Water Journal : Water Journal November 2015
NOVEMBER 2015 WATER 37 Feature Article water use. The coef cients in the model were used to replicate actual results at the farm level for the 3,200 farms with usable metered data, and a good t of modelled water use compared to measured water use was achieved. The results were applied to the entire ILG to spatially distribute irrigation water use across the state. The model estimated that overall, agricultural lands are irrigated at about 0.40 metres/year, which is remarkably similar to the Australian Bureau of Agricultural and Resource Economics and Sciences (ABARES) estimates of Australia's national mean agriculture irrigation (0.45 metres/year). Note that Florida has an average rainfall total of around 1.27 metres/year. Historic estimates of Florida irrigation, as prepared by the water management districts, reported demand of about 0.55 metres/year. The model developed from the metered data estimated that farmers were actually using about 27 per cent less overall water than previously thought. AGRICULTURAL ACREAGE PROJECTIONS AND ASSOCIATED WATER REQUIREMENTS With estimates of current demand in hand, the challenge was to project irrigation behaviour 20 years into the future, in a rapidly urbanising state. The state required forecasts of both water use and its spatial location. Data was analysed to de ne trends for projecting future water needs; behaviour, a central theme of the study, is de ned by needs. Previous econometric modelling has identi ed spatially varying trends in behaviour including irrigation intensity, crop mix changes and land use conversion. Agriculture is an inherently risky operation and management behaviour takes a variety of risks -- from prices to labour shocks to weather -- into account. Projections were developed for future agricultural water use in ve- year increments for 20 years. Future scenarios were modelled using two parallel threads of information: future crop revenue expectation; and long-term trends in land use change. Land use change was tackled rst. In some parts of the state, such as the Miami metro area, agricultural land is likely to be smaller in 20 years, while in others -- namely the northern parts of the state -- substantial tracts of agricultural land are currently unirrigated. Irrigation practices from the western parts of the US, namely Texas and California, have shown evidence of migrating eastward over the past 30 years, as shown in Figure 2. Some areas of Florida that have traditionally been dry-land farming have seen new operators install sophisticated irrigation systems, and harvest two or more crops in elds that were historically single crop. Using agricultural census data that is collected every ve years, analysis was conducted at the county level to identify long-term trends. Techniques to detect the best- tting trends were applied and used to identify how much agricultural land is likely to be irrigated in each county, 20 years into the future. Areas where irrigation was projected to be added or removed were identi ed in GIS. R script was developed to automate the process. Figure 3 shows selected counties with the long-term irrigation trends, based on activity since 1987. Once the areas identi ed for likely future irrigation were identi ed, future water use was estimated. Long-term global revenue projections by crop type were used to estimate future crop returns. The econometric model was used to simulate future conditions, substituting long-term average weather trends for climate factors, and projected revenues for crop returns. Crop mixes have shifted in recent years as growers respond to subsidies, market shifts and structural agronomic changes. For areas forecast as newly irrigated, prevalent irrigation equipment and crop mix in the local Figure 2. Irrigation withdrawals, 1985--2010. Table 1. Regression variables. Variable Measure; all, at farm eld level NP Crop-speci c annual revenue; approximately 70 crops, aggregated to 12 crop categories Soils Soil type, aggregated to 8 broad soil categories PERIRR Percentage of permitted acreage that is irrigated IRR Type of irrigation system FF Variable for freeze protection L Vector of location attributes, including latitude/longitude coordinates and Water Management District variables RF Mean annual rainfall, based on nearest rain station ET Evapotranspiration, based on satellite data Figure 1. Map of agricultural lands and irrigated agricultural lands developed for modeling current and future water demands in agriculture.
Water Journal September 2015
Current Feb 2016