I plotted the residual chlorophyll concentrations at the times of the primary and secondary chlorophyll bloom peaks. This was to double check if there were any significant trends in seasonal peak chlorophyll concentrations. Using the the residual chlorophyll concentrations did not show any significant overall trends though.
I also went back and changed the dates that I used when comparing the day of year that a bloom started. I changed it from the dates of the highest concentration in that seasonal bloom to the date that the bloom first started to increase. My thinking was that the date of the peak wasn't as important since the peak tended to occur at several different stages of the bloom. I felt that the triggering of the start of the bloom was instead what we were looking to for signs of a trend. However this again did not show any significant trend leading to a possible conclusion that the start times of the blooms are not getting significantly earlier or later year to year.
Yesterday I started working with the sea surface temperature data. This required gaining an understanding of netcdf files and how to use them. Once I had the right MATLAB files downloaded I started going through Carrie's preexisting code to try to clean it up a little and make it more open to possible changes to the dates we choose to analyze. Also I downloaded the land sea mask and took a while to write a code that could use it.
Today I continued cleaning up the code a little bit and looked into getting the right dates to show up. NOAA uses the same serial date format as MATLAB only it starts in 1800 instead of starting at 0 A.D. The simplest way I found of fixing this was to just add 657,436, the equivalent of 1800 years, to all the time data points when first loading them in. I double checked this and found it left the day and month unchanged but now all the years are correct displaying 1981 instead of 181.
Subscribe to:
Post Comments (Atom)
Dealing with different times is a common problem with Matlab, or indeed, any data analysis project.
ReplyDeleteAnother one you often run into are unix timestamps, which are seconds from 1970. People often get confused, but the easy fix is:
matlab_time = unix_time/(24*60*60) + datenum(1970,1,1);
It's also a good idea to leave the explicit function call to datenum in your code, as the next person to read it might get confused as to why you added 657,436.
So, no luck with the bloom start date? Have you had a chance to look at the residuals from the long term (multi-year) mean, or only the monthly mean? I also wonder if there's a relationship between the length of the blooms, or rather the bloom concentrations integrated over the length of the bloom, from year to year?
At any rate, keep fighting (with Matlab ;) and Good Luck!