Abstract

Introduction

The management of the three locust species under the APLC charter (AFFA, 2000) – the Australian plague locust, Chortoicetes terminifera (Walker), the spur-throated locust, Austracris guttulosa (Walker), and the migratory locust, Locusta migratoria (L.) – relies on a preventive control program with the location and treatment of nymphal bands and adult swarms early in breeding sequences that potentially lead to plagues (Hunter et al., 1998). C. terminifera is a common pest of agriculture in Australia and the APLC has undertaken control of the species in most years since its creation in the mid-1970s, including major outbreaks in 1979, 1984, 1987, 1992-93 and 1999-2000. Outbreaks of A. guttulosa and L. migratoria are less common and have been restricted to Queensland and northern New South Wales. Plagues originate from several consecutive generations of successful breeding, often with migration to different areas between each. C.terminifera and L.migratoria can complete 3-5 generations in 12–18 months. For A.guttulosa, which has only one generation per year, it takes several years for outbreaks to develop.

Figure 1. Eastern Australia showing APLC infrastructure, operational area and potential locust habitat.

The strategy of preventive control to limit outbreaks depends on locating dense populations early in the sequence of generations. Improvements in the availability of environment and weather data allows the modelling of locust development and environmental variables influencing distribution and has aided the detection and minimisation of outbreaks. The opportunities for efficient aerial locust control are limited to periods of high aggregation coincident with fine weather and, for C.terminifera and L.migratoria, these may be only 2-3 weeks in each generation. Locating early population concentrations and predicting the timing of lifestage changes is therefore critical to the reduction of outbreaks. The way the system is used operationally to combine outputs of various environmental models with those of locust ecology is discussed in relation to critical events during the life cycle of C. terminifera, and in the context of the 2000 outbreak.

Background

With a potential infestation area of more than 2 million km² covering the sparsely populated, remote inland areas of four States (Fig. 1), finding population concentrations of highly clumped and highly mobile insects requires an efficient and rapid means of index monitoring (Thomson, 1998), a network of landholder reporting and a method of identifying areas where environmental and habitat conditions are favourable for survival, breeding and population increase. Locusts aggregate into swarms that can migrate distances up to 500km with the prevailing winds in a single night. These mass migrations may continue over several nights and occur through a range of altitudes, often making the prediction and detection of invading populations difficult. The rapid location of gregarious populations is necessary to allow for more intensive aerial assessment and control if warranted.

Geographic information systems (GIS) enable the integration of those environmental factors which determine habitat suitability, and those which influence locust distribution and recruitment, with known distributions to model development across the entire monitoring area. In this way predictions of the timing, reproductive success and likely numbers can be made for known, and potential populations at any location.

Decision Tools

The primary management decisions - the early intervention strategy, control opportunities, constraints and methodology, the use of spotter and spray aircraft - were made over years of research and operations. Operational campaign decisions - population thresholds to launch or cease campaigns, control agent, technique and identification of targets - require intensive field assessment.

The Decision Support System system consists of a set of computerised secondary decision tools (Norton & Mumford, 1993) that have been added or modified as technology, information sources and operational needs change. They provide access to databases, simulation models and spatial analyses and maps to support both forecasting and operations.

Forecasting support tools

• Daily weather analyses – rainfall, temperature,windfields
• Gridded accumulated rainfall for consecutive days
• Locust distributions- survey data
• Location-specific weather data for development models
• Automatic modelling of potential and known populations
• Wind trajectory analysis
• Bulletin maps

Operations access tools

• Field survey data collection, transfer and GIS processing
• Display interface to survey and control information
• Historical locust databases
• Landuse limitations – nature conservation, organic producers
• Property boundaries – state DCDB extracts
• Aerial navigation maps
• Property control maps
• DGPS target logfile conversion and management

Online support Tools

• Bureau of Meterology: regional forecasts, observations
• GMS IR hourly cloud images and loops
• Weather Watch Radar
• 4 – 10 day rainfall forecasts and longterm (SOI ) outlook.
• ADFA – Insect monitoring Radar (Bourke)
• State Agriculture Agency Websites during locust campaigns.

