Earth observations for irrigation management : Swaziland's sugar industry

Sugarcane production uses over 90% of Swaziland's water resources which are connected either downstream or upstream with neighboring Mozambique or South Africa. This first ever country-wide irrigation performance assessment of the sugarbelt was conducted using open-source satellite remote sensing data to try and identify well-performing 'bright spots' and low-performing 'hot spots' growers and provide sufficient information on how to improve irrigation performance and productivity through improved water management practices. Ready-made, spatially distributed precipitation and ET maps were combined with secondary data to quantify equity, adequacy, reliability and water productivity. TRMM 3B43 Version 7 and CHIRPS Version 1.8 precipitation products were calibrated based on Geographic Difference Analysis (GDA) and independent rain gauge station measurements ware used to validate the products. Un-calibrated CHIRPS was found to be the best performing precipitation product with a RMSE=10%, MAPE=7% and a Bias of 6 mm/year. Despite the minimal computational requirements, the formulated CMRSET and SSEBop mean ensemble product fairly reflected ET spatial variation across the study area. The annual ET of irrigated sugarcane in Swaziland averaged from November 2005 - October 2010, ranges from 582 to 1,259 mm/year with an average of 1,058 mm/yr and 80 mm/year standard deviation. High ET pixel values corresponded to high production areas managed by Millers and Estates and lower ET pixel values were predominately on small and medium scale growers managed by Individual Growers. This research investigated over 300 sugarcane growers by segmenting them into three clusters: (i) according to land holding size (ii) according to management regime (ii) according to the location of the irrigation schemes. Five growing seasons, from 2005/06 to 2009/10, were investigated and the overall irrigation performance of the sugarbelt was expressed using the average coefficient of variation for all irrigation performance indicators. The results showed that large scale growers, managed by Miller-cum-planters and Estates are the 'bright spots' and small and medium scale growers, managed by Individual Growers and some Farmers Associations are the 'hot spots'. Most of these under-performing growers are located at Big Bend, Siphofaneni, and Vuvulane whilst North-based growers (RSSC and Tabankulu) are the 'bright spots' in the sugarbelt. The average industry scope of improvement was found to be 23% which implies a possible 23% increase in sugarcane yield while maintaining the same ET. Alternatively, the 23% can be perceived to be indicative of the amount irrigation water resources that can be saved while maintaining the current production levels. Water productivity in the sugarbelt can be better improved by focusing on interventions aimed at increasing sugarcane yield levels other than ET. A comparison of water abstraction data with direct withdrawals derived from remote sensing revealed remote sensing to be a promising alternative tool for assessing withdrawals in Swaziland. The average irrigation depths obtained from flow measurements was 621 mm/year and 516 mm/year was calculated using remote sensing data. The new insight provided by this research on irrigation performance in the sugarbelt is pivotal for informing intervention directed towards increasing productivity, especially on small and medium scale growers. It was concluded that the use of globally available, open access, ready-made ET and precipitation products enabled the timely and fairly accurate estimation of the spatiotemporal variation of ET and rainfall utilized in performing this comprehensive countrywide irrigation performance assessment. These products are not only cost effective and have minimum computational requirements, but without these data and financial requirements for conventional irrigation performance studies is prohibitive for developing countries like the Kingdom of Swaziland