Agricultural Research Centre, PMB 19, Trangie NSW 2823
Phone 02 68808000, Fax 02 68887201
Email : firstname.lastname@example.org
The ability to place current rainfall and pasture conditions in the historical record, and to forecast pasture production in future months, are important tools for all who desire to manage land in a productive and sustainable way. Information currently available to land managers in NSW focuses on actual rainfall and does not have the capacity to evaluate the effectiveness of that rainfall in growing or maintaining pastures.
From 1997 to April 2000 the Aussie GRASS project (Australian Grassland and Rangeland Assessment by Spatial Simulation) developed a national spatial modelling framework using the GRASP pasture growth simulation model. This model is run at the beginning of each month and produces a range of products in the form of maps. Some Queensland land managers have had access to these products for several years.
Starting in February 2001, a group of land managers from western NSW have been collaborating in evaluation of these Aussie GRASS products for their area. Each month 100 individuals, located evenly throughout the area, receive the products either by mail or on the Internet. Maps include rainfall and pasture growth over a range of periods relative to the same period of months in the last 40 years, and forecasts of rainfall and pasture growth in the next three months. Evaluation Forms focus on the accuracy and usefulness of these maps in the enterprise of each collaborator. Forms returned over a 12-month period will be assessed and results used in decisions on future use of the products in western NSW.
The lower the average annual rainfall and the more variable the climate, the greater the interest in whether it will rain and whether the rain which falls will produce growth. As depicted on Figure 1, average annual rainfall for most of western NSW is < 400 mm; for approximately half of the area it is < 300 mm and for small areas in the far west it is < 200 mm. The non-dependable nature of this rainfall is accentuated by the fact that in much of western NSW rainfall is aseasonal. Whereas rainfall in Victoria is concentrated in the winter months (> 1.3:1), in the area from the Victorian border to a line running west from approximately Byrock and then north-west through Milparinka, rainfall is typically uniform throughout the year, with the possibility that no one period will receive sufficient rain to grow pasture. To the north of this area rainfall is increasingly concentrated in the summer months.
Figure 1 Map showing average annual rainfall for NSW based on the 30 years of records from 1961 to 1990. The line running from just east of Bourke and Cobar and through Deniliquin is the 400 mm isohyet. The line west of Ivanhoe is the 300 mm isohyet and the two small areas on the SA border receive <200 mm.
In rural areas, discussions concerning the weather frequently focus on comparison of current conditions with conditions in past years. The tendency is to remember a certain good season as the standard for hopes and expectations in future years. This is actively promoted by those selling land, who often use the good year as a measure of a property’s potential. Thus new people come into an area with unrealistic expectations of the amount of rain and pasture growth that might occur on their newly acquired property. Even those who have bought in over the last 15 or so years, have a minimal understanding of what constitutes an average or normal season.
The low average annual rainfall, the aseasonal nature of that rainfall, and unrealistic expectations, all compound to increase the difficulty experienced by land managers in managing the land in a productive and sustainable way.
Tools currently available to assist land managers in NSW are limited to climate information on actual rainfall, the Southern Oscillation Index (SOI) and Sea Surface Temperatures. The monthly Regional review produced by NSW Agriculture adds areas of NSW suffering drought, and situation statements for major crop types and pastures and for agricultural production on a regional basis. The reality is that none of this information has the capacity to place current pasture growth, as distinct from rainfall, in the context of the historical record, or to forecast future pasture growth. And yet, in making decisions regarding almost every aspect of land management, it is present and future pasture conditions which are the critical factors.
The National Drought Alert Strategic Information System Project (NDASIS; Queensland Department of Primary Industries - QDPI, 1993-1996; Brook 1996; Carter et al. 1996) and the subsequent Australian Grassland and Rangeland Assessment by Spatial Simulation Project (Aussie GRASS; Queensland Department of Natural Resources - QDNR, 1997-2000; Hall et al. 2001; Richards et al. 2001) worked towards providing a range of products including pasture growth information. Funded by Land and Water Resources Research and Development Corporation (LWRRDC), these projects focussed on setting up a national spatial modelling framework for the whole of Australia. The main simulation model developed for this framework was the GRASP pasture growth model. Calibration and validation of this model was achieved for Queensland during the NDASIS project and attempted for the other states during the Aussie GRASS project.
With the objective of further evaluating the accuracy and usefulness of materials produced by the Aussie GRASS system for western NSW, selected products are currently being assessed by 100 collaborators located across the Western Division and Hay Plain.
