Mapping irrigated areas in Afghanistan over the past decade using

article i nfo Agricultural production capacity contributesto foodsecurity inAfghanistanand islargely dependentonirrigated farming, mostly utilizing surface water fed by snowmelt. Because of the high contribution of irrigated crops (N80%) to total agricultural production, knowing the spatial distribution and year-to-year variability in irrigated areas is imperative to monitoring food security for the country. We used 16-day composites of the Normalized Difference Vegetation Index (NDVI) from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor to create 23-point time series for each year from 2000 through 2013. Seasonal peak values and time series were used in a threshold-dependent decision tree algorithm to map irrigated areas in Afghanistan for the last 14 years. In the absence of ground reference irrigated area information, we evaluated these maps with the irri- gatedareas classifiedfrommultiplesnapshots of thelandscape during thegrowing seasonfromLandsat5 optical andthermalsensorimages.Wewere able toidentify irrigatedareasusing Landsatimagery by selecting asirrigat- ed those areas with Landsat-derived NDVI greater than 0.30-0.45, depending on the date of the Landsat image and surface temperature less than or equal to 310 Kelvin (36.9 ° C). Due to the availability of Landsat images, we were able to compare with the MODIS-derived maps for four years: 2000, 2009, 2010, and 2011. The irrigated areas derived from Landsat agreed well r 2 = 0.91 with the irrigated areas derived from MODIS, provid- ing confidence in the MODIS NDVI threshold approach. The maps portrayed a highly dynamic irrigated agricul- ture practice in Afghanistan, where the amount of irrigated area was largely determined by the availability of surface water, especially snowmelt, and varied by as much as 30% between water surplus and water deficit years. During the past 14 years, 2001, 2004, and 2008 showed the lowest levels of irrigated area (~1.5 million hectares), attesting to the severe drought conditions in those years, whereas 2009, 2012 and 2013 registered the largest irrigated area (~2.5 million hectares) due to record snowpack and snowmelt in the region. The model holds promise the ability to provide near-real-time (by the end of the growing seasons) estimates of irri- gated area,whichare beneficial for food security monitoring as well as subsequent decision making for thecoun- try.While themodelisdeveloped for Afghanistan, itcan beadopted with appropriate adjustments inthe derived threshold values to map irrigated areas elsewhere.