The DSS is built using a commercial vendor GIS system (ArcInfo, Arcview – ESRI) and uses scripting to coordinate data ingestion, processing, modelling and visualisation. It depends on regular internet FTP data feeds of reported and modelled weather data from the Australian Bureau of Meteorology (BoM) and locust distribution data from surveys as input to development and movement models. The software was chosen as it provided for creating a user interface, map creation, both vector and raster manipulation and analysis, linkage to the operating system and a relatively simple means of automating tasks.

The operation of the DSS for forecasting involves modelling critical lifestage events and likely outcomes for the current and offspring generation for multiple populations from initial distribution estimates based on survey or reports. It is designed to give experienced staff access to all relevant primary or derived information. For operations it provides map overviews and detailed regional views of the current locust situation that can be combined with relevant environmental conditions and infrastructure.

System Inputs

Locust Information

(a) Locust survey data

Locust count data from transects taken on survey are the primary data source for estimating population distributions. APLC staff sample locust populations by recording numbers and lifestages of each species on 250m transects at intervals of approximately 10km along highways and property roads. These data, along with information on vegetation condition at each transect stop are entered into a menu-driven data collection application on a GPS-linked palmtop PC. Transect counts indicate densities and the presence of gregarious populations and are used to extrapolate possible distribution across uniform habitat areas.

(b) Locust Reports

Reports from landholders, rural extension staff, State agriculture agencies and even travellers are crucial to the early location and assessment of the extent of locust infestations. Reports extend the information base beyond the public roads and are used to direct survey. Telephone or fax reports come through APLC field bases or direct to HQ in Canberra where the information is entered into a database and converted to GIS point features and are added to other distribution data. The exact location, species, lifestage, density and contact information are recorded to enable decisions about priority and timing of follow-up to be made.

(c) Locust Control records

Control is only undertaken by the APLC with consultation and permission from landholders. Details of APLC insecticide applications are recorded during campaigns. Weather conditions, control agent and formulation, aircraft and delivery parameters and GPS coordinates for each spray block are recorded and sent to HQ nightly during campaigns. Target details are entered in a database and summaries are distributed to stakeholders on a daily basis. In addition regional summary maps or individual property target maps can be generated as required.

Differential GPS (DGPS) has been used in spray aircraft for navigation and alignment of individual spray runs in recent years. During the 2000-01 the collection full DGPS logfiles from individual spray blocks was trialled as these data provide the most complete and accurate record of insecticide application. The DGPS logfiles are obtained either directly from pilots on flashcard, or from the contracting company after the end of a campaign. Files in various formats can be converted to line or polygon features within the GIS and archived. Procedures for managing this information source are being developed.

Locust Light Traps

A network of APLC-sponsored light traps in Qld, NSW and SA (Figure 1) provide regular information on the nocturnal flight activity of locusts. Data are faxed on a monthly basis, but large counts are generally reported by telephone. They assist in the interpretation of locust migrations by identifying the species involved and nights of high activity.

Weather Information

(d) Rainfall & Temperature

Daily rainfall and temperature data are collected from the Bureau of Meterology (BoM) SILO internet site via scheduled automatic FTP scripts and are converted to GIS grid and tabular databases soon after collection. Both interpolated gridded surfaces (Koch et al.,1983) and source point record files are available for the previous 24 hours. Temperature data files as maximum and minimum gridded surfaces and as point records are collected, processed and stored in the same way.

At the end of each month updated rainfall records from the wider property network are downloaded from the National Climate Centre (BoM) and the daily rainfall grids are automatically regenerated using the more complete data. Rainfall data are also stored as rainfall ‘events’, surfaces accumulated from a sequence of consecutive rain days up to a week long. Events are useful for visualisation of the total rainfall received in an area over several days and for summarising rainfall over several months. These events are also used for initialising potential model cohorts, based on rain amount above regional thresholds for sustained vegetation response.

The GIS automatically generates weather files formatted for the Dymex locust development model. These files use longterm mean temperature values for future dates and are updated with current values each day as new data are received. Files are created for a representative set of reporting stations and also as mean values for individual habitat units.