Achievements of National Drought Alert Strategic Information System Project (NDASIS) and Australian Grassland and Rangeland Assessment by Spatial Simulation Project (Aussie GRASS)
The national spatial modelling framework developed by NDASIS divided the whole of Australia into grid cells, each ∼∼ 25 square kilometres (5 x 5 km) in size. The pasture growth model GRASP, which runs within this framework (Littleboy and McKeon 1997), simulates/mimics pasture growth (ie new growth on perennial plants plus new annual plants), pasture biomass, soil water, ground cover and pasture utilisation by grazing animals.
All data fed into the model has to be in spatial format (ie a form which gives a unique value for each location or grid cell). Climatic data includes daily rainfall, temperature, humidity and radiation. Thus a comprehensive network of telegraphic rainfall recording stations is essential. Additional data sets required by GRASP are soil type, vegetation community, numbers of domestic stock, other grazing animals and tree cover.
The model is run, during the first week of each month, on the QDNR’s powerful Supercomputer. This computer system calculates the required variable (eg pasture growth) grid cell by grid cell. For any one grid cell the computer takes the soil data, the vegetation data, the % tree cover, the rainfall and other climatic data and the herbivore numbers for that cell, and substitutes them in the equations making up the GRASP model. The model then calculates the value for the desired variable for that cell. The values for each grid cell for any one variable are then combined into maps, coloured according to the calculated values. Maps can depict absolute values (eg pasture growth as kg dry matter/Ha for the previous month) or values relative to the historical record (detailed climatic data is available in digital form for the last 40 years).
The Aussie GRASS modelling system can be run for any period of time in any year for which digital complete climatic data is available. Thus pasture growth in past years can be calculated and the current conditions of pasture growth placed in the context of pasture growth for the same month or period of months in the historical record. With this capacity, maps of rainfall and pasture growth over the last three, six, 12 and 24 months, relative to the same period of months in the historical record, can be produced.
In addition to pasture growth, GRASP calculates Total Standing Dry Matter (TSDM equals weight in kg per hectare of all the grasses and herbs present in the pasture, including old and new growth, after all moisture has been removed). The GRASP model also has the capacity to forecast probability values of exceeding median rainfall or median pasture growth over the next three months.
During the Aussie GRASS project, researchers in NSW Agriculture and the Department of Land and Water Conservation worked together to adjust the Aussie GRASS system to the conditions of the grazing lands of western NSW. In cooperation with the Bureau of Meteorology, 18 additional telegraphic rainfall stations were established, more accurate and complete spatial data sets were provided and field data for ground-truthing GRASP was collected. The field data was collected using the technique termed “spider mapping” (Hassett et al. 2000, Clipperton and Bean 2000). The spider mapping data was supplemented with biomass observations from 334 Rangeland Assessment Program (RAP) sites.
For calibration and validation purposes, an average value of pasture yield for each grid cell was calculated using all the estimated values located within that one 5 x 5 km grid cell. The average value for each grid cell was then compared with the value predicted by GRASP for the same grid cell at the same period of time (Figure 2). Sixty-six % of the data collected in spider mapping and from RAP sites was used for calibration of the model and the remaining 33% for validation. In addition, the Normalised Difference Vegetation Index (NDVI is based on brightness values of the Near-infrared and Red bands and is a representation of the "greenness" of vegetation at the time the satellite passed over a designated area) calculated from the NOAA satellite data was used in a complementary calibration process (Figure 3). Calibration was carried out separately for each vegetation community. If there were significant differences in the observed and predicted values, the value of the relevant variable was changed and the model re-run. This procedure continued until the fit between the predicted and observed values for each vegetation community was acceptably close.
Figure 2 Observed and predicted Total Standing Dry Matter values for vegetation communities in western NSW following validation. The observed values for each community represent the mean of all calibration observations, on a grid cell basis, for that community collected as part of the spider mapping program and at the 334 RAP sites. The predicted value for each vegetation community is the mean of all values for the same grid cells and on the same dates as the observations were made. The numbers of the data points are the codes for each vegetation community as used in the GRASP model (Richards et al2001).
At two periods during the Aussie GRASS project, workshops were held in western NSW to explain the project and seek feedback from stakeholders. During the second round (March to May 2000) considerable interest was shown in the Aussie GRASS products. Interest in the potential of the maps had also been evident in 1998 when the GRASP model was run for the January 1997 to June 1998 period for all points within the Wentworth Rural Lands Protection Board (RLPB) for which detailed climate data was known. For this 18-month period, the model calculated rainfall as having been in the 32.7 to 4.8 percentile range and pasture growth even lower in the 14 to 1.9 percentile range. On the basis of this the Wentworth RLPB's request for Drought Exceptional Circumstances Status was granted.
