Content uploaded by John Carter
Author content
All content in this area was uploaded by John Carter on Aug 24, 2015
Content may be subject to copyright.
University of Nebraska - Lincoln
DigitalCommons@University of Nebraska - Lincoln
Drought Network News (1994-2001) Drought -- National Drought Mitigation Center
8-1-1996
A Prototype National Drought Alert Strategic
Information System for Australia
Ken D. Brook
Resource Sciences Centre, Queensland Department of Natural Resources (QDNR), Brisbane, Australia
John O. Carter
Resource Sciences Centre, Queensland Department of Natural Resources (QDNR), Brisbane, Australia
is Article is brought to you for free and open access by the Drought -- National Drought Mitigation Center at DigitalCommons@University of
Nebraska - Lincoln. It has been accepted for inclusion in Drought Network News (1994-2001) by an authorized administrator of
DigitalCommons@University of Nebraska - Lincoln. For more information, please contact proyster@unl.edu.
Brook, Ken D. and Carter, John O., "A Prototype National Drought Alert Strategic Information System for Australia" (1996). Drought
Network News (1994-2001). Paper 10.
hp://digitalcommons.unl.edu/droughtnetnews/10
This article appeared in the June 1996 issue of Drought Network News.
A Prototype National Drought Alert Strategic Information System for Australia
Introduction
Defining and categorizing drought in a quantitative and scientific man-
ner are important national issues for Australian state and Commonwealth
governments, landholders, and agribusiness. The challenge for modelers of
Australia’s grasslands is to integrate biological models, geographic informa-
tion systems (GIS), satellite imagery, economics, climatology, and visual
high-performance computing into an Internet-deliverable application that
can provide easily understood monitoring and prediction advice in near real-
time—a national drought alert strategic information system.
Although NOAA satellite-derived imagery has been somewhat useful in
the broad-scale spatial assessment of green cover, especially the spatial
response of vegetation to rainfall events (Smith, 1994; Dudgeon et al., 1990;
Filet et al., 1990), it has inherent limitations in providing a total solution for
drought and rangeland monitoring; biomass relationships are not good, tree
cover confounds the signal, and a future projection of the current situation is
not inherent. Also, the interpretation of the imagery does not usually consider
the effects of soil type, vegetation structure, or rangeland “condition.”
Similarly, rainfall analyses alone do not necessarily reflect the quantity and
quality of pasture available on the ground. In the recent 1991–95 record-
breaking drought in Queensland, rainfall analyses did not map the drought-
declared southwestern areas of the state as droughted, and, conversely,
coastal areas of the state were classed as droughted by rainfall analyses, when
there was no community push for their declaration. Measures of rainfall
effectiveness expressed as measures of plant biomass are required for
drought definition. Improved assessments of the quantity and quality of
biomass are needed, as well as consideration of herbivore densities and
future climatic scenarios.
Land Condition Alerts
The Queensland Department of Natural Resources (QDNR) systems
approach to the management of native grasslands recognizes that rainfall
drought occurs at a regional scale, and that impacts on livestock and natural
resources can be forecast using simple models of soil water, plant growth,
and animal performance. Our vision for a comprehensive national drought
alert strategic information system is a system that consists of the best
combination of rainfall analyses, seasonal climate forecasts, satellite and
terrestrial monitoring, and simulation models of meaningful biological
processes. The information would be available on an Internet web server run
as a joint facility by the various state departments.
The QDNR drought research program aims to predict the occurrence of
land condition alerts on a quarter to half local-government area (shire) basis.
This will provide a rational basis for large-scale management decisions by
graziers, extension workers, and politicians. Land condition alerts are likely
to be issued when approaching an El Niño summer with low pasture cover,
high stocking rates, fragile soils, and a high probability of a change in the
species composition of the pasture. Land condition alerts currently comprise
maps of ground cover and pasture utilization that are output layers from the
spatial simulation.
Given the varying perceptions of actual drought compared to long-term
probabilities, the model also needs to be run over the historical weather
record to determine real-time risk as determined by the current plant biomass
and stock numbers.
