DOE OSTI2008
The primary objective of this study was to analyze the spatial variability of temperature and relative humidity across Kansas (KS) and Oklahoma (OK) for sparse and dense networks by comparing data from (1) the Surface Meteorological Observing System (SMOS) installations at the Atmospheric Radiation Measurement (ARM; Peppler et al. 2007) Program's Southern Great Plains site and (2) the Oklahoma Mesonet (OKM; McPherson et al. 2007). Given the wealth of observations available from these networks, this study provided the unique opportunity to determine, within a quantifiable statistical limit, an optimal distance between stations deployed for observation of the climatological values of temperature and relative humidity. Average distances between a given station and its closest neighboring station for the ARM SMOS ({approx} 70 km) and the OKM ({approx} 30 km; Brotzge and Richardson 2003) networks provided an excellent framework for comparisons of sparse and dense observations (Figure 1). This study further lays groundwork for a future investigation to determine the necessary spacing between observations for initialization of gridded numerical models. The spatial variability of temperature and relative humidity was examined over KS and OK by comparing observations between station pairs located in three primary domains: (1) a sparse domain in KS, consisting only of ARM SMOS stations; (2) a dense domain centered in northern OK, consisting of both ARM SMOS and OKM stations; and (3) a dense domain centered in central OK, also consisting of both ARM SMOS and OKM stations (Figure 2). In addition, the ARM SMOS stations in OK were utilized to create two secondary sparse domains. Before the observations were compared, quality control (QC) beyond the standard ARM range test was added through implementation of tighter range tests specified by data quality objectives (DQOs). Furthermore, instances of poor-quality data were removed from the data set on the basis of ARM data quality reports (DQRs). Finally, to account for spatial differences in terrain, temperature observations were corrected to mean sea level by using a standard lapse rate of 6.5 C km{sup -1} and the elevation of each observing station. For the comparison, a central station was chosen in each domain. Observations during the time period 2004-2006 from each of the other stations within a respective domain were compared to those from this central station. The Pearson correlation coefficient ({rho}) and root-mean-square difference (RMSD) were the statistics used to quantify the relationship between station pairs. For each domain, the {rho} and RMSD values were plotted against the distance separating each station pair, and a least-squares (LS) regression line was fitted to the values. The regression slopes and intercepts were compared between the various domains. The results of this analysis demonstrated positive correlations between all individual station pairs for both temperature and relative humidity. In addition, the {rho} and RMSD values for both temperature and relative humidity exhibited, in general, a linear relationship with distance from a central station. The calculated slope and intercept values were comparable across most domains, and spatial differences in temperature were smaller than those for relative humidity. The findings suggest that although the sparse networks studied might provide an accurate spatial representation for climatological values of temperature and relative humidity over the specific distances between stations, the relative importance of the temperature and relative humidity observations is a critical consideration in network design.
54 ENVIRONMENTAL SCIENCES↗