ASHRAE 4775-2005 Integration of ASOS Weather Data into Model-Derived Solar Radiation《整合ASOS气象数据 将其纳入模型中得出的太阳辐射RP-1226》.pdf

《ASHRAE 4775-2005 Integration of ASOS Weather Data into Model-Derived Solar Radiation《整合ASOS气象数据 将其纳入模型中得出的太阳辐射RP-1226》.pdf》由会员分享，可在线阅读，更多相关《ASHRAE 4775-2005 Integration of ASOS Weather Data into Model-Derived Solar Radiation《整合ASOS气象数据 将其纳入模型中得出的太阳辐射RP-1226》.pdf（15页珍藏版）》请在麦多课文档分享上搜索。

1、4775 (RP-1226) Integration of ASOS Weather Data into Model-Derived Solar Radiation Brian N. Belcher Arthur T. DeGaetano, PhD ABSTRACT Recent changes from manual to automated weather observing systems at most airports have introduced biases that influence building energy calculations. In particular,

2、these biases compromise the accuracy of model-derived solar radi- ation. This paper summarizes changes in methods and instru- mentation used to observe weather over the past decade, the consequences that these changes have on building energy calculations, and a new solar radiation model developed to

3、 help alleviate these consequences. This new model uses obser- vations from the US. National Weather Services Automated Surface Observation Systems (ASOS) to estimate global hori- zontal, direct normal, and diffuse horizontal solar radiation and allows for application to diferent climatic regimes wh

4、ile minimizing season-specijc and cloud-condition-specijc mean error biases. Model evaluation reveals that errors in solar radiation estimation are comparable to other contem- poravy solar radiation models. INTRODUCTION Over the past decade, observation methods and instru- mentation used to report h

5、ourly weather conditions have changed at most airports with the gradual implementation of automated systems such as the U.S. National Weather Services Automated Surface Observation Systems (ASOS). Users of weather data must be aware of the differences in observing methods and instrumentation, as wel

6、l as the biases these changes have produced, before using the data. One such application that is affected by these biases is the use of weather data to estimate solar radiation. While the demand for solar radiation measurements is high, budget constraints have often limited the establishment of netw

7、orks, resulting in sparse solar radiation observations. Due to this continual scarcity of data over time, the use of model estimates of solar radiation has been necessary for building energy calculations. The changes in weather observing practices must be accounted for in order to maintain sufficien

8、tly accurate model estimates. The objectives of this project were to present recent changes in weather observing methods, describe the influ- ences of these changes on observed weather data, and develop procedures to estimate global horizontal, direct normal, and difise horizontal solar radiation fr

9、om these data. In this paper, the analyses through which the objectives were met are summarized. The changes in observational instrumentation and methods instituted over the past decade that lead to biases are first discussed before an analysis of the biases themselves. Then, current methods of sola

10、r radiation estimation are discussed and necessary developmental changes are described. Finally, model evaluation results from cross-vali- dation procedures are presented. CONVERSION FROM MANUAL TO AUTOMATED OBSERVATIONAL SYSTEMS As automated observing systems were commissioned over the past decade,

11、 new instrumentation and methods of observation were introduced at stations that had previously used manual observation throughout their period of record (www.nws.noaa.gov/asos; awos.htm1; wwwl .faa.gov/atpubs/AIM). Since systems were designed for application to the aviation community, the extent o

12、f observations was limited to those that provide the most benefit to airport operation yet fall within budget constraints. Subsequently, the term “automated systems” refers specifi- cally to ASOS at stations that are currently operational. Brian N. Belcher is a research support specialist at the Nor

13、theast Regional Climate Center, Ithaca, New York. Arthur T. DeGaetano is director of the Northeast Regional Climate Center and associate professor in the Department of Earth and Atmospheric Sciences, Cornel1 University, Ithaca, New York. 02005 ASHRAE. 363 Changes in instrumentation Specifications an

14、d Observational Methods Numerous differences exist between current and previous observation practices. Manual observations are often instan- taneous assessments of weather conditions, without spatial limitations. ASOS measurement of ambient temperature and dew-point temperature are not instantaneous

15、 but rather aver- ages over the previous five minutes. The use of time-averaging results in less variability of observations through short time intervals and the resulting temperature-time series are consis- tent with the assumption that temperatures change smoothly and slowly over time. These avera

16、ges may be more represen- tative than instantaneous measurements during most weather conditions since a larger sampling frequency reduces obser- vational uncertainty. The difference between the use of aver- age and instantaneous measurements is largest during rapidly changing weather conditions; how

