Prioritize...
After reading this section, you should be able to define a time series and describe the importance of having long datasets.
Read...
As with Meteo 815, we need to be cautious before we begin an analysis. Before we dive into all the applications of a time series, we will first talk about the time series data itself. In this lesson, you will learn a very simplistic approach meant to illustrate how the components of a time series all add together. We will hit it with more powerful tools in later lessons. For now, let’s begin by defining a time series.
What is a time series?
A time series is a sequence of measurements spanning a period of time. Usually, these measurements are taken at equal intervals. Although we inadvertently used time series data in Meteo 815, we generally discussed the random samples of observations, not the sequence. In fact, the assumption for many of the statistics we computed in Meteo 815 included the need for randomly sampled data. For a time series, the key assumption will instead be that each successive value of the dataset represents consecutive measurements of the variable, usually at equal time intervals.
Why do we care about a time series?
There are two main goals, in this course, for examining a time series. The first is to identify the nature of a weather or climate phenomenon. We might observe something occurring and ask why that event occurred. We could use a time series to investigate that particular event in more detail.
Check out the video below for an example:
[THUNDEROUS SOUNDS]
PRESENTER: Here's the latest from Earth Now.
[BIRDS CAWING]
Extreme weather in the Northern Hemisphere is increasingly blamed on arctic amplification. What is arctic amplification, and how does it affect our weather? The display first shows a satellite image of Arctic sea ice from September 1980.
[WAVES ROLLING]
Ice is colored white. The temperature differences between the cold poles and the warm tropics, combined with the Earth's rotation, cause air to flow eastward fastest over the mid-latitudes, where most of us live. This is called the jet stream and is shown here in red and blue.
Approaching North America, the jet stream moves northward over the Rocky Mountains, then dips southward, forming a trough towards the east coast, then northward as a ridge over the Atlantic Ocean. It continues in a similar long wave pattern over Europe, Asia, and the Pacific. Storms develop and track along the jet stream, pushing cold air south, shown as the blue portion of the jet stream, and warm air north, shown as the red portions of the jet stream.
Now, the display shows the recent arctic sea ice from September 2012, the lowest ice extent ever recorded by satellites. The current typical jet stream location is now shown, with the previous ice extent and jet stream location typical of 30 years ago shown in pink for comparison.
[BIRDS CAWING]
During the past few decades, general warming in the atmosphere has accelerated arctic ice melt, leading to more seasonal ice cover, which is not as white and therefore absorbs more sunlight, in turn causing more warming. The Arctic has warmed twice as fast as the rest of the Northern Hemisphere. This is arctic amplification.
Scientists have observed that the reduced temperature difference between the North Pole and the tropics is associated with slower west to east jet stream movement and a greater north south dip in its path. This pattern causes storms to stall and intensify rather than move away, as they normally used to do. At mid-latitudes, more extreme weather results from this new pattern, including droughts, floods, cold spells, and heat waves.
[WAVES ROLLING]
And that is how arctic amplification is affecting our weather.
[WAVES ROLLING]
It shows how the Arctic affects extreme weather. You can find a description here of the video, but let me briefly describe the premise. The number of extreme weather events has increased over the past several decades. This could be due to a number of reasons, including better monitoring. But one hypothesis is that Arctic Amplification (rapid Arctic warming - twice as fast as the rest of the Northern Hemisphere) may be a cause. So by examining the time series of extreme weather events and Arctic ice melt, a scientist could understand the interaction in more detail.
The second reason we analyze a time series is for forecasting. By identifying patterns in time, we can extrapolate to predict occurrences in the future. We can also use a time series to identify the dominant time scales across which variability occurs. This tells us a lot about which extrapolation forecasts will be useful. Meteo 825 will focus on forecasting so it’s key that you understand this course to utilize the information for the follow on.
Check out the video below:
[ON SCREEN TEXT]: Ocean temperatures influence the atmosphere above, impacting weather and climate. Temperatures at the ocean's surface are constantly changing. Over time, seasonal patterns emerge. A band of warm weather shifts north and south as the seasons change.
The El Niño/Southern Oscillation (ENSO) cycle is a climate pattern that occurs approximately every five years. ENSO is marked by changing temperatures across the eastern tropical Pacific Ocean. ENSO has two phases: warmer than average and cooler than average When surface temperatures along the eastern tropical Pacific become warmer than average the phase is called El Niño. El Niño drives extreme weather worldwide. El Niño's opposite phase is called La Niña. In 2010, El Niño's warm waters cooled quickly, giving way to a strong La Niña. The 2010-2011 La Niña was behind many of the year's weather disasters. Scientists use sea surface temperature observations to forecast weather and understand the global climate system.
ENSO is an oscillation that will be discussed in more detail later on. The video is showing how we can use the ENSO cycle to predict weather events (like droughts, floods, cold temperatures, etc.) for the season. By analyzing the time series of ENSO and these other weather events, we can use the relationship to forecast future weather events.
In short, performing a time series analysis can be quite valuable both for investigating connections and utilizing the results for forecasting.
What do we need to consider when examining a time series?
Generally speaking, when we perform a time series analysis the main goal is to detect a reproducible pattern. This means that, in most cases, we will assume that the data has a pattern with some random noise and/or error. The random noise is what makes it difficult to pull out patterns.
Throughout the lesson, we will examine some potential issues that may arise with a time series. Whenever you begin a time series analysis, it would be beneficial to consider the points listed below (which will be discussed in more detail later on) as they may pose a problem in the actual analysis. We can correct these potential problems if we detect them before conducting the time series analysis:
- Are there any missing values?
- Are there any abrupt changes or shifts (could be natural or artificial)?
- Is there constant variance over the time period?
- Are there regularly repeating patterns (daily, monthly, etc.)?
- Is there a trend (steady increase or decrease over time)?