Dynamically Characterizing a Variety of Phenological Responses of Semi-Arid Areas to Hydrological Inputs using Multi-Year AVHRR NDVI Time Series and HYDRO1k-Type Terrain Parameters

This report explores the interaction of a variety of phenological responses in a semi-arid environment to hydrological forcing terms. We analyze the interannual departure of AVHRR NDVI values from normal seasonal variations in the context of local climate, hydrological and topographic controls. We have recently reported on several new techniques to smoothly interpolate and minimize noise in weekly and biweekly NDVI data for the purpose of characterizing both average annual variations and interannual variations in the time series of 1 km x 1 km pixels. These least-squares procedures tend to fit the upper data envelope to minimize the effects of atmospheric obfuscations, while minimizing model roughness to allow higher order models to better track subtle time variations in the raw data. We apply these techniques to mapping the dynamic behavior of multi-year NDVI time series from the NOAA- 14 and NOAA-16 satellites for a 150 x 150 km study area of the Great Basin in west-central Nevada. These time series, as currently produced and delivered by the U.S. Geological Survey's National Center for Earth Observation and Science, have been compensated for sensor drift and atmospheric water vapor, and in this area are georegistered to an accuracy of better than 0.5 km. Our study area provides a microcosm of a broad range of vegetation classes including irrigated agriculture with peak annual NDVI values of up to 0.7, semi-arid grasslands having NDVI values of 0.6 or higher during "wet" years, and non-vegetated playas (alkali salt flats) with typical annual NDVI values of 0.07. We show that the high precision of georegistration, combined with the higher resolution of our smoothing algorithms and the refined level of digital elevation models (DEMs) for this area, allows one to draw obvious and meaningful conclusions on the impact of elevation, hill slope and hydrology on a variety of local phenologies over the scale of the intermontane valleys. Adopting the USGS procedures for defining the HYDRO1k metrics of aspect, flow direction, slope etc., we refine the grid scale from the current HYDRO1k GTOPO30 DEM dimension of 1 km to a local DEM for our study area having a grid scale of 0.25 km. We employ higher-order 9 point finite differences to compute local topographic gradients, then aggragate (or integrate) the "HYDRO1k-type" parameters to the 1 km pixel dimensions of the NDVI data. We then perform a multivariate comparison of the derived-hydrologic parameters with characteristic phenological behaviors from the interannual NDVI modeled time series. For example, as one would expect, in spite of similarities of peak NDVI values in a particularly "wet" year, irrigated agricultural sites are well- discriminated from natural semi-arid grassland due to the multivariate controls from observed precipitation, surface water runoff, topographic slope, and the intrinsic fine structure in the behavior of the interannual NDVI time series. NDVI time series from montane areas provide interesting insight into the time of disappearance of snow cover, as well as the relation of summertime phenology to elevation and slope. A striking pattern emerges regarding the similitude between seasonal surface water runoff and interannual trends in phenology that corroborates the potential of NDVI data to monitor and characterize long term trends in the response of phenology to hydrological processes.