Wednesday, 13 January 2016: 11:15 AM
Room 228/229 ( New Orleans Ernest N. Morial Convention Center)
A majority of atmospheric and land surface variables are observed as time series, i.e. evenly spaced in time and measured successively. In some cases, meteorological observations are non-stationary, because sensors are moved through space as meteorological conditions change. Examples include surface temperature observations from MODIS/ASTER (MASTER) overflights or microclimate variables from mobile transects. Statistical methods to analyze time series generally assume stationarity of the underlying dataset, which can be achieved through time-detrending. The most common technique to detrending time series from mobile transects is curve fitting by linear regression, subtracting a least-squares-fit straight line as a function of time from each observation to yield a zero mean residue. Detrending surface temperatures from mobile transects using a global linear function over all observations introduces errors, as heating and cooling rates of surfaces vary significantly by surface type and sun-exposure. In this presentation, we investigate the sensitivity of linear time-detrending to shaded and non-shaded surface types. As non-stationary time series, we use a mobile transect of Power Ranch, a master-planned community in Gilbert, Arizona, USA. A golf cart performed 1.5 hour transects every 3 hours from 6 AM to 9 PM on August 13, 2015, a heat wave day, to sample tree-shaded and sun-exposed microclimates over five surface types (asphalt, concrete, grass, rock, sand). Surface temperatures were observed using an Apogee SI-111 infrared radiometer attached to the cart and a handheld Delta-TRAK Infrared Thermometer. A comparison of surface temperatures time-detrended over all observations, by surface type, and by sun exposure highlights the importance of surface types and sun-exposure to achieve stationarity at the microscale.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner