A Remote Sensing-based Biophysical Modelling Approach for Estimation of Crop Yield in Irrigated Agriculture for Australian Conditions

With rising world population, irrigated agriculture faces the problem of
accommodating increases in demand for food while maintaining the
sustainable use of limited water resources. By 2030, demand for cereal for
human and animal consumption is expected to increase by 50% from its
level in 2000. Such high demand must be met despite less water being
available for irrigated agriculture. Irrigated agriculture accounts for 40% of the world's food production and uses over 70% of consumptive water.
However with climate change and the pressure of population, there is no
possibility that this level of water usage can be sustained into the future and
thus there is an urgent need to improve water use efficiency and water
productivity in the irrigated agriculture sector. In order to achieve improved
efficiency and productivity, accurate crop yield estimation is crucial. This
will also contribute to improved water resource management, food security
at national and international levels, and food trade planning.

Many biophysical models have been developed for the estimation of crop
yields in Australia and globally. For example, the Maizeman model was
developed in Australia for estimating the yield of maize crops at farm scale
while the ORYZA2000 model has been widely used for the estimation of
rice crop yields. However, these models are only able to provide crop yield
information at the farm scale and are not able to provide total crop yield
production data at the irrigation system or catchment/basin scale. A number
of research projects have been undertaken to develop catchment-scale yield
models for forecasting the crop yields of different crop types using state-ofthe-art remote sensing techniques in many countries. However, there have
been very limited applications of remote sensing of broad acre irrigated
crops grown in Australian conditions. This study has tried to estimate the
spatial variability of crop yield using a remote sensing modelling technique
in order to fill this gap in the research and to provide a practical and accurate model over the Coleambally Irrigation Area located in the Riverina
region of NSW.

This study focused on yield estimation for maize/corn, rice and wheat, three
crops commonly grown in the Coleambally Irrigation Area (CIA) which is
located in the Murray Darling Basin. This study developed CIA-specific
relationships between the Leaf Area Index (LAI) and the Normalised
Difference Vegetation Index (NDVI) using satellite and on-ground
measurements for the three crops grown in various soil types. The novel
biophysical models developed are able to use the estimation of groundbased
LAI in the CIA. In addition the biophysical models, incorporated
with the object-oriented modelling technique, have been used for classifying
the three irrigated crops in the area. In the classification phase, high and
medium level spatial resolution optical satellite imagery was used in order
to obtain highly accurate classification maps. The overall accuracy of the
classification map obtained in this research was 78% for winter 2010 and
85% for summer 2010/11. The average producer and user accuracies for
winter were 79% and 78% respectively and for summer 85% and 86%
respectively.

The biophysical models and classification maps developed were integrated
to extract biomass growth over the cropping season in the CIA. By
integrating biomass production with the crop specific harvest index, the crop
yield for each crop for the two seasons, winter and summer, was estimated a
few weeks before the harvest. The estimation of crop yield for corn, rice
and wheat were in 93, 91% and 86% in agreement with the published data.
In conclusion it can be stated that the general reliability and accuracy of this
robust method is promising and the method has proved to be an effective
tool in regional yield estimation in Australian conditions. The method
proved to be stable and accurate for operational use for crop yield
estimation at the irrigation system level across Australia.