stock numbers) was derived from an in-depth study of the agricultural and pastoral potential of Queensland. Decades of sheep production and climatic data spanning differing seasonal conditions were collated for analysis. A ruminant biology model adapted from Grazplan was used to simulate pregnancy rate. Empirical equations predict mortalities, marking rates, and weight characteristics of sheep of various ages from simple climatic measures, stocking rate and reproductive status. The initial age structure of flocks was determined by running the model for several years with historical climatic conditions. Drought management strategies such as selling a proportion of wethers progressively down to two-tooth and oldest ewes were incorporated. Management decisions such as time of joining, age at which ewes were cast-for-age, wether turn-off age and turning-off rate of lambs vary with geographical area and can be specified at run time. The model is run for sequences of climatic conditions generated stochastically from distributions based on historical climatic data correlated in some instances. The model highlights the difficulties of sustaining a consistent supply of sheep under variable climatic conditions.

Potential Human Exposure to Australian Bat Lyssavirus, Queensland, 1996-1999

Two human deaths caused by Australian bat lyssavirus (ABL) infection have been reported since 1996. Information was obtained from 205 persons (mostly adults from south Brisbane and the South Coast of Queensland), who reported potential ABL exposure to the Brisbane Southside Public Health Unit from November 1,1996, to January 31, 1999. Volunteer animal handlers accounted for 39% of potential exposures, their family members for 12%, professional animal handlers for 14%, community members who intentionally handled bats for 31%, and community members with contacts initiated by bats for 4%. The prevalence of Lyssavirus detected by fluorescent antibody test in 366 sick, injured, or orphaned bats from the area was 6%. Sequelae of exposure, including the requirement for expensive postexposure prophylaxis, may be reduced by educating bat handlers and the public of the risks involved in handling Australian bats.

Response of tropical turfgrasses to recycled water in southern Queensland

The effects of recycled water (effluent) on 8 tropical grasses growing in 100-L bags of sand were studied in Murrumba Downs, just north of Brisbane in southern Queensland (27.4°S, 153.1°E). The species used were: Axonopus compressus (broad-leaf carpetgrass), Cynodon dactylon (bermudagrass 'Winter Green') and C. dactylon x C. transvaalensis hybrid ('Tifgreen'), Digitaria didactyla (Queensland blue couch), Paspalum notatum (bahiagrass '38824'), Stenotaphrum secundatum (buffalograss 'Palmetto'), Eremochloa ophiuroides (centipedegrass 'Centec') and Zoysia japonica (zoysiagrass 'ZT-11'). From May 2002 to June 2003, control plots were irrigated with potable water and fertilised monthly. Plots irrigated with effluent received no fertiliser from May to August 2002 (deficient phase), complete fertilisers at control rates from September to December 2002 (recovery phase) and nitrogen (N) only at control rates from January to June 2003 (supplementary phase). In October 2002, the average shoot weight of plants from the effluent plots was 4% of that from potable plots, with centipedegrass less affected than the other species (relative growth of 20%). Shoot N concentrations declined by 40% in the effluent plots from May to August 2002 (1.8 ± 0.1%) along with phosphorus (P, 0.46 ± 0.02%), potassium (K, 1.6 ± 0.2%), sulfur (S, 0.28 ± 0.02%) and manganese (Mn, 19 ± 2 mg/kg) concentrations. Only the N and Mn concentrations were below the optimum for grasses. The grasses grew satisfactorily when irrigated with effluent if it was supplemented with N. Between January and June 2003 the average weight of shoots from the effluent plots was 116% of the weight of shoots from the control plots. Shoot nutrient concentrations were also similar in the 2 regimes at this time. The recycled water supplied 23% of the N required for maximum shoot growth, 80-100% of the P and K, and 500-880% of the S, calcium and magnesium. The use of recycled water represents savings in irrigation and fertiliser costs, and reductions in the discharge of N and P to local waterways. Effluent is currently about 50% of the cost of potable water with a saving of about AU$8000/ha.year for a typical sporting field.