(e) Windfields

Information from a regional atmospheric circulation model used by the BoM provide the source data for an insect migration simulation to investigate possible locust mass migrations. Regular analyses and prognoses from the Limited Area Prediction System (LAPS) numerical weather prediction system (Puri et al.,1998) are available for subscribers in netCDF file format. Six-hourly analysis outputs are collected via automatic FTP and converted to hourly HDF files for wind vectors and temperatures at altitudes from 300m to 1800m. Prognosis files are collected so that predictions can be made up to 24 hours ahead of the latest analysis. In this way possible night flights of 12 hours duration can be analysed on the next morning.

Habitat Information

(f) Survey Data

Ground vegetation condition information is recorded on survey for the perennial, ephemeral and forb components. It is measured on a visible scale from green shooting, indicating a response to recent rain, to fully dry. This information indicates habitat suitability and, in the absence of further rain, when the drying of vegetation would cause mortality of nymphs.

NOAA NDVI Relative Change Imagery.

Regular composite image mosaics of AVHRR data are now available on the internet, but most are geared to higher biomass ecosystems than the arid areas with sparse vegetation cover where many locust outbreaks originate. Small variations in NDVI values in low-cover environments are often difficult to detect. Environment Australia has made NDVI imagery, modifed to highlight small variations in greenness, available to the APLC since 1998. Pixel values for 14-day image composites are recalculated as the current NDVI value relative the recorded historical range of values for each pixel. Image files are converted to ArcInfo grids and then displayed with colour ramping from green to dry as a backdrop to locust information and habitat type information using Arcview.

Studies have shown that the relative NDVI imagery is suitable for monitoring ground vegetation condition in response to rainfall over a range of arid and semi-arid landscapes (Deveson et al., 1998). It is used as a means of rapid visual comparison of the relative trends in vegetation condition across the full range of locust habitat types.

(g) Habitat Type

A map of potential locust habitats (Figure 1) is used to separate those inland landscapes that often support swarm development and breeding from those where locusts are rarely found. It is used to direct survey to areas with a higher historical frequency of infestation and also to stratify NDVI imagery to highlight suitable habitat in green condition. The map also provides the framework of uniform habitat units used to model the development of known and potential populations. The habitat classification is based upon the association of historical locust occurrence with open tussock grasslands on clay soils, calcareous earths and stony downs. It was generated using a combination of available continental spatial datasets of soil type and vegetation cover. The area of potential habitat, those landscapes capable of supporting swarm development and breeding, is about half of the two million km² area of operations.

Spatial Datasets

• Topographic Features - AUSLIG Topo 10M & Topo250K, DCW 1:1million.
• Cadastral Framework - Extracts from State DCBDs Qld, NSW & Victoria.
• Land Use Limitations – Organic producers – certified or ‘in progress’.
• Nature Conservation – reserve boundaries, Point locations for rare & threatened species, mound Springs (SA). Specific habitat polygons (Plains Wanderer).
• Infrastructure – APLC lightraps, ADMS airfields.
• Administrative Boundaries – Shires; NSW, Counties & Hundreds; SA, NSW RLPB Districts, APLC operational areas, BoM forecasting districts.
• APLC specific databases – locust control 1977-01, locust survey 1986-01
• Vegetation and Soils – various scales.

Software Components

Data Collection and Direct Data Transfer

Separate application programs handle the data collection and remote transfer of text files on the palmtops (HP 320LX). The field data entry application remains open on the palmtop during survey and a series of pick-list dialogue boxes provide for recording of vegetation conditions, species development stage and density. The final dialogue window displays current GPS coordinates which are added to the current record. Records are held in a repository file and are periodically written to ASCII export files. Navigation output from the GPS is split so that current location is available to the HF radio for emergency beaconing as well as to the application

A WindowsCE specific software application, ‘vxHpc’, manages remote networking and export file transmission from the palmtop using the ‘Zmodem’ text transfer protocol. The palmtop serial connection cable is swapped from the GPS to the HF-radio modem when data files are sent to Canberra. New survey data files are deposited directly to the GIS server and automatically converted and added to GIS point coverages. Data files can also be emailed from field bases at the end of a survey.