17、ever, a lag time on the order of minutes may be present in an ASOS temperature-time series during these conditions. The assessment of cloud cover has changed quite dramat- ically in both spatial and temporal extent, and these changes influence the accuracy of model-derived solar radiation. Although

18、automated systems do provide a higher vertical resolution (- 15 m, 50 ft) when compared to human perception (- 30 m, 100 ft), the vertical extent of automated observations is subject to ceilometer limitations (3840 m, 12,600 ft) but also to the different methodologies used. Analyses were performed t

19、o determine biases that may be influential to building energy calculations. Data Used in Calculating Instrument Biases. In order to determine these potential biases, it was first necessary to find concurrent manual and ASOS hourly observations with which comparisons could be made. Since these observ

20、ations were not available together at a single location, juxtaposed stations were used for comparison. Ten pairs of stations were selected throughout the US to perform this analysis (Table 1). Each pair was evaluated for one year of concurrent data, with each time period wholly occurring after METAR

21、 reporting began (July 1996). Analysis of Data and Bias Detection. Recent studies have shown negative biases in ASOS observations (ASOS values are lower) for dry-bulb temperature (Gunman and Baker 1996; McKee et al. 1997) and peak wind speed (Lock- Table 1. Station Pairs Used for Bias Analysis Green

22、ville, SC Topeka, KS Dallas, TX KGMU KGSP 13.0 7/1/97 - 7/1/98 KFOE KTOP 14.0 7/1/96 - 7/1/97 KDAL KDFW 18.6 1/1/97 - 1/1/98 I - I Buffalo, NY Gallup, NM KIAG KBUF 25.1 7/1/96 - 7/1/91 KGUP KRQE 29.6 1/1/99 - 1/1/00 I I St. Paul, MN I Miami. FL I KTMB I KMIA I 23.6 I7/1/96- 7/1/97 I KFCM KSTP 35.0 1

23、/1/97 - 1/1/98 I Denver, CO I KAPA I KDEN I 33.5 I1/1/97- 1/1/98 I 364 ASHRAE Transactions: Research 0.40 0.35 $ 0.30 v in O r 5 0.25 b 0 0.20 u p! 0.10 C 0.15 f 0- U 0.05 0.00 I than ASOS both vertically (above 3840 m, 12,600 fi) and hori- dr few sct bkn ove obs multi sky coverage Figure 1 Frequenc

24、y of all types of possible sky coverage reports using ten pairs of weather observing stations. Frequencies for yew, ” “sct, ” “bh, and “ovc ” sky coverages only include hours in which a single cloud layer is reported. Frequencies for “multi” include all hours in which more than one cloud layer is re

25、ported, regardless of the sky coverage of each layer. Also given arefrequencies for obscured (“obs 7 and clear (“clr 7 reports. The number of hours evaluated was 60475 for manual and 86533for ASOS observations. hart 1999). Similar results are obtained for dry-bulb tempera- ture and instantaneous win

26、d speed using the ten station pairs described. Using a different sample of stations, Guttman and Baker (1996) find that the magnitudes of dry-bulb tempcrature differences range between -2F and 0F (-lC and OOC), a result comparable to this analysis. Such biases can influence estimates of energy usage

27、 for building heating or cooling. Despite the difference in the type of wind measurements used between this analysis and Lockhart (1 999), a similar percent- age of stations (60%-70%) show a negative bias for ASOS measured wind speed, introducing a potential influence on structural design as well as

28、 building energy calculations. A tendency toward positive bias in ASOS measured dew-point temperature is also present; however, large ( 1 OF) negative biases are present at Buffalo, New York, and Denver, Colo- rado. Although not performed in this analysis, precipitation measured by ASOS has also bee

29、n shown to have a negative bias (McKee et al. 1997). The differences in the characteristics of cloud cover observations between manual and ASOS methods are impor- tant to evaluate for the purpose of solar radiation estimation, since both the amount of cloud cover as well as the height of cloud layer

30、s are strongly related to the amount of solar radia- tion observed at the Earths surface. ASOS observations report clear conditions nearly twice as frequently as manual observ- ers (Figure i). This difference can be explained since a manual observer is able to assess sky conditions to a greater exte

31、nt zontally (horizon to zenith). It is therefore logical to assume that nearly half of the hours reported as clear by ASOS are actually contaminated by cloud cover, either above the ceilo- meters vertical limit or near the horizon. These differences in cloud observation method also result in a highe

32、r frequency of manual cloud cover reports for most sky conditions (including reports with multiple cloud layers), with the exception of single-layer overcast skies. Similar biases are present when manual observations supplement ASOS observations, a prac- tice that is common at some larger airports.