Nutrient uptake in tropical turfgrasses growing in winter in southern Queensland

The effects of fertilisers on 8 tropical turfgrasses growing in 100-L bags of sand were studied over winter in Murrumba Downs, just north of Brisbane in southern Queensland (latitude 27.4°S, longitude 153.1°E). The species used were: Axonopus compressus (broad-leaf carpetgrass), Cynodon dactylon (bermudagrass 'Winter Green') and C. dactylon x C. transvaalensis hybrid ('Tifgreen'), Digitaria didactyla (Queensland blue couch), Paspalum notatum (bahiagrass '38824'), Stenotaphrum secundatum (buffalograss 'Palmetto'), Eremochloa ophiuroides (centipedegrass 'Centec') and Zoysia japonica (zoysiagrass 'ZT-11'). Control plots were fertilised with complete fertilisers every month from May to September (72 kg N/ha, 31 kg P/ha, 84 kg K/ha, 48 kg S/ha, 30 kg Ca/ha and 7.2 kg Mg/ha), and unfertilised plots received no fertiliser. Carpetgrass and standard bermudagrass were the most sensitive species to nutrient supply, with lower shoot dry weights in the unfertilised plots (shoots mowed to thatch level) compared with the fertilised plots in June. There were lower shoot dry weights in the unfertilised plots in July for all species, except for buffalograss, centipedegrass and zoysiagrass, and lower shoot dry weights in the unfertilised plots in August for all species, except for centipedegrass. At the end of the experiment in September, unfertilised plots were 11% of the shoot dry weights of fertilised plots, with all species affected. Mean shoot nitrogen concentrations fell from 3.2 to 1.7% in the unfertilised plots from May to August, below the sufficiency range for turfgrasses (2.8-3.5%). There were also declines in P (0.45-0.36%), K (2.4-1.5%), S (0.35-0.25%), Mg (0.24-0.18%) and B (9-6 mg/kg), which were all in the sufficiency range. The shoots in the control plots took up the following levels (kg/ha.month) of nutrients: N, 10.0-27.0; P, 1.6-4.0; K, 8.2-19.8; S, 1.0-4.2; Ca, 1.1-3.3; and Mg, 0.8-2.2, compared with applications (kg/ha.month) of: N, 72; P, 31; K, 84; S, 48; Ca, 30; and Mg, 7.2, indicating a recovery of 14-38% for N, 5-13% for P, 10-24% for K, 2-9% for S, 4-11% for Ca and 11-30% for Mg. These results suggest that buffalograss, centipedegrass and zoysiagrass are less sensitive to low nutrient supply than carpetgrass, bermudagrass, blue couch and bahiagrass. Data on nutrient uptake showed that the less sensitive species required only half or less of the nitrogen required to maintain the growth of the other grasses, indicating potential savings for turf managers in fertiliser costs and the environment in terms of nutrients entering waterways.

Sustainable native forest management. Case studies in managing private native forest in southeast Queensland.

The introduction describes productive forest in Queensland and summaries the principles of native forest management that achieve optimum productivity. Case study 1 deals with thinning an even-aged regrowth forest. It shows how thinning the stand actively manages the future composition and structure to improve productivity in the best stems and increase the commercial value of the next harvest. Case study 2 describes restoring productivity in a high-graded spotted gum - ironbark forest. It shows that defective and non-saleable trees should be removed so they do not repress the future stand; and that regeneration should be thinned, retaining the best trees in adequate growing space. Case study 3 discusses on-farm value adding for hardwood forests. It shows how long-term viability and maximum productivity and returns depend on the best management practices and knowing how to obtain the best returns from a range of forest products. Case study 4 examines integrated harvesting in a eucalypt forest. It shows how integrating the harvest enables the full range of timber products are harvested and sold for their maximum value while reducing the amount of waste.

Queensland Coastal Wetland Resources: the Northern Territory Border to Flinders River

Protection of coastal wetland environments is an important prerequisite to effective and sustainable fisheries management and conservation of habitats for the use of future generations. Mangroves, saltmarshes and seagrasses directly support local and offshore fisheries through the provision of food, shelter, breeding and nursery grounds. As such, these vegetated wetland environments along with sandbars, mudflats, rocky foreshores and reefs have significant economic value as well as their intrinsic aesthetic and ecological values. This report summarises the results of the mapping undertaken in the Gulf of Carpentaria Region from the Queensland/Northern Territory border eastwards to the western bank of the Flinders River (hereafter called the Gulf Study Area). The study was undertaken in order to: 1. document and map coastal wetlands of the Gulf Study Area; 2. document levels of existing disturbance to and protection of these wetlands; 3. examine existing recreational, indigenous and commercial fisheries of the region; 4. evaluate the conservation values of the areas investigated from the viewpoint of fisheries productivity and as habitat for important and/or threatened species for future FHA/Marine Protected Area (MPA) declaration. Dataset URL Link: Queensland Coastal Wetlands Resources Mapping data. [Dataset]

Queensland Coastal Wetland Resources: Sand Bay to Keppel Bay.