GIS User Interface

The DSS user interface is built as a set of menus for accessing the separate information management tools with data generally displayed as maps. Weather data are accessed through ArcInfo, while survey, control and modelled locust cohort information use Arcview projects to take advantage of the zoom and pan capabilities The DSS is designed to bring together information sources used by forecasting and operations staff to aid decision making and to allow modelling of likely outcomes for the current locust generation. The implementation of the various tools has followed operational requirements. A full range of contextual spatial datasets including topographic map data are incorporated into the GIS displays, (Figure 2.) The DSS operates through a range of scales, from a continental synoptic view of major locust species distributions, to regional views of pastoral individual holdings. All APLC historical data are accessible through Arcview project files.

Wind Trajectory Model

The migration modelling tool PMTRAJ (Rochester 1999) is used interactively to investigate likely migration paths from various take-off points through a user-defined set of flight parameters. The PMTRAJ software package combines the GenSIM (Rochester et al., 1996) simulation program with an interface to the ArcInfo GIS and uses the collected outputs of the BoM LAPS model. It has a trajectory generation module which traces wind vectors at given altitudes from a given set of points (Figure 2) and a redistribution module which uses a specified starting distribution. The program runs a simulated flight model defined by a set of parameters including take-off time, flight height, altitude and duration. The model can also be run backwards so that movements can be analysed from source or destination. Outputs from modelled flights can be also be used as inputs to the redistribution model for subsequent nights to analyse complex migration events.

Locust Development Model

Dymex population modelling software is incorporated into the DSS with associated development models. Dymex (CSIRO) consists of a model Builder program and a Simulator which runs specific models with selected weather files (Maywald, et al., 1999). The models incorporate current biological knowlege of each species and are easily modified within the Builder program. Locust development rates are a function of body temperature while recruitment and mortality are linked to rainfall and vegetation state. Dymex is run interactively by selecting a weather file for a reporting station, initialising cohort numbers and specifying the duration of the model run. ArcInfo is used to initiate automatic runs for multiple locations and display the predicted outputs for any date. These can be started with population profiles from survey data and use weather files calculated from BoM daily rainfall and temperature surfaces. Dymex outputs a table or file with cohort numbers in each defined lifestage for each time-step. For the modelled runs outputs are entered into attribute tables for each habitat unit.

The model for C. terminifera includes a temperature-dependent module for egg-diapause, which is significant in the southern part of its range in winter (Hunter, 1983). Eggs may also enter quiescence if laid in dry soil, and resume development after rain wets soil at egg-pod depth (~10cm). The principal modelled mortality factor is the desiccation of grass and forb food sources and is based upon regional seasonal characteristics of vegetation and soil type

DSS Operational Use

In practise the DSS is used as a set of information sources applied in logical steps to accumulate knowledge about known or potential populations during the course of each generation. The aim is to identify those areas where high density locust populations are likely to occur and to model the timing of critical lifestage events in which detection and intervention is possible. Weather data specific to known populations are input to locust development models to predict the timing of egg-laying, hatching, nymphal instar stages, fledging and migration. Survey is directed to these areas to identify if intensive ground and aerial survey, and possible control is warranted. Previous survey data provides the initial distributions and subsequent surveys enable the correction of model outputs to actual populations for the next generation. In areas where maturing adults are observed, subsequent surveys are timed when the development model indicates that nymphs of the following generation are likely to be found. Data from survey for nymphs and young adults provides the distribution of possible source areas for migrating populations for the next generation. Diapause becomes significant in the timing of winter generations in the southern half of the range of C. terminfera so the autumn distribution of gregarious adults provides an initial baseline for the spring generation.

The critical event that begins a locust generation in a region is rainfall sufficient to allow adults to mature eggs and oviposit (Hunter 1982) or to trigger the development of quiescent eggs in C. terminifera. Females can lay three pods of eggs at weekly intervals so multiple potential cohorts can result from most rain events. The location of rainfall

events relative to areas of potential habitat provides an initial stratification of landscapes where population aggregation or egg-laying is likely. In the arid and semi-arid inland (250-500 mm/pa) the areas affected by rain events at any one time may be only a small proportion of the total and migration of C. terminifera to small regions of green vegetation to breed is common. The intersection of these areas with potential migration pathways calculated using the trajectory model starting from known populations of suitable age gives possible sources of immigration and intitiation of new cohorts.

