GIS and Remote Sensing Technologies for Efficient Agricultural Water Use

Professor Shahbaz Khan
CSIRO Land and Water Division,
Locked Bag 588, Wagga Wagga, NSW 2678

International Centre for Water,
Charles Sturt University, Locked Bag 588, Wagga
Wagga, NSW 2678, Australia
Shahbaz.khan@csiro.au

Cooperative Research Centre for Irrigation Futures

Abstract
Efficient water use is the key for sustainable management of water resources. In inefficient rice growing areas, groundwater recharge is caused by inefficient irrigation practices, high leakage rates from light textured soils, poor surface and groundwater drainage and inappropriate crop rotations. Shallow watertable and secondary soil salinisation have a major impact on the long term sustainability of these areas. Land and Water Management Plans have been developed in these irrigation areas to address sustainability issues by implementation of on-ground works and community actions to control waterlogging and salinity. The proposed on-ground works and community actions can achieve the desired benefits only if their effectiveness can be assessed prior to implementation, restrictions on rice areas can be quantified and resulting changes in environmental conditions can be measured on the landscape.

This paper describes how GIS databases, remote sensing and hydrological modelling techniques are helping land and water management actions in the rice growing regions.

1. Introduction
The major rice based irrigation areas in Australia are situated in southern NSW comprising of a total area of around 150,000 hectares situated in the Murrumbigdee, Coleambally, Murray Valley Irrigation areas and districts and along the rivers and creeks (Humphreys, 1999). The rice growing areas are experiencing rising watertables and soil salinity, which threaten the sustainability of irrigated agriculture. Due to the limited unconfined storage, discontinuous nature of underlying aquifers and limited regional groundwater discharge it is necessary to limit rice growing to suitable lands which do not effect surrounding areas significantly. Currently rice growing area restrictions are in place to reduce recharge to groundwater (Humphreys et al, 1994).

In addition to rice area restrictions, Land and Water Management Plans (LWMPs) are currently being developed and implemented to address sustainability issues. These plans enlist specific sustainability targets for a 30 years period such as:

Sustainable productivity of farms
Achievement of on farm water balance bench marks
Extent of saline areas to be less than a certain proportion of total irrigation area by a certain number of years
Levels of salinity in the drainage waters at key locations

In order to achieve these targets a number of actions such as adoption of best management practices, growing of rice on only suitable lands, on-farm and regional management activities and education and extension endeavours are planned.

Currently rice growing areas in the Murray Valley and Irrigation Districts are being monitored using satellite images and GIS methods. The GIS offers unique opportunities to integrate spatial data from different sources with the natural resources management models (Goodchild, 1993). Digital description of these rice growing areas with the hydrological models in a GIS environment can be used to assess and differentiate climatic and management impacts on shallow watertables and soil salinity. These models can also be used to evaluate the local and downstream impacts of a number of management concepts e.g. drainage, conjunctive water use and sustainable hydraulic loading.

2. Measurement of Rice Growing Areas
Barrs et al. (1994) gave details of rice classifying methods using satellite imagery, with varying degrees of accuracy stated. De Soyres (1989) showed 25-50 percent reduction in costs over the conventional techniques for medium-scale line maps and further reductions in costs for digital elevation mapping, ortho-images and spatiomaps when satellite images (Landsat) were used. Gastellu-Etchegorry (1990) analysed SPOT and Landsat capabilities for spatial feature determination and concluded that Landsat-MSS (TM) data (30 m pixel size) can allow identification of features larger than 5 hectares and whereas SPOT-XS (P) data (10 m pixel size) can allow analysis of features larger than 0.16 hectares for length to width ratios less than 4.

Prior to 1996/97 irrigation season, Murray Irrigation Ltd (MIL) were using hard copy aerial photographs and planimeters to measure rice crop areas. To record the history of areas sown to rice, measured areas were transferred to a file system by hand colour coding for every irrigation season thus a drawing or photocopy was produced for each of the 1580 rice growing farms. At this stage a visual comparison was done to verify the crop was grown on suitable soil. Measured areas were then keyed into a data base and hydraulic loading (amount of water applied per unit area) was calculated using meter readings from dethridge water measuring outlets. This entire process used to take around 6 to 8 weeks for 3 people to complete. MIL has now adopted digitisation of satellite imagery for quantification of rice areas.

Timing of rice growing activities is dependent on crop variety and climate. Rice bays are filled and sown from late September to early December. The crop emergence varies due to climate, variety, water management and turbidity. Draining begins late February for early varieties and harvest can continue into June. The purpose of rice area measurement is to assess whether rice is being grown on the suitable land and within the allowable rice area limits Therefore irrigation companies need to measure the total amount of ponded water regardless of crop density or coverage.

The optimum timing for satellite data acquisition for rice area measurement is mid December to early February. The Murray Irrigation Area is split across two Landsat flight paths. The revisit interval of Landsat could mean a large change in crop development between scenes or data could be missed entirely due to cloud cover.

