Date of Graduation


Document Type


Degree Name

Master of Arts in Geography (MA)

Degree Level





Jason Tullis

Committee Member

Jesse Casana

Second Committee Member

Jackson Cothren


Earth sciences, Classification, Phenology, Time series


Land surface phenology describes events in a seasonal vegetation cycle and can be used in a variety of applications from predicting onset of future drought conditions, to revealing potential limits of historical dry farming, to guiding more accurate dating of archeological sites. Traditional methods of monitoring vegetation phenology use data collected in situ. However, vegetation health indices derived from satellite remote sensor data, such as the normalized difference vegetation index (NDVI), have been used as proxy for vegetation phenology due to their repeated acquisition and broad area coverage. Land surface phenology is accessible in the NDVI satellite record when images are processed to be intercomparable over time and temporally ordered to create a time series. This study utilized NDVI time series to classify areas of similar vegetation phenology in the northern Fertile Crescent, an area from the middle Mediterranean coast to southern/south-eastern Turkey to western Iran and northern Iraq. Phenological monitoring of the northern Fertile Crescent is critical due to the area's minimal water resources, susceptibility to drought, and understanding ancient historical reliance on precipitation for subsistence dry farming. Delineation of phenological classes provides areal and temporal synopsis of vegetation productivity time series. Phenological classes were developed from NDVI time series calculated from NOAA's Advanced Very High Resolution Radiometer (AVHRR) imagery with 8 × 8 km spatial resolution over twenty-five years, and by NASA's Moderate Resolution Imaging Spectroradiometer (MODIS) with 250 × 250 m spatial resolution over twelve years. Both AVHRR and MODIS time series were subjected to data reduction techniques in spatial and temporal dimensions. Optimized ISODATA clusters were developed for both of these data reduction techniques in order to compare the effects of spatial versus temporal aggregation. Within the northern Fertile Crescent study area, the spatial reduction technique showed increased cluster cohesion over the temporal reduction method. The latter technique showed an increase in temporal smoothing over the spatial reduction technique. Each technique has advantages depending on the desired spatial or temporal granularity. Additional work is required to determine optimal scale size for the spatial data reduction technique.