Figure 2. Wind trajectory model graphic output (left) and overview of spur-throated locust distribution and survey (right) from DSS interface.

Australian locust species frequently make nocturnal, long-distance migration flights that have been observed from radar studies in the 300-900m altitude range (Drake & Farrow, 1983). C. terminifera adults generally migrate in large numbers after a 7-10 day period of fat accumulation following fledging, but subsequent migrations between the first and subsequent egg-layings also occur. With numerous, asynchronous populations in the inland at most times, migrations of C. terminifera may occur any time from October to May. The association of these wind-assisted flights with the passage of rain-bearing weather systems provides a means of isolating the most likely timing of migrations.

DSS Performance 2000-2001

The events leading to the 2000 locust outbreak in eastern Australia were a sequence of widespread, unusually heavy rains in the inland from November 1999 to April 2000 with an already established initial outbreak population in NSW. The generation sequence involved migrations between rain areas in favourable habitats. APLC forecasts and surveys enabled control at nymphal and adult stages at various locations at each generation, despite logistical problems of flooding and rain. Survey data established the extent of concentrated populations so that resources were appropriately assigned, and also verified regions where populations were low and therefore not contributing to the outbreak.

Figure 3. Examples of AVHRR relative greenness imagery showing a range of dry (pale) to green (dark) vegetation and Australian plague locust survey data (symbols) for nymphs in October 2000, Broken Hill (left) and adults in March 2001, Orientos (right).

During the entire outbreak the only unexpected locust event was the enormous population multiplication which resulted from the January 2000 generation in southwest Qld. During January, APLC undertook control of nymphal bands and adult swarms in the Windorah-Noccundra region, but the extent of this population may have been much larger than those areas identified. In late January and early February adult locusts migrated into northwest NSW and northern SA around the same time as flood rains soaked these regions. The resulting successful breeding produced enormous numbers of bands from Lake Eyre to White Cliffs in March. An unexpected finding during this phase was large numbers of bands in the dunefields of the Strzelecki Desert. These areas had been mapped as largely unsuitable habitat, but with the exceptional rainfalls and high population, areas of inter-dune corridors were green enough to support bands.

The daily rainfall estimates provide an immediate indication of areas where locust activity is possible, but even the surfaces generated from later updated records may overestimate the the areas of high rainfall and, therefore, grass response, or miss significant local falls away from rain gauges. Earlier assessment of relative greenness index imagery from AVHRR showed that fortnightly composite images consistently detected ground vegetation response to rainfall in uniform locust habitat. In some cases the imagery was superior to rainfall estimates as it detected response to rainfall which was not reported. In addition it measures the variable directly related to locust mortality – ground vegetation condition. During the 2001 season the relative NDVI images several times indicated a potential to help identify localised locust concentrations, particularly where rainfall was patchy ( Figure 3 ). Larger numbers of locusts have been found in areas shown to be green on the NDVI imagery. Our problem in making operational use of this imagery is obtaining it soon enough to use it to direct survey rather than see the association later. At present there are very few organisations with such a near real-time need for processed AVHRR data.

The introduction of the palmtop PC data collection and wireless transfer system has proved to be reliable over recent locust seasons. Problems with field data transfer that have occurred have mostly resulted from connection problems that develop from field conditions. The palmtops have the advantage of compactness and instantaneous return to an application between uses compared to other possible input devices. Current field data on locust distribution, density and life-stage are available in the GIS within a day of collection. This system allows free direct data transfer that is independent of other communications networks. It runs on an HF radio system already used for remote area safety and voice communication and has proved operationally stable over several seasons.

Conclusion

The incorporation of range of information sources relevant to the detection and modelling of locust populations into a DSS maximises the opportunities for early intervention and allows management decisions to be based on best available field and modelled data. It also provides the opportunity to extend the value of collected data to test model hypotheses or possible scenarios by repeatedly running spatially explicit development models. Locust distribution data collected on field survey is an integral part of this system. The introduction of a palmtop PC - HF radio direct data transfer system has improved the timeliness and accuracy of data available for forecasting, and reduced errors and time wasted in multiple data handling.

