Laserfiche WebLink
With this method, data from the dependent time series (new station) are used during any period <br />where the two stations have overlapping data. TSTooI assigns the name and ID of the dependent <br />time series to the combined station. The combined stations are indicated as such in the TSTooI <br />input command files provided in Appendix B. <br />2. Investigate and Determine Appropriate Methods for Filling Missing Monthly Data <br />Daily climate data stored in HydroBase are obtained from NOAA. Consistent with NOAA's <br />approach, monthly average temperature and total precipitation values in HydroBase are <br />calculated from daily data. If there are nine days or less missing within a particular month, a <br />monthly value is calculated based on the available daily data. If more than nine days of data are <br />missing in a particular month, no monthly value is calculated (the monthly value is missing). <br />Frost dates in HydroBase are calculated using NOAA's method of determining frost dates from <br />minimum daily temperature data. NOAA utilizes the date of June 30th as the end of spring, and <br />July 1st as the beginning of fall for frost date analysis. There may be climate stations located in <br />the higher altitudes where a detailed investigation of local irrigation practices, growing seasons, <br />and minimum daily temperature data show that a different date may be more appropriate in <br />determining the last spring and first fall dates. However, it is our recommendation that the June <br />30th and July 1st dates be used for the basin-wide SPDSS analyses. <br />Missing monthly climate data were filled using techniques adopted in previous DSS modeling <br />efforts, which generally involved filling missing data using regression with stations located in <br />close proximity. Prior to filling, a cursory review of daily climate data was performed. During <br />this review, we encountered some extreme values, e.g. 900.0° Fand -60.0° F daily temperature <br />values. These extreme values were removed from the analysis and considered as missing data. <br />2.1 Temperature and Precipitation <br />The following four regression techniques can be easily applied in TSTooI for temperature and <br />precipitation data: <br />^ Annual regression (one equation) with linear data transformation <br />^ Annual regression (one equation) with logarithmic data transformation <br />^ Monthly regression (twelve equations) with linear data transformation <br />^ Monthly regression (twelve equations) with logarithmic data transformation <br />Each regression technique was tested to determine which predicted data that best matched <br />historical data (minimized the residual). As described when filling streamflow data in the <br />SPDSS memo, "Task 2 -Identify Key Streamflow Gages and Estimate Streamflows for Missing <br />Records", the correlation coefficient (R) was not used as the factor for deciding the best <br />regression technique because the coefficients from twelve monthly equations cannot be <br />compared directly to the coefficient resulting from one annual equation. <br />The monthly linear regression technique provided for the smallest residuals between predicted <br />and historical data for both temperature and precipitation data. The R-values were generally <br />high with respect to temperature data but significantly lower with precipitation data. An <br />example of the accuracy of predicted data using monthly linear regression can be seen in Figures <br />3and4. <br />Task 53_2.doc 7 of 17 <br />