I am looking at how rainfall can influence volcanic activity.
I have rainfall data from three different datasets. One is a gridded dataset at 0.25 x 0.25 degree resolution that offers monthly precipitation in mm/month. It spans the time period from January 1982 - December 2016. The second is a gridded dataset at 1 x 1 degrees resolution that offers daily precipitation in mm/day. Data is available from 1891 to 2016. The last is a land station/gauge based dataset that offers daily precipitation in mm/day and the distance to the volcano varies depending on the country. The time range also varies depending on the station, but can cover up to 100 years of rainfall.
I also have eruption times for multiple different volcanoes from 1900 to now.
What is the best way to analyse these datasets to see the effect rainfall can have on volcanic eruptions? I am new to statistics and research so any advice would be helpful!
I have two approaches based on past papers.
1. Make a timeseries of the rainfall data, overlay the eruption times over this and put in reference lines for the mean and two standard deviations above and below the mean. This is to see if there's any significant peak in rainfall before an eruption.
2. Use fourier transform analysis or spectral analysis to identify the wet and dry seasons in the rainfall data. Then carry out binomial probability analysis, determining what the distribution of eruptiutions in the wet season would be if they were randomly distributed, and then find the actual distribution.
Any thoughts on these methods or more methods would be appreciated!
The Pettitt test can be used to detect the abrupt change point in time frame, while the Mann-Kendall (MK) test and Sen’s Innovative trend analysis can be performed to analyze the rainfall trend.
Personally, I would probably try your method 1 as more visual representation. That seems to help me. If there seems to be some relationship, then continue to analyze in more statistical detail. The difficulty arises when a similar series of events might be linked to an eruption, and also might not be linked at some other time. Also, you might want to look at rainfall totals for various time periods before an eruption, days, weeks, months, or even annual. And are there any underground sources of water contributing, or responses to tidal, sea level, planet alignment, temperature or seasonal.
Thank you Muawya Ahmed Hussein, I had not heard of the Pettitt test so hopefully that might produce something interesting.
- Badges
- Science topic