The GRASP Simulation Model
The core simulation model used by the project is the GRASP (GRASs
Production) pasture simulation (McKeon and Littleboy, 1996). This model
has been developed as a result of a number of Queensland Department of
Primary Industries (QDPI) and Rural Industries Research and Development
Corporation (RIRDC) research projects. GRASP combines two successful
approaches in modeling plant growth—those of Fitzpatrick and Nix (1970)
and McCown et al. (1974). The soil water budget is simulated using four
layers (0-10 cm, 10-50 cm, 50-100 cm, and a deeper tree layer), and the
processes of runoff, drainage, soil evaporation, and transpiration are sepa-
rately calculated on each day from inputs of rainfall and pan evaporation
(Rickert and McKeon, 1982). The soil variables required include available
soil-water range for each layer, maximum rate of bare soil evaporation, and
infiltration parameters. A daily plant-growth index is calculated from a soil
water index, plant growth response to average air temperature, vapor pres-
sure deficit, solar radiation, and nitrogen availability. Plant growth is calcu-
lated as a function of growth index, plant density, and potential regrowth. As
green cover increases, plant growth is calculated from a combination of
temperature response, transpiration, radiation, temperature, and nitrogen
uptake. Tree density, expressed as trunk basal area, is used to adjust the soil
water balance and pasture production of the model. A number of nitrogen
submodels are being evaluated to improve estimates of pasture quality, plant
growth rates, and liveweight gain. On a daily basis, animals remove green
and dry biomass, and animal liveweight gain is calculated from empirical
relationships derived from grazing trials.
The main plant parameters are derived from calibrating the model
against field data. In the Queensland Grass Under Nutritional Stability, Yield
Nitrogen and phenological Development (GUNSYNpD) program (McKeon
et al., 1990), a group of pasture ecologists is collaborating in calibrating the
GRASP model from more than 75 site and treatment combinations. Some
calibrations have now been achieved using site data from other states.
Validation has also been conducted from use of other pasture production
measurements and historical grazing trial data.
The model produces estimates of pasture growth, biomass in green and
dead pools, green cover, soil moisture, runoff, animal liveweight gain and
pasture utilization on a daily basis, and can be run forward up to 180 days into
the future (although typically only 90 days). When pasture production is
combined with stock estimates, calculations of the degree of pasture utiliza-
tion can be made and displayed as maps of feed availability and land
condition, with a resolution of a quarter to half a shire. These maps form a
core product of the strategic information system.
The National Spatial Model
A national spatial modeling framework has been developed that allows
agricultural simulation models to be run at a continental scale on a 5 km grid.
The model structure takes advantage of the high performance vector archi-
tecture of Cray supercomputers, which allow simultaneous transformations
of state vectors. To attain maximum speed-up, at every model iteration (a
daily time-step), the spatial model simultaneously evaluates all simulation
state-variables across the entire 250,000 land pixels. The model is capable of
efficiently running any daily time-step biological simulation model, pro-
vided the model is re-coded to run for all pixels a day at a time.
The spatial framework comprises some 3,400 lines of FORTRAN 90,
while GRASP has another 2,000 lines of FORTRAN 90. The Agricultural
Production Systems Simulator (APSIM) model was also included in the
spatial framework for modeling wheat crops, but program execution was
quite slow because APSIM has not been re-coded for optimum use of the
Cray’s architecture.
Seasonal Climate Scenarios
The spatial model uses interpolated historical and near real-time meteo-
rological data to run the model up to a known point in the past or the current
day. At the present stage of system development, the spatial model simulates
Ken D. Brook and John O. Carter
Resource Sciences Centre, Queensland Department of Natural Resources (QDNR)
Brisbane, Australia
This article appeared in the June 1996 issue of Drought Network News.
pasture yields into the future an ensemble (5-10) of analogue years derived
from an analysis of the current Southern Oscillation Index phase (Stone,
1992). The mean and coefficient of variation of all relevant system-state and
accumulated variables from the ensemble runs are displayed as the final
output scenario. This system is also used to hindcast past scenarios to test the
accuracy of the system.
A limited evaluation of the system ran predictions ahead from October to
December (~90 days), and from October to the following March (~180
days). Spatial correlations for a 90-day forecast of total standing dry matter
varied from 0.41 in the exceptional El Niño year of 1982 to 0.75 in 1994. The
180-day predictions were worse, ranging from 0.18 in 1982 to 0.42 in 1994.
In general, only the 90-day prediction would be used.
