Remote sensing of irrigated crop types and its application to regional water balance estimation

Abstract

The strengths of moderate to coarse resolution satellite remote sensing in both identifying crop types and estimating crop area has resulted in the widespread use of this technology for agricultural monitoring. Although the spectral information and cost of these remote sensing data are attractive, their spatial resolutions are often perceived as being inadequate for agricultural management at both the individual holding and the paddock level in the rice areas of New South Wales (NSW). Conversely, fine resolution remote sensing (e.g., aerial photography) very often contain spatial detail that will allow management decisions to be made at the paddock level, but these data can be expensive to acquire and subsequent manual digitisation of crop areas is labour intensive when performed each year. This raises at least two associated research questions for the rice industry in southern NSW: (1) ‘how is the rice area best mapped when considering cost, accuracy, timing, and complexity while reconciling the above issues? ‘; and (2) ‘how can spatial accuracy (concerning both areas and positions) be measured and related to relevant management practices in order to influence decisions?’. Additionally, many operational users of remote sensing data perceive it as being an overwhelming data source as it often requires time consuming training and expensive computer software. This results in a further series of issues: (3) ‘can remote sensing be used operationally within the NSW rice industry so that simple methods can be applied using inexpensive software with minimal training in order to achieve similar or increased accuracies?’. Furthermore, use of spatially accurate GIS paddock boundaries has been shown to increase crop classification accuracy. However, this raises further questions: (4) ‘what is the influence of spatial error on management decisions?’; (5) ‘how can the accuracy of GIS data be measured?’; and (6) ‘how are these issues altered when considering the other major summer crops in the region?’. As satellite hyperspectral data (e.g., >100 spectral bands per image) are now available this again raises some questions, such as: (7) ‘does this extra spectral information content translate into additional or more accurate agricultural metrics’; and (8) ‘what is the current capacity in the rice industry of NSW to process this sort of information quickly as to impact management decisions?’. These and other related issues have made up the vast majority of the research from project 1105. Recommendations have been made wherever possible regarding the improvement of spatial analysis or mapping efficiencies. Importantly, the research from project 1105 has been adopted by the local industry – this is proof of ‘impact’ as opposed to only producing ‘outcomes’. The work reported here has concentrated on practical issues with an emphasis on transferring the knowledge gained to industry partners. Prior to addressing these issues, a comprehensive literature review concerning the utility of remote sensing in rice base irrigation systems was performed to ensure that past, present and current opportunities (and constraints) concerning the use of time series remote sensing in the local, national and international context were known and understood. Due to wanting to optimise research results by acquiring as many images as possible with our operating budget for image acquisition all new research (as opposed to the literature review) was conducted on the smallest irrigation areas in southern NSW: Coleambally Irrigation Area (CIA). Before methods can be transferred to the other irrigation areas (i.e., Murrumbidgee and Murray Valley Irrigation areas) some assessment of the similarities of the irrigation systems in terms of non-rice crops and their phenology needs to be performed.