33、Perez et al. (2001) document these reporting biases between adjacent large and small airports and attribute these biases at least in part to the presence or lack of augmentation by manual observers. In addition to sky coverage differences, the higher vertical resolution ofASOS allows for a greater a

34、bility to resolve cloud layers that are separated by small heights. In general, ASOS has a higher relative frequency of reports with cloud layers separated by less than 457 m (1500 fi). Despite the higher vertical resolution of ASOS, manual observers report slightly more occurrences of multiple clou

35、d layers than ASOS due to human ability to detect clouds beyond the spatial limitations of ASOS instrumentation. These differences between manual and ASOS reports can affect solar radiation estimation when dependent upon the number of cloud layers present. REVIEW OF AVAILABLE SOLAR RADIATION MODELS

36、Due to the scarcity of solar radiation measurements, a great deal of research has focused on the estimation of solar radiation. Various methods have established empirical rela- tionships between measured solar radiation and daily temper- ature range (Bristow and Campbell 1984; Goodin et al. 1999; Be

37、chini et al. 2000), sunshine hours (Revfeim 1981; Andretta et al. 1982), precipitation (Reddy 1987), or a combination of these and other variables including cloud cover, relative humidity, and wind speed (Ra0 and Bradley 1983; Winslow et al. 2001; Zhang et al. 2002). Of these variables, cloud cover

38、is the most direct factor associated with the day-to-day variabil- ity ofsolar radiation amounts reaching the surface ofthe Earth. Other studies have used cloud cover as a random variable to estimate solar radiation, based on probability distributions of variables (climatologies) in different region

39、s (Kasten and Czeplak 1980; Stanghellini 1981; Chen et al. 1983; Brinsfield 1984; Gautam and Kaushika 2002). While such a method is appropriate for estimating solar radiation over long time scales, the errors on an hourly or daily time scale are much larger than if observed conditions are utilized.

40、Algorithms that estimate solar radiation from data obtained via polar-orbiting and geostationary satellites have been derived and improved over recent decades. With a possi- ble resolution of approximately 1 km, these estimates can be very important in regions that lack desirable surface observa- ti

41、on coverage. The visible band of systems such as Geosta- tionary Operational Environmental Satellites (GOES) are ASHRAE Transactions: Research 365 used to detect cloud cover, which, in turn, can be used to provide estimates of solar radiation at the Earths surface (Gautier et al. 1980; Diak and Gaut

42、ier 1983). Additional methods of solar radiation estimation in recent years include sophisticated semi-physical models that use commonly observed surface weather variables to estimate solar radiation. These models generally incorporate solarposi- tioning algorithms in order to determine the amount o

43、f solar radiation reaching a horizontal terrestrial surface (without cloud or atmospheric effects). Parameters such as aerosol absorption, cloud transmissivity, and precipitable water content are then either estimated empirically or calculated through the use of physical equations before using them

44、to estimate global horizontal solar radiation for observed weather conditions. The National Renewable Energy Labora- tory (NREL) developed such a model (Maxwell 1998) to esti- mate hourly global horizontal, direct normal, and diffuse radiation for production of the National Solar Radiation Data Base

45、 (NSRDB). This model, however, requires cloud opacity information, which is no longer reported. DeGaetano et al. (1 993) and Petersen et al. (1 995) each developed regional semi-physical models based on the model of Meyers and Dale (1983). The model of DeGaetano et al. (1993), developed at the North

46、east Regional Climate Center (NRCC), provides slightly less biased estimates than the NREL model for a sample ofNortheast stations through incor- porating surface weather condition information into the model (fog, haze, etc.) and using aerosol and cloud transmissivities appropriate for the region. A

47、lthough this model does not calculate hourly direct normal or diffuse radiation, it does not rely on cloud opacity for radiation estimation, making it an ideal base for development of a new model using ASOS obser- vations. This base model (subsequently referred to as the “NRCC model”) is described i

48、n depth in the next section. Various methods exist for calculating direct normal and diffuse radiation, and each is considered here for inclusion into the base NRCC model. One approach involves methods such as that of Stuhlmann et al. (1990) and Pinker and Laszlo (1 992), which improved previous alg

49、orithms that use satellite data by accounting for scattering and multiple reflection between the surface and atmosphere. These improvements allowed for the estimation of direct and diffuse components of solar radiation from satellite data. Other methods determine appropriate empirical aerosol and cloud transmissivities for direct and diffuse radiation, in addition to those used for global radiation, within the context of a semi-physical model. Para- metric models such as the Iqbal model (Iqbal 1983) or the ASHRAE (1999) algorithm use meteorologi