Off nadir viewing capability of SPOT gives excellent revisit opportunity during periods of cloud cover. SPOT imagery has higher spatial resolution and could be used by the irrigation company for other larger scale mapping applications.

Two remote sensing methods were considered for rice area measurement:

Classification of Landsat imagery with Normalised Vegetation Index (NDVI).
On screen digitising of Spot Panchromatic imagery

Identification of Rice Paddocks Using Landsat Data
Landsat band 3 (Red, 0.63-0.69 µm wavelength) and band 4 (Near Infrared, 0.76- 0.90 µm wavelength) were used to construct normalised vegetation index (NDVI) i.e.

NDVI=(Red-Near Infrared)/(Red+Near Infrared)

Because vegetation has a low visible reflectance and high near infrared reflectance, by using this index water bodies appear black and high vegetation areas appear brighter than lower vegetation areas.

Results of NDVI are given in Fig-1. On a farm scale, it appears that after classification of rice areas, the data could still require a considerable amount of manual editing. Errors of commission can occur in the classification process with the inclusion of other water bodies such as storage dams and waterways.

On Screen Digitising of SPOT Panchromatic Imagery
The climate in MIL area and timing of imagery acquisition results in a large contrast between irrigated and non-irrigated areas making visual identification of rice crops very easy from the SPOT panchromatic data. (Fig-2)

ESRI’s Arcview software was selected and scripts were written, in house, to streamline the digitising process. The operator is able to type in a landholding reference number and the program zooms to that area of interest. The rice area is identified visually and digitised on screen. The program then writes that area to the landholding selected. Identical areas measured in previous years do not have to be re-digitised; they can be added to the current year by clicking on a button then clicking on that area. This function has greatly reduced the time taken to measure crops each year. MIL now has five seasons of rice growing captured digitally. (Fig-3)

The rice growing areas can be overlaid on soils maps and electro-magnetic surveys to identify leaky paddocks, which can help reduce groundwater recharge to shallow watertables. Other uses of digital rice area data are to provide ready crop statistics, crop approval and environmental reporting. The spatial distribution of rice areas provides input to the spatially distributed hydrologic models, which are described in the next section.

Figure-1 Normalised Vegetation Index Image

Figure-2 On Screen Digitising of Spot Panchromatic Imagery

Figure-3 ArcView GIS database of Rice Growing Areas

3. Conceptual Models of Irrigation Systems
Applications of GIS and hydrological models for natural resource management have been described by several authors including Bradley (1993), Alaric (1994), Lilburne et al. (1998) and Belmonte et al. (1999). This section describes how GIS, remote sensing and hydrology are being integrated for the environmental management of rice growing areas.

A conceptual model is developed to explain the hydrogeology of the irrigated regions. It illustrates the geologic formations, hydrological flows in and out the region and the capacity of the aquifers. The structural contours of geologic formations based on hydrogeological maps of Australian Geological Survey (Hennessy, et al., 1994) are digitised in the GIS environment. The spatial reference for this spatial database is Zone 54, Zone 55 and Datum is Australian Geodetic 1984.

Standard procedure in the development of a 3D conceptual model includes spatial data acquisition, data processing in ArcGIS, and model building in ArcScene.

A vast quantity of data/information from different data sources goes into the construction of a conceptual model. ArcGIS serves as a tool that enables multiple types of data to be integrated for both analytical and visual analysis. ArcGIS’s 3D Analyst adds functionality to ArcMap to provide three-dimensional visualization, topographical analysis, and surface creation capabilities. ArcScene also allows advanced spatial data visualization and interaction in a 3D environment.

Following major spatial data/datasets have been collected:

DEMs
Remotely sensed images
Rivers
Channels
Localities
Irrigation area boundaries
Watertable levels
Geological formations

Original data available for use in the 3D conceptual model are of widely varying spatial reference, scales and accuracy. They are all processed to a geographic coordinate system (WGS84) in ArcGIS. Colour composite of Landsat 7 ETM+ images (RGB 742) are used to drape onto the DEM (250m resolution) for representing the ground level surface (e.g. landscape and crops). Model domain and other boundary conditions are defined by irrigation boundaries, channel and stream network and major localities. Geological framework is formed by the top elevation of geological formations. The structural contours of them are firstly digitised from the hydrogeological maps of Australian Geological Survey, and then interpolated into representative geologic surfaces. Groundwater depth data are incorporated to show the subsurface watertable level. All the above data layers are then overlayed in ArcScene. The hydrologic characteristics and properties are assigned to each layer to illustrate/reflect system losses and gains, regional groundwater flow, vertical interactions between the formations, recharge due to irrigation and rainfall, leakage to and from the drainage channels, surface-groundwater, agriculture and environment interactions, and so on. A vertical exaggeration is applied for presentation emphasis. Interactive perspective viewing can be performed to the model. The final 3D conceptual model can also be animated fly-throughs, and exported for presentation and analysis.