Figure 3 Time series for the period 1982 to 1992 of observed (circles) and predicted (squares) Normalised Difference Vegetation Index (NDVI) values for the belah- bluebush vegetation community of western NSW (Richards et al. 2001).
At completion of the Aussie GRASS project, for the western areas of NSW, the GRASP model had been calibrated and validated to the point where at least the relative maps were judged to be reasonably accurate.
Towards the end of 2000 NSW Agriculture, recognising the potential usefulness of the Aussie GRASS products to land managers, agreed to internally fund an evaluation of the products for NSW. A contract drawn up with QDNR allowed for continued production of the required NSW maps over a two year period. It was planned to involve 100 collaborators, half receiving the Report by post and the other half accessing it on the Internet.
Over a 12-month period starting in February 2001 the Report would be produced each month and, together with an Evaluation Form, posted to the collaborators. Collaborators would fill in and return the form for analysis. After the initial six months, evaluation would be extended to include whether the Report would be appropriate as part or total replacement of NSW Agriculture’s Regional Review. Data from analysis of responses would be used as the basis for decisions on future use of the products.
The primary criteria applied in seeking collaborators were 1 - that collaborators be located as evenly as possible over the whole of the area, 2 - that they include representatives of all the major stakeholders and 3 - that the major vegetation communities or country types be included.
A list of potential collaborators was made. Each individual, family or group was then contacted directly by phone. The planned Report and Evaluation Form were described and interest in acting as a collaborator sought.
By the beginning of February, from the list of potential collaborators, 111 had agreed to act as collaborators. Only 22 of these had an Internet connection of sufficient speed to receive the Report by this media. Thus 89 chose to receive a hard copy by post.
Stakeholders represented in the final list include landholders, staff and Directors of RLPBs, stock and station agents, fire control personnel, Shire Councils, agency people, and officer-bearers of both NSW Farmers and the Pastoralist's Association of the West Darling.
Nature of Report
Based on input from stakeholders at the workshops, the Report consists of a range of information all brought together in the one integrated Report. As the maps need to be reproduced in full colour and such reproduction is expensive, the size of the Report is limited to one double-sided A3 page. The number of maps which can be included is limited to 17. The Report focuses on maps placing current rainfall and pasture growth conditions within the historical record, and maps forecasting future pasture growth.
Maps 2 to 6 show rainfall one month ago, two months ago, during the last three months, the last six months and the last 12 months, all relative to the relevant month or period of months over the last 40 years. Maps 8 to 12 show pasture growth for the same periods, all relative to the relevant month or period of months over the last 40 years.
Values on these relative maps are presented as percentile ranges. To allocate a certain value to a percentile range, all the values in the historical record for that variable are listed in ascending order. For example if one is looking at rainfall during the month of January : values for rainfall during the month of January during the last 40 years are listed in ascending order. These values are then divided into groups of four (10% of 40). If the current month of January received rain in the range of the lowest four values, then rainfall for the current month was in the 0-10 percentile range (equals the first decile). If the current month of January received rain in the range of the second lowest group of recordings from past years, then rainfall for the current month was in the 10-20 percentile range (equals the second decile), and so on. If the current value was in the fifth or sixth groups of past recorded values, this 40-60 percentile range (equals fifth and sixth decile) is said to represent average or normal conditions. At the other end of the range, if in the current month rainfall was in the range of values greater than the fifth highest value in past years, rainfall in the current month was in the 90-100 percentile range (equals the tenth decile). This systematic mathematical analysis of the historical data using percentile ranges removes the tendency to over-emphasise conditions in the good years and gives a better assessment of what is a realistic expectation for a particular month or season of the year.