Resources and time have not yet permitted an exhaustive test of the
predictability of all the model outputs, analysis of the spatial variance of the
simulation predictions (from coefficient of variation maps), a complete
testing of all years, or a comparison of different sources of predicted
meteorological information such as the Australian Bureau of Meteorology’s
own analogue system or General Circulation Models.
Percentile Outputs
The central aspect of the Australian government’s Exceptional Circum-
stances criteria for rural assistance during drought is that these criteria would
only be met a few times each century—that is, conditions are in the percentile
5-10 class or lower. This project aims to produce percentile views of
meaningful biological and agricultural variables. So it is possible to con-
struct a percentile view of grassland production and condition that is more
aligned with actual droughtedness than are rainfall percentile maps. This
month’s or season’s grass biomass can be compared with the last 30 or 100
years of biomass that would have existed at that location. Preliminary
analysis suggests that the year ranking according to rainfall may be
quite different from the year ranking from simulated percentile pasture
growth.
The current system can generate probability distributions of selected
variables (such as total standing dry matter) from 1957 on, but interpolated
daily rainfall data exist that would potentially enable the model to run from
the 1890s if raster surfaces of the other meteorological variables such as
temperature, humidity, and radiation could be derived.
The Current State of Operation
A prototype of a national spatial-modeling framework has been devel-
oped and demonstrated. This substantial technological achievement was
possible only through the excellent collaboration of a large number of QDNR
and interstate scientists across a number of disciplines. The system now
produces pasture biomass, percentile pasture biomass, ground cover, and
utilization maps for inclusion in Queensland’s applications for Rural Adjust-
ment Scheme Exceptional Circumstances assistance from the Australian
Commonwealth Government. Further refinement of the system in other
Australian states is required, and a second phase of technological develop-
ment and refinement of the system is expected to commence in 1997.
Acknowledgments
This work has been funded by Queensland Government Special Treasury
Initiatives and has also received assistance from the Australian Land and
Water Resources Research and Development Corporation.
References
Dudgeon, G. S.; C. S. Nilsson; and W. B. Fry. 1990. Rangeland vegetation
monitoring using NOAA-AVHRR data: 1. Improving the spatial and
radiometric accuracy of NDVI data; pp. 208–17. In Proceedings of the
Fifth Australasian Remote Sensing Conference, Perth, Western Australia.
Filet, P.; G. Dudgeon; J. Scanlan; N. Elmes; J. Bushell; M. Quirk; R. Wilson;
and A. Kelly. 1990. Rangeland vegetation monitoring using NOAA-
AVHRR data: 2. Ground truthing NDVI data; pp. 218–27. In Proceedings
of the Fifth Australasian Remote Sensing Conference, Perth, Western
Australia.
Fitzpatrick, E. A.; and H. A. Nix. 1970. The climatic factor in Australian
grassland ecology. In R. M. Moore, ed. Australian Grasslands; pp. 1-26.
ANU Press, Canberra.
Hutchinson, M. F. 1991. The application of thin-plate smoothing splines to
continent-wide data assimilation. In J. D. Jasper, ed. Data Assimilation
Systems; pp. 104–113. BRMC Research Report No. 27, Bureau of Meteo-
rology, Melbourne.
McCown, R. L.; P. Gillard; and L. A. Edye. 1974. The annual variation in
yield of pastures in the seasonally dry tropics of Queensland. Australian
Journal of Experimental Agriculture and Animal Husbandry 14:328–33.
McKeon, G. M.; K. A. Day; S. M. Howden; J. J. Mott; D. M. Orr; W. J.
Scattini; and E. J. Weston. 1990. Northern Australian savannas: Manage-
ment for pastoral production. Journal of Biogeography 17:355–72.
McKeon, G. M.; and M. Littleboy. 1996. The ‘Surfair’ GRASP Documenta-
tion Set. Internal DPI publication, in press.
Rickert, K. G.; and G. M. McKeon. 1982. Soil water balance model:
WATSUP. Proceedings of the Australian Society for Animal Production
14:198–200.
Smith, C. G. 1994. Australian vegetation watch. Rural Industries Research
and Development Corporation, Final Report, ref: DOL-1A.
Stone, R. C.; and A. Auliciems. 1992. SOI phase relationships with rainfall
in eastern Australia. International Journal of Climatology 12:625–36.
Tothill, J. C.; and C. Gillies. 1992. The pasture lands of northern Australia.
Tropical Grassland Society of Australia, Occasional Publication No. 5.
