To apply the techniques of feature-based visualization to climate data sets, we need to think about what to consider as features of special interest. The basic parameters available such as temperature, refractivity, and geopotential height have rather well-known dependencies between each other, so additional insight is anticipated mainly from investigating newly derived parameters.
Two derived parameters are considered: The linear trend calculated as a moving 10-years-difference and the signal-to-noise ratio (SNR) defined as the ratio of the trend to the detrended standard deviation. To detect and explore regions sensitive to climate change in time and space, the features of interest are composed of high values for the linear trend while maintaining a high SNR. To obtain these features the following parameters are generated:
• Smoothed Data yav: To generate the linear trend the data y is first smoothed using a moving arithmetic average with an averaging timeframe of 11 years.
• Linear Trend b: The linear trend per year bi (where i denotes the center year of the current timeframe) is calculated as a moving 10-years difference between the data of year i + 5 and year i - 5 (Eq. 1). Due to the exponential character of the refractivity with height, the relative trend (in %) is generated in this case in relation to the first value of the current timeframe (Eq. 2).
Was this article helpful?