Maps 13 and 14 in the Report are forecast maps. These show the probability of exceeding median rainfall and median pasture growth in the next three months. These are calculated by GRASP based on analogue years. Years in the historical record which had the same phase of the SOI (the five phases of the SOI are consistently positive, consistently near zero, consistently negative, rapidly rising and rapidly falling) for the same two months as the last two months in the current year are called analogue years. Rainfall values for these analogue years are compared to the median value (value which divides the historical observations in half) over all years in the historical record. To produce the forecast map of pasture growth over the next three months, GRASP is run using the climate data for the next three months from the analogue years. The predicted pasture growth for these analogue years is then compared to the median value of pasture growth over all years in the historical record. If rainfall or pasture growth in say eight out of 10 analogue years was greater than the median value, then the probability of exceeding median rainfall/pasture growth is 80%. At the other end of the spectrum, if rainfall or pasture growth in only 2 of the 10 analogue years was greater than the median value the probability would be 20%. Low probabilities (0-20%) give a strong indication that rainfall or pasture growth in the next three months may be lower than the average or normal. High probabilities (80-90%) give a strong indication that rainfall or pasture growth may be higher than the average or normal. It is important to appreciate that a probability is not a prediction. Even within an 80 or 90% probability of exceeding the median in the next three months, there is the possibility that the next three months will follow the path of the two or one years out of the 10 analogue years in which the rainfall or pasture conditions were lower than the median value.
In addition to the relative and forecast maps, two absolute value maps are included. Map 1 depicts actual rainfall for the previous month, as measured at the network of telegraphic stations. This is included to assist collaborators to calculate the difference between rainfall measured at their property and that calculated by interpolation between adjacent telegraphic stations. In evaluating the accuracy of the pasture growth maps, a mental adjustment needs to be made by the respondent for the difference in measured rainfall (includes localised storms) and the value recorded at the telegraphic stations. The pasture growth map is based on the latter.
Map 7 depicts actual pasture growth (kg dry matter/Ha) in the preceding month as calculated by GRASP. Although it is realised that many collaborators do not have the tools to calculate the amount of pasture growth which occurred in their paddocks over the previous month in terms of kg/Ha, they would be able to evaluate the accuracy of this map in relative terms. For example, if the map shows pasture growth during the last month as 20-50 kg DM/Ha to the north of their property and 10-20 kg DM/Ha to the south, and the respondent can see that new pasture on the ground is an order higher in the north than in the south, then they conclude that the map is accurate. Or, if the map shows pasture growth for the whole area as 10-20 kg DM/Ha and the respondent knows that there has been no significant pasture growth on their property over the last month, they conclude that the map is not accurate.
Figure 4 Sample section of the maps included in the Monthly Seasonal Report as at 1st May 2001. Maps 7 to 9 are three of the pasture growth maps, and 13 and 14 are the two forecast maps
Map 15 depicts Total Standing Dry Matter for the previous month relative to the historical record. Calculations of TSDM are made as at the last day of the previous month (eg 30 June) by taking the value calculated at the last day of the preceding month (31 May), adding the amount of pasture predicted to have grown during the month of June, and subtracting the amounts calculated to have been consumed by grazing animals during June or which would have disintegrated through decomposition/decay. Comparison of Map 15 and Maps 7 to 12 gives a good indication of the nutritional value of available forage. For example if pasture growth has been limited or practically non existent over the last three months, but TSDM is relatively high, it can be assumed that much of the available forage has been in the paddock for longer than three months and will be declining in nutritional value.
Maps 16 and 17 are fire products. The map showing Curing Index shows the proportion of the total standing dry matter which is dead. GRASP does not include in this calculation litter which has detached from the plants. The percentage is obtained by dividing the dead TSDM by all TSDM. The map of Potential fire risk only reflects the amount of fuel available and the degree of curing. In using this map, it is essential to take into account weather conditions. The actual level of risk will be strongly influenced by factors such as dampness of the pasture and temperature of the day.
All maps included in the Report which are generated by the GRASP model are Experimental prototypes. Their accuracy at this stage cannot be guaranteed.
To assist in predicting future climatic conditions, the current Sea Surface Temperature Anomalies map and the Southern Oscillation Index and SOI Phases Chart are included in the Report.
The Southern Oscillation Index has dropped further during April (average value + 4.87 to + 1.39), however it remains positive but in a neutral SOI phase. Sea surface waters in the eastern equatorial Pacific Ocean have further warmed up and their geographical spread has increased. A warming trend in the surface waters of the Indian Ocean was also recorded during April. The majority of computer simulation models that develop seasonal forecasts suggest neutral sea surface temperatures in the Pacific Ocean for at least September. Looking ahead till December 2001, half of the models predict neutral conditions and the other half predict warm conditions. Indicating that the possibility of El Nina type conditions cannot be ruled out.
The National Climate Centre’s temperature outlook is that of warmer-than-average conditions across the State.