4. Evaluation of Land and Water Management Options
The conceptual models are used to develop calibrated surface and groundwater models which enable integration of biophysical processes with crop production and economics components for different management scenarios. These scenarios can be modelled in consultation with local groups to enable ready adoption of modelling results. The models and framework are capable of simulating following scenarios at the farm and irrigation area levels:

the do nothing case (i.e. no change);
change in agricultural enterprise;
change in irrigation water quality;
change in volume and quality of drainage water;
change in water allocation; and
change in commodity pricing, water pricing and other costs including environmental costs.

The integrated decision support system allows identification of rice areas, areas of higher recharge and leaky channel and river reaches. Overlaying the surfacegroundwater interaction model results with the GIS model grid and channel network layers can help identify groundwater hotspots in the irrigation areas (Fig-5).

Figure-4 Conceptual Model of the a Rice Growing Irrigation System

Figure-5 Surface network features overlaid on model grid

Groundwater model simulations give predictions of watertable heights under different scenarios. Spatial distribution of watertable heights combined with the aquifer hydraulic properties can be used to derive groundwater flow vectors. The groundwater flow vectors can be used to identify groundwater recharge and discharge zones to understand how management actions at farm level affect watertables in other areas (Khan et al, 2000 a & b). Therefore groundwater models coupled with GIS databases provide a powerful tool for the environmental management of irrigation areas.

5. Conclusions
The following conclusions are drawn from this work:

Using GIS for rice monitoring has improved the efficiency of crop measurement and the associated administrative processes, e.g., customer inquiries, information searches, approval for growing crops, crop statistics and information presentation.
GIS has enabled other doors to open as new technology arises. The adoption of EM survey for soil suitability in rice bays has been possible by importing surveys directly into GIS and overlaying them on image data to pinpoint approved and unsuitable rice growing areas.
Integration of GIS databases with conceptual and dyanmic models is helping identify on farm and regional impacts of irrigation management practices.
Visualisation capabilities of GIS helps community interaction with the modelling scenario outcomes and therefore provides a useful mechanism for acceptance of complex modelling results by the farmer groups.

Acknowledgements

The author is grateful to CRC for Sustainable Rice Production, CRC for Irrigation Futures, CSIRO Water for a Healthy Country Flagship and Murray Irrigation Limited for providing assistance with the various aspects of this work. Technical inputs from a number of colleagues including Dr Yun Chen, Dr Akhtar Abbas and Tariq Rana are acknowledged.

References

Alaric V. (1994). Realizing the potential of remote sensing and GIS in ecosystem management planning, and policymaking. In: V., Alaric, (Ed.), Remote Sensing and GIS in Ecosystem management, Islands Press, Washington, pp 346-352
Barrs H. D. and Prathapar S. A. (1994). An inexpensive and efficient basis for monitoring rice areas using GIS and remote sensing. Aust. J. Exp. Argic. 34:1079-83. Belmonte A. C., Medrano Gonzalez J., Vela Mayorga A., Fernandez C. S. (1999).
GIS tools applied to the sustainable management. Agricultural Water Management, 40:207-220.
Bradley R.L. (1993). Geographical information systems for agricultural decision support. Agric. Eng. 48: 102-105.
Gastellu-Etchegorry J. P. (1990). Remote Sensing in Indonesia. In: Gastellu- Etchegorry, J. P. (Ed.) Satellite remote sensing for agricultural projects. World Bank Technical Paper No. 128.
Goodchild M. F. (1993). The state of GIS for environmental problem-solving. In: Goodchild M. F., Parks B. O., Steyaert L. T. (Eds.), Environmental Modeling with GIS. Oxford University Press, pp 8-15.
Humphreys E. (1999) Rice water use efficiency in the southern Murray-Darling Basin. Proceeding of Rice Water Use Efficiency Workshop, Cooperative Research for Sustainable Rice Production. March 1999. 3-11.
Humphreys E., van der Lely A., Muirhead W. and Hoey D. (1994). The development of on-farm restrictions to minimise recharge from rice in new south wales, Australian Journal of Soil and Water Conservation 7(2): 11-20.
Khan S, Xevi, E., O’Connell N., Madden J. C., and Zhou F. (2000a). A farm scale hydrologic economic optimisation model to manage waterlogging and salinity in irrigation areas. Proceedings of the Fourth Biennial Engineering Mathematics and Applications Conference, EMAC 2000, RMIT University, Melbourne September, 2000. 179-182.
Khan S., Short L. and Best L. (2000b). Groundwater dynamics in the CIA: CSIRO Consultancy Report to the Coleambally Irrigation Corporation Limited.
Lilburne L., Watt J., Keith V. (1998). A prototype DSS to evaluate irrigation management plans. Computers and Electronics in Agriculture, 21:195-205.