Protection of coastal wetland environments is an important prerequisite to effective and sustainable inshore fisheries management and conservation of habitats for use by future generations. Mangroves, saltmarshes, seagrasses and non vegetated habitats directly support local and regional inshore and offshore fisheries through the provision of food, shelter, breeding and nursery grounds. As such, these wetland environments have significant economic value as well as their intrinsic aesthetic and ecological values. This report summarises the results of the mapping undertaken in the Central Queensland Coast from Sand Bay to Keppel Bay (hereafter referred to as the Study Area). The study was undertaken in order to: 1. document and map the coastal wetland communities along the Queensland coastline from Sand Bay (20.93°S, 149.04°E) to Keppel Bay (23.65°S, 151.07°E); 2. document levels of existing disturbance to and protection of the wetlands; 3. examine existing recreational and commercial fisheries in the region; and 4. evaluate the conservation values of the areas investigated from the viewpoint of fisheries productivity and as habitat for important and/or threatened species. Dataset URL Link: Queensland Coastal Wetlands Resources Mapping data. [Dataset]

Queensland Coastal Wetland Resources: Cape Tribulation to Bowling Green Bay

This report provides key resource data for the ongoing assessment of the requirement for additional Marine Protected Areas (e.g. FHAs under the Queensland Fisheries Act 1994) in regions of high fish habitat value in northern Queensland from Cape Tribulation to Bowling Green Bay (hereafter referred to as the Study Area). The study also provides baseline information on the coastal wetlands within this Study Area for consideration in the Ramsar site nomination process. The Study Area extends from Cape Tribulation (16o 6’S, 145o 24’E) to Bowling Green Bay (19o 30’S, 147o 24’E) in tropical north Queensland. The project aimed to: 1. document and map the coastal wetland communities of the Study Area; 2. document levels of existing disturbance to and protection of the wetlands; 3. examine existing recreational, indigenous and commercial fisheries resources in the region; 4. evaluate the conservation values of the areas investigated from the viewpoint of fisheries productivity and as habitat for important and/or threatened species for future FHA/MPA declaration. Dataset URL Link: Queensland Coastal Wetlands Resources Mapping data. [Dataset]

Queensland Coastal Wetland Resource Investigation of the Bowen Region: Cape Upstart to Gloucester Island

The wetland resources of the Queensland coastline have been mapped as a baseline dataset for Marine Protected Area investigation and particularly Fish Habitat Area (FHA) declaration, Ramsar site nomination and continued monitoring of these important fish habitats. This report summarises the results of the mapping undertaken in the Bowen region from the East Coast of Cape Upstart (Abbot Bay) to Gloucester Island (encompassing Edgecumbe Bay). The study was undertaken in order to: 1. document and map the coastal wetland communities within the Bowen region; 2. document levels of existing disturbance to and protection of the wetlands; 3. examine existing recreational and commercial fisheries in the region; and 4. evaluate the significance of the coastal wetlands in the region. Dataset URL Link: Queensland Coastal Wetlands Resources Mapping data. [Dataset]

Resource Assessment of the Tidal Wetland Vegetation of Western Cape York Peninsula, North Queensland, Report to Ocean Rescue 2000

A project to allow the resource assessment of tidal wetland vegetation of western Cape York Peninsula, in north Queensland, was undertaken as part of the longterm assessment of the coastal fisheries resources of Queensland. The project incorporated a littoral invertebrate fauna component. Extending from May 1993 to December 1994, fieldwork was undertaken in May 1993, November 1993 and April 1994. The aims of this project were to: • obtain baseline information on the distribution of marine plants of western Cape York Peninsula; • commence a preliminary assessment of the littoral invertebrate fauna and their habitat requirements with a view to extending knowledge of their biogeographic affinities; • perform biogeographic classification of the tidal wetlands at a meso and local scale for marine conservation planning; • evaluate the conservation values of the areas investigated from the viewpoint of fisheries productivity and as habitat for important/threatened species. Dataset URL Link: Queensland Coastal Wetlands Resources Mapping data. [Dataset]

