Remote sensing is a valuable source of data that can provide a synoptic perspective critical for understanding biophysical relationships at a regional scale. Because of this, remote sensing has been a popular tool readily accepted into agricultural research and management. Since the launch of LANDSAT-1 in 1972, scientists and managers have been using remote sensing for crop identification, area measurements, yield prediction, and crop damage assessment. More recently, remote sensing has been seen as a source of spatial data for precision agriculture, although currently these systems are not widely operational. Remote sensing along with climate data and GIS technology can also be used for modelling _E and ma for regional analyses of water use efficiency. As fine to moderate remotely sensed data is now available free of cost, the use of remote sensing in agricultural management is more appealing than ever. The current availability of very fine spatial resolution data as well as the anticipation of hyperspectral data also broadens the scope of remote sensing and its usefulness regarding agricultural management.
The common thread in crop type identification applications is an attempt to achieve greater accuracy from remotely sensed data. In order to accomplish this, researchers have looked into various alternatives. Most of these alternatives have to do with the type of sensor (i.e., optical or microwave), number of images (i.e., single-date or multi-date), timing of the imagery, or processing technique. Although these characteristics certainly make a difference in the results attained, the trait that seemed to be most relevant was an appropriate use of the spatial data in combination with process understanding. Appropriate ground validation data and accuracy assessment is also critical for testing and reporting results.
Area measurement of crops from remote sensing is largely straightforward. However, positional accuracy and pixel size can both affect the results attained in this procedure. The scale of the remote sensing data, therefore, should be appropriate for the level of accuracy desired. This means that clear management objectives should be outlined prior to areal measurements. Individual areal measurements may vary more widely and be less accurate than summed (overall) measurements because errors of underestimation are offset by errors of overestimation when a number of areas are added together.
Strong correlations can exist between single date VI's or multi-date cumulative or integral VI's and crop yield. This relationship can provide early estimates of crop yield as well as allow for the assessment of crop damage. However, these remote sensing techniques are fairly limited because VI's tend to saturate rather quickly. They also are heavily dependant on the timing of the imagery acquired in relation to the physiological stage of the crop and can be influenced greatly by a lack of data due to things like cloud cover. The alternative crop simulation models can also provide fairly good predictions of crop yield alone. These models, however, usually simulate crop yield well for `normal' seasons, whereas variations from normal can create rather large errors. This is why using remote sensing to recalibrate simple growth models is often preferred over using either approach separately.
Thermal remote sensing can be linked with meteorological data, through resistance energy balance models, to estimate crop water use. For operational management of irrigated areas, having access to data with the required spatial resolution at the required frequency is an issue. Mapping ma prior to the irrigating of furrow or drip irrigated crops (i.e. those other than rice in southern NSW), would allow irrigation amounts to be better targeted. Regional estimates of WUE appear to be best developed using a regional GIS data `input-output' definition. Required data sets include rainfall, water application, and yields. These data sets are measured in NSW; for some, the main issue is their availability.
