Laserfiche WebLink
CRDSS <br />TASK MEMORANDUM 1.05-6 <br />Study of System Integration Issues <br />Database - Tools for Data Filling and Manipulation <br />1.0 ISSUE <br />Time-series (TS) type data contained in the historical database are needed to calibrate both the water <br />rights planning and consumptive use models. The quality of the real-time management system is <br />directly related to the integrity of the TS calibration data. This memorandum discusses the TS data <br />filling and manipulation tools to be implemented in the CRDSS. <br />2.0 DISCUSSION/ANALYSIS <br />TS data consist of time-stamped data, usually recorded at constant data intervals, e.g., daily, <br />monthly, yearly. This type of data commonly contains missing values or gaps where data have not <br />been recorded. It will be important to identify which TS should not contain missing data. <br />Depending on the use of the data, some TS can contain missing data. For example, a base network <br />of meteorological stations might be selected. Stations that are not part of this network would be used <br />to estimate missing data for stations in the basic network. Also, TS not required to support the <br />modeling tasks will be allowed to contain missing data. <br />TS data will be displayed and data gaps identified through a GUI, as addressed in Task <br />Memorandum 1.05-18. TS data will also be analyzed with software developed at RTi that produces <br />a summary of the data as well as a statistical analysis of the TS. Once data gaps are identified and <br />characterized, appropriate steps to fill the gaps will be taken. <br />Meteorological Time-Series <br />Data gaps for meteorological TS data (precipitation, temperature, snow, relative humidity, solar <br />radiation, etc.) will generally be filled by using information from surrounding stations. Surrounding <br />stations will be weighted based on their distance and difference in elevation from the station with <br />missing data. The technique used to estimate missing data from surrounding stations will take into <br />account the differences between station means by adjusting the data based on the ratio of station <br />means for a selected historical period. To compute station weights, stations will be categorized, for <br />example by quadrant, in relation to the station being estimated. This will ensure that any directional <br />effects in the data are taken into account by weighting stations from all points of the compass. <br />If there are no stations surrounding a station with missing data, the data will be stochastically <br />generated using a technique that conserves the station mean, probability density function, and <br />spectral components of the TS. This procedure will be handled on a case-by-case basis. In general, <br />the station statistics will be estimated from available station data, data from other stations, and <br />regional relationships based on physical characteristics. In some cases annual or seasonal data may <br />be available when monthly or daily data are required. In these cases the longer-term data will be <br />disaggregated, preserving the shorter period statistics. <br />1 <br />A275 05.10.94 1.05-6 Ritsch, Day <br />