Duckweed-based Wastewater Treatment Systems: Design Aspects and Integrated Reuse Options for Queensland Conditions

Point sources of wastewater pollution, including effluent from municipal sewage treatment plants and intensive livestock and processing industries, can contribute significantly to the degradation of receiving waters (Chambers et al. 1997; Productivity Commission 2004). This has led to increasingly stringent local wastewater discharge quotas (particularly regarding Nitrogen, Phosphorous and suspended solids), and many municipal authorities and industry managers are now faced with upgrading their existing treatment facilities in order to comply. However, with high construction, energy and maintenance expenses and increasing labour costs, traditional wastewater treatment systems are becoming an escalating financial burden for the communities and industries that operate them. This report was generated, in the first instance, for the Burdekin Shire Council to provide information on design aspects and parameters critical for developing duckweed-based wastewater treatment (DWT) in the Burdekin region. However, the information will be relevant to a range of wastewater sources throughout Queensland. This information has been collated from published literature and both overseas and local studies of pilot and full-scale DWT systems. This report also considers options to generate revenue from duckweed production (a significant feature of DWT), and provides specifications and component cost information (current at the time of publication) for a large-scale demonstration of an integrated DWT and fish production system.

Turf for tough times - keeping grass cover with less water, Proceedings of a seminar held at Redlands Research Station, Cleveland, Queensland, 6 June 2007.

Suitable for gaining some insights into important questions about the management of turf in dry times. Improve your product quality and avoid unnecessary losses. Can varieties help? How important are soils in conserving moisture and how do I measure my soil's condition? How can I make the best use of available water? Can water retaining amendments assist in establishing turf? Is recycled water a good option? Contains research results from turfgrass trials conducted by Queensland Government scientists for Queensland conditions.

Influence of plant traits on production costs and profitability of strawberry in southeast Queensland.

In Queensland the subtropical strawberry ( Fragaria * ananassa) breeding program aims to combine traits into novel genotypes that increase production efficiency. The contribution of individual plant traits to cost and income under subtropical Queensland conditions was investigated, with the overall goal of improving the profitability of the industry through the release of new strawberry cultivars. The study involved specifying the production and marketing system using three cultivars of strawberry that are currently widely grown annually in southeast Queensland, developing methods to assess the economic impact of changes to the system, and identifying plant traits that influence outcomes from the system. From May through September P (price; $ punnet -1), V (monthly mass; tonne of fruit on the market) and M (calendar month; i.e. May=5) were found to be related ( r2=0.92) by the function (SE) P=4.741(0.469)-0.001630(0.0005) V-0.226(0.102) M using data from 2006 to 2010 for the Brisbane central market. Both income and cost elements in the gross margin were subject to sensitivity analysis. 'Harvesting' and 'Handling/Packing' 'Groups' of 'Activities' were the major contributors to variable costs (each >20%) in the gross margin analysis. Within the 'Harvesting Group', the 'Picking Activity' contributed most (>80%) with the trait 'display of fruit' having the greatest (33%) influence on the cost of the 'Picking Activity'. Within the 'Handling/Packing Group', the 'Packing Activity' contributed 50% of costs with the traits 'fruit shape', 'fruit size variation' and 'resistance to bruising' having the greatest (12-62%) influence on the cost of the 'Packing Activity'. Non-plant items (e.g. carton purchases) made up the other 50% of the costs within the 'Handling/Packing Group'. When any of the individual traits in the 'Harvesting' and 'Handling/Packing' groups were changed by one unit (on a 1-9 scale) the gross margin changed by up to 1%. Increasing yield increased the gross margin to a maximum (15% above present) at 1320 g plant -1 (94% above present). A 10% redistribution of total yield from September to May increased the gross margin by 23%. Increasing fruit size increased gross margin: a 75% increase in fruit size (to ~30 g) produced a 22% increase in the gross margin. The modified gross margin analysis developed in this study allowed simultaneous estimation of the gross margin for the producer and gross value of the industry. These parameters sometimes move in opposite directions.