| |
Title: |
 |
Development of a Statistical Model for Estimating Spatial and Temporal Ambient Ozone Patterns in the Sierra Nevada, California |
| |
Authors: |
|
Preisler, Haiganoush K.; Arbaugh, Michael J.; Bytnerowicz, Andrzej ; Schilling, Susan L. |
|
| |
Journal: |
|
TheScientificWorldJOURNAL |
|
| |
Year: |
|
2002 |
|
| |
Volume: |
|
2 |
|
| |
Page Range: |
|
141-154 |
|
| |
Article Type: |
|
Research Article |
|
| |
Domains: |
|
Atmospheric Systems
|
|
| |
DOI: |
|
10.1100/tsw.2002.86 |
|
 ?
|
Article clouds not enabled. Press the image to enable for this session.
|
| |
Synopsis: |
|
A spatial temporal model of ozone distribution over the Sierra Nevada using robust modern regression techniques was developed using data from a network of passive ozone samplers. The study indicated that characterizing the distribution of ozone exposure using passive monitors and a statistical model with auxiliary variables might be a practical alternative to dense networks of active monitors when seasonal ozone patterns are of interest. Also, it may be helpful in assessing the outputs of system models based on atmospheric chemistry and transport theory. |
|
| |
Keywords: |
|
generalized additive models, jackknife estimates, passive samplers, random effects |
|
| |
|
|
|
|
| |
|
|
[Related TSW Articles]
[Export to EndNote]
[Open Choice]
|
|
| |
|
|
|
|
| |
|
|
Abstract |
|
| |
|
|
Statistical approaches for modeling spatially and temporally explicit data are discussed for 79 passive sampler sites and 9 active monitors distributed across the Sierra Nevada, California. A generalized additive regression model was used to estimate spatial patterns and relationships between predicted ozone exposure and explanatory variables, and to predict exposure at nonmonitored sites. The fitted model was also used to estimate probability maps for season average ozone levels exceeding critical (or subcritical) levels in the Sierra Nevada region. The explanatory variables — elevation, maximum daily temperature, and precipitation and ozone level at closest active monitor — were significant in the model. There was also a significant mostly east-west spatial trend. The between-site variability had the same magnitude as the error variability. This seems to indicate that there still exist important site features not captured by the variables used in the analysis and that may improve the accuracy of the predictive model in future studies. The fitted model using robust techniques had an overall R2 value of 0.58. The mean standard deviation for a predicted value was 6.68 ppb. |
|
| |
|
|
Related articles in:
|
|
| |
|
|
|
|
| |
|
|
| |
Post Your Comment |
 |
|
|
| |
|
|
|
|
| |
|
|
|
|
| |
|
|
|
|