Area 1 = Northern half of the Lower Rainfall Region
The probability of exceeding median rainfall for the next three months (Map 13) falls in the 40-60% range for the entire northern area. The probability of exceeding median growth for the next three months is slightly less favourable (Map 14). Probabilities of 0-20% were calculated for portions of the Walgett, Brewarrina, Bourke and Cobar boards. Probabilities of 30-40% were calculated in the northern portion of the Broken Hill board and in the north-west of the Wilcannia board and Milparinka boards. On a more positive note, in a band running from the central area of the Wilcannia board across to the NW corner of the Cobar board, the probability of exceeding median pasture growth was calculated as 60-70%. It is expected that feed quality will further decline. Since rainfall and pasture growth predictions give no indication of exceeding median values in the next three months, consideration needs to be given to the adequacy of feed availability to meet the needs of livestock, both domestic and non domestic. Special consideration needs to be given to pregnant and lactating livestock as they have much higher nutritional requirements. A glove box guide to “Tactical Grazing Management” for the semi-arid woodlands has been designed to help graziers make such decisions. Copies can be purchased from NSW Agriculture offices.
Area 2 = Southern half of the Lower Rainfall Region
The probability of exceeding median rainfall during the next 3 months (Map 13) is in the 40-60% range for all of the southern area. The probability of exceeding median pasture growth over the same three months (Map 14) is 40-70% for the Broken Hill and Wentworth boards and much of the Hillston and Balranald boards. Exceptions to this 40-70% probability are 0-20% in the central and southern areas of the Balranald board, the southern portion of the Hay board and virtually all of the Riverina board, and 20-30% in an area east of the town of Hillston and an area in the south-west of the Hay board. It is concluded that for most of the southern region there is little or no indication that rainfall and/or pasture growth will exceed median values in the next three months. This, coupled with some of the lowest rainfall on record for April, indicates the need to match stock numbers and pasture availability/quality.
Figure 5 Sample section of comments on the Monthly Seasonal Report as at 1st May 2001.
All the maps described above are arranged on pages 2 and 3 on the inside section of the folded A3 Report (Figure 4). Comments on Rainfall, Pasture Growth/TSDM and Outlook, fill pages 1 and 4 on the outside of the folded A3 Report (Figure 5). These comments are subdivided into comments on the Northern half of the Lower Rainfall Region and those relevant to the Southern half of the Region. These are written each month by the NSW Agriculture Rangeland Extension Officers located at Bourke and Dareton. General comments on Outlook are written by the Department’s Agroclimatologist based in Orange.
Preparation of the Report and its distribution
The maps generated by the Aussie GRASS system in the first week of each month, are posted by QDNR on the Aussie GRASS web site on 6th or 7th of the month. The chosen maps are down-loaded into the map portion of a Microsoft Word Template. The current Sea Surface Temperature Anomalies map is down-loaded from the NOAA web site and the current Southern Oscillation Index and ‘SOI Phase’ Chart from the Long Paddock web site of the QDPI. Comments received from NSW Agriculture staff at Bourke, Dareton and Orange are edited and placed into the text portion of the Word Template.
The documents are colour printed on high grade A3 paper and then commercially colour photocopied in Dubbo.
Each month an accompanying letter is produced and, when necessary, additional explanatory material written. These, together with the Report, Evaluation Form and Reply Paid envelop are posted to all collaborators.
The Report is converted to four quarto pages in .pdf format and posted, with an electronic Evaluation Form, on NSW Agriculture’s external web site on a password protected page.
The Evaluation Form focuses on assessing the accuracy and usefulness of the Report in a range of enterprises/businesses. In the introductory section respondents are asked to select the appropriate category of enterprise/business (eg landholder or fire control personnel), write the RLPB or Boards in which they are located/work, the name of their property and the major country type/s on which their enterprise/business is located.
Question 1 asks about the level of accuracy of each category of map on each country type. Five levels of accuracy are given, with boxes for Excellent, Good, Moderate, Poor and Very Poor. The categories of map are Rainfall (Nos 1-6), Pasture Growth (Nos 7-12), Total Standing Dry Matter (No 15), Curing Index (No 16) and Potential Fire Risk (No 17).
Question 2 asks, if there is a change in colour on the maps in the area of their enterprise/work, does the location of the colour boundary reflect a change which can be seen on the ground. This question gives a Yes/No choice for each category of map as listed in Question 1.
Question 3 asks for the level of usefulness/value of each category of map to the respondent in their enterprise/business. The respondent can choose from five levels of usefulness (Very useful, Useful, Moderately useful, Not very useful and Not useful at all). To the categories of maps listed in Question 1 and 2, is added the category of Forecast maps.