Rainfall events are unpredictable in extent and timing and migrations can take place in any direction, so modelling locust outcomes beyond the current and offspring generation remains problematic. These elements of the DSS are therefore used to analyse events as they happen. Rainfall remains the key influence on the timing and location of successful locust reproduction and migration, the emergence of quiescent eggs, and provides a geographic focus for more widespread populations. It therefore acts as a synchronising force in an otherwise chaotic system.

Survey and report data are still the essential starting and recalibration points for modelling and forecasting locust outcomes. The possible outcomes from potential populations increases enormously with each generation, rain event and possible migration.

The incremental development of decision tools has taken advantage of advances in weather and environmental information products, primarily from the Bureau of Meterology. At the same time the internet has enabled access to the data in ‘near real-time’. Six yeas ago rainfall information came as a selected list of individual reports, which arrived as a letter from each State more than a week after the event. The prospect of just that list coming via email so it could be massaged into a database and interpolated seemed revolutionary. The GIS framework has accommodated these new data types and enabled the collection, analysis and presentation of synthesised information to be largely automatic. In addition the ready access to map-based historical survey and control data, along with linkage to development models which simulate population numbers as well as timing, enhances the use of a growing corporate knowledge base. It also provides a means of analysing and retesting those data and of extending the use of spatial modelling to improve our understanding and management of locust population dynamics.

References

APLC (2000). Australian Plague Locust Commission Annual Report 1998-99. Department of Agriculture, Fisheries and Forestry – Australia.

Deveson, E.D., Odjnanovich, L & Cridland, S. (1998) NOAA NDVI imagery detects ground vegetation response to rainfall in potential locust habitats. APLC Research Reports -1998.

Drake, V.A. & Farrow, R.A (1983). The nocturnal migration of the Australian plague locust, Chortoicetes terminifera (Walker) (Orthoptera: Acrididae): quantitative radar observations of a series of northward flights. Bull. Ent. Res.,73, 576-585.

Hunter, D. M.(1982). Adult development in the Australian plague locust, Chortoicetes terminifera (Walker)(Orthoptera: Acrididae) Bull.Ent.Res., 72,589-595.

Hunter, D. M. (1983). The maintenance of body temperature in adult Australian plague locusts. J. Aust. Ent. Soc., 22,135-136.

Hunter D.M., Strong K. & Spurgin P.A.(1998). Management of populations of the spur-throated locust, Austracris guttulosa (Walker), and migratory locust, Locusta migratoria (L.) (Orthoptera: Acrididae), in eastern Australia during 1996 and 1997. J.Orthopera Res. ,7, 173-178.

Koch, S.E., Des Jardins, M. & Kocin, P.J.(1983). The Interactive Barnes Objective map scheme for use with satellite and conventional data. Journ. Climate & Appl. Meterology., 22, 1478-1503.

Maywald, G.F., Timmers, P., Bottomly,W. & Sutherst, R.W. (1999). Dymex: modelling natural systems. Professional Model Simulator User’s Guide. CSIRO Publishing, Australia.

Puri, K., Dietachmayer, G., Mills, G.A., Davidson, N.E., Bowen, R.A., and Logan, L.W. (1998). "The new BMRC Limited Area Prediction System, LAPS." Australian Meteorological Magazine 47, No 3, 203-223.

Rochester, W. A. (1999). The migration systems of Helicoverpa punctigera (Wallengren) and Helicoverpa armigera (Hubner) (Lepidoptera:Noctuidae) in Australia. PhD thesis, The University of Queensland, Brisbane.

Rochester, W. A., Dillon, M. L., Fitt, G. P., and Zalucki, M. P. (1996). A simulation model of long-distance migration of Helicoverpa spp. Moths. Ecological Modelling , 86, 151-156.

Thomson, W. L., White, G. C. & Gowan C. (1998). Monitoring vertebrate populations. Academic Press, San Diego. (1998)