Question 4 asks for what purpose(s) was the respondent able to use the maps that they considered ‘Very useful’ or ‘Useful’ in Question 3.
Question 5 asks for suggestions for improvement in presentation or nature of the Report.
Analysis of responses
Each month the answers given in the returned Evaluation Forms are entered into a data base for analysis.
The level of accuracy for the different categories of map will be analysed : 1 - according to RLPB, 2 - according to country type and 3 - according to month. With the exception of the rainfall maps, the aim of these analyses will be to assess the performance of the GRASP model in the different regions and country types and at different times of the year. If the level of accuracy of a particular type of map in a particular type of country is significantly lower than is acceptable, more work will need to be carried out on calibration of the model for these vegetation communities and/or the particular time of the year.
Responses to Question 3 on the level of usefulness of each category of map, will be analysed initially according to enterprise/business. Further analysis will look at any trend with time in the level of usefulness as listed by individual respondents. This should indicate whether or not experience gained by monthly contact with the maps leads to an increasing awareness and understanding of their usefulness.
Responses to Question 4 will be analysed according to enterprise/business. In analysis of Questions 3 and 4, it will be important to allow for the influence of time of year and differing seasonal conditions, on variation in the level of usefulness and the number and types of decisions for which the maps were used.
Land managers in the grazing lands of western NSW experience many difficulties in managing their land in a productive and sustainable way. These difficulties are increased by the lack of information available to assist them to place current pasture conditions in the historical record and to forecast pasture production in future months.
The Aussie GRASS project, building on the achievements of the NDASIS project, worked towards provision of required spatial data sets and calibration and validation of the GRASP pasture growth model for grazing areas outside Queensland. On completion of this project in mid 2000, the Aussie GRASS system was producing maps which for western NSW were judged to be reasonably accurate, especially maps of pasture growth relative to the historical record.
Recognising the need for such information, starting in February 2001 NSW Agriculture has been running an evaluation of the accuracy and usefulness of these products for land managers in western NSW. Responses from the 100 collaborators, located evenly across the Western Division and Hay Plain, will be analysed and the data used in decisions regarding future use of the Aussie GRASS products in this area.
The authors wish to acknowledge the work of Rob Richards of the Department of Land and Water Conservation in all aspects of the Aussie GRASS Project as implemented in western NSW. We also thank Wayne Hall, the Brisbane-based Coordinator of the Aussie GRASS project, for his encouragement, direction and backup. All those who have agreed to act as collaborators in Evaluation of the alternative approach to seasonal monitoring are thanked for their willingness, time and effort.
Brook, K. (1996). Research Summary: Development of a National Drought Alert Strategic Information System. Volume 1. Final Report on QDPI 20 to Land and Water Resources Research and Development Corporation, Brisbane.
Carter, J., Flood, N., McKeon, G., Peacock, A. and Beswick, A. (1996). Model framework, Parameter derivation, Model calibration, Model validation, Model outputs and Web technology: Development of a National Drought Alert Strategic Information System. Volume 4. Final Report on QDPI 20 to Land and Water Resources Research and Development Corporation, Brisbane.
Clipperton, S. and Bean, J. (2000). A technique for rapid acquisition of spatial ecological data. Proceedings Centenary Symposium, Australian Rangeland Society, Broken Hill, 21 - 24 August, 2000.
Hall, W., Bruget, D., Carter, J., McKeon, G., Peacock, A. and Brook,K. (2001). Australian Grassland and Rangeland Assessment by Spatial Simulation (Aussie GRASS). Final Report for the Climate Variability in Agriculture Program, April 2001.
Hassett, R.C., Wood, H.L., Carter, J.O. and Danaher, T.J. (2000). A field method for statewide ground-truthing of a spatial pasture growth model. Australian Journal of Experimental Agriculture, 40, 1069-1079.
Littleboy, M. and McKeon, G.M. (1997). Subroutine GRASP: grass production model; documentation of the Marcoola version of subroutine GRASP. Evaluating the Risks of Pasture and Land Degradation in Native Pastures in Queensland, Appendix 2. Final Project Report for Rural Industries and Research Development Corporation project DAQI24A.
Richards, R., Watson, I., Bean, J., Maconochie, J., Clipperton, S., Green, D., Hacker, R. and Beeston, G. (2001). Australian Grassland and Rangeland Assessment by Spatial Simulation (Aussie GRASS): Southern Pastures. Final Report for the Climate Variability in Agriculture Program, April 2001.