More on Microseisms

The previous post gave me some confidence to suspect that storms in the north Atlantic basin may give rise to much of the microseism activity I see on my seismograph. Earlier this winter I ran across a public US Navy web page that shows the results of computer modeling of significant wave height in the north Atlantic and have been visually comparing those plots to my microseism activity. At least on some qualitative level this new information continues to support the notion that the microseisms I see result from storms at sea. Take a look at the two examples below.

On Jan 14, 2017, the seismograph was quiet so I looked at the significant wave height model to find that it was also very quiet:



Today, Feb 14, 2017 the seismograph is very noisy:


Take a look at the Significant Wave Height now.  The wave heights at the center are in the upper 30 to mid 40 feet values!


I had actually predicted that this would happen on Feb 11, 2017 by looking ahead at the Significant Wave forecast given by the model.  If I can find access to the gridded data or come up with a way to at least roughly re-digitize these charts I will try to see if there is a simple relationship and correlation between the wave height, the distance from the seismograph to the center, etc. and the amplitude of the microseisms.

This suddenly makes the microseisms, which used to be more of an annoyance, now almost as interesting to watch as the earthquakes.

Whence Microseisms?



Normal Seismogram


Noisy Seismogram showing Microseisms








In the twelve years or so that I have been observing earth motions on my home built seismograph I have been puzzled by the occasional appearance of noisiness in the seismometer trace. I have read about these microseisms and tried unsuccessfully to identify a source of them. The literature always identifies storms at sea as the main cause but I discounted that explanation because of my location in Ohio and the fact that I don’t have a professional, broadband seismometer. So I looked at local wind speeds and wind gusts buffeting the house or moving the trees so much that their roots move. I have looked at local barometric pressure and temperature changes even monitoring the temperature inside the enclosure which contains my seismometer. I have wondered whether there is something in my amplifier electronics that might explain these tiny signals. Nothing worked out.

Earlier this month, I noticed that they were back and that they were growing quite strong. Then it occurred to me that Hurricane Matthew was approaching the east coast of Florida at the same time. Maybe the microseisms I see can be attributed to storms at sea!

I began a closer look at this by downloading time series data about the hurricane from the National Hurricane Center ftp site.  I wrote a python script to load the data from the downloaded file and plot the track of Hurricane Matthew and other information I thought might be important such as the maximum sustained wind, central pressure, and the distance of the eye from my home in Millersburg, Ohio which I calculated from the location data.


Note that P0 marks the location of the seismograph in Millersburg, Ohio.

I also downloaded a bundle of specialized microseism analysis tools from the IRIS (Incorporated Research Institutes for Seismology) software site.  This software is designed for advanced study using professional grade, broadband, multi-channel seismographs.  As such, I have most likely not applied it correctly, in fact bypassed important calculations, etc.  Having said that I hope that the “relative” analysis I did shows some direction toward a qualitative relationships between storms at sea and the microseisms I see recorded on my seismograph.  Here is what I found.


This data includes Local, Secondary, and Primary Microseism energy.  I chose to use the Primary band data since they all show similar qualitative response.  Hurricane Matthew dissipated  on Oct 10 and the NHC data file was closed and no more data was added.  More on that later.  I snipped the microseism graph off at Oct 10 and compared to the Hurricane central pressure and distance to the eye:


Actual (non-inverted) data

This is a little hard to visualize in part because one would think that the microseism energy might be inversely related to both the distance and the central pressure.  We can look at the same information taking this into account by plotting the reciprocal of both quantities:


Inverted data (see text)

The last chart is the closest to correlated data that I have seen in twelve years or so of trying to figure this out.  Not perfect due to many known and unknown reasons but something for me now to be aware of when I see this noisiness show up.

The proximity of the storm seems predominant.  The central pressure seems to lead the microseism energy.  And finally, the original microseism graphs that went out to Oct 13 showed the noise staying high long after the NHC downgraded the storm.  Presumably, the ocean does not calm immediately after the storm dissipates.  An unmentioned complication in this study is that Hurricane Nicole was also present and strengthening during this time period further out in the Atlantic.

New Live Seismograph Display


A year or so ago, I upgraded my seismograph software from the original IRIS AmaSeis program to an upgraded version based on Java called jAmaSeis.  I ran the new software on a little Asus netbook.  The helicorder display that I uploaded to the Live Seismograph page on this site was just a screen shot of the active window every five minutes.  The little netbook was at its performance limit and the screenshot made it impossible to do any analysis anyway.

This fall I built up a Linux desktop computer running XUbuntu 14.04LTS and installed jAmaSeis on it.  I downloaded from the IRIS website and followed the basic instructions for installation given on the same site.  With a little fiddling around, I got the software running.  I plugged in the Dataq DI-145 USB Analog to Digital converter…the software could not find it.  I fiddled around with it for a long time with no success.  I could see the device appear in /dev as ttyACM0 when I plugged it in.  After giving up on it a couple of times I finally created a symbolic link from ttyACM0 to ttyS32 where the seismograph software would detect it.  I do this manually every time the computer reboots or the ADC is unplugged.  I plan to write a udev rule to do this automatically.

To get the helicorder display for the Live Seismograph page I use an open source package of seismology tools for Python called Obspy.  jAmaSeis writes the data stream in one hour long segments in sac formatted files arranged in a time based directory structure.  The Obspy stream manipulation tools allow one to easily build a continuous 24 hour data stream from the jAmaSeis files and plot it in a moderately flexible way in a Python script.  This method yields several improvements over the screen shot.  The date and time are unambiguously shown on the vertical axis and the traces alternate through four colors to differentiate the hour in which an event occurs more clearly.  The Python script runs as a cron job every five minutes.

The Obspy package also provides tools to parse QuakeML documents which I obtain from a USGS feed inside the same script that plots the data.  After parsing each event, I use the obspy.core.util.locations2degrees tool to find the distance from the epicenter to my station.  The script then annotates the helicorder display with seismic events selected using magnitudes and distances that might be detected with my seismograph.  This selection is arbitrary so there will be some that show on the trace without annotation and others will be annotated when there is nothing showing on the trace.

M4.0 – 8km S of Galesburg, Michigan


A Magnitude 4.0 earthquake occurred in southern Michigan at 12:23 EDT this afternoon generating a very clean signal on my seismograph at Millersburg in northeastern Ohio.  The analysis I did using jAmaseis estimated the distance to the epicenter to be 353 km.  The Google Earth ruler measured 348.5 km.  The signal overlaid on the jAmaseis travel time curves is shown below.


Software Defined Radio — Installing the NooElec NESDR Mini 2 for Linux


I will take the opportunity now to stop and review where I am with the Satellite Tracking project that my older son and I started several years ago. In 2009 we recorded several passes of a now decommissioned amateur radio satellite named HAMSAT (VO-52). With attention to the timing accuracy of the recording, I was able to estimate slant range from the receiver to the satellite as a function of time using a simple Doppler Shift model. I have then recently developed a Python script to estimate the actual location (eg. latitude/longitude of the ground track, altitude, and azimuth/elevation) of the satellite given the range/time data from three ground stations of known latitude and longitude.

The project now has some momentum in the direction of a real test sometime in the future raising the question of what radio receivers I should consider. My son has suggested the use of Software Defined Radio, a concept I like very much because it is in line with one of the unstated goals of most of my projects to do as much as you can with as little financial cost as possible. A month or so ago he mentioned Gnu Radio and a hardware device known generically as a DVB-T Dongle based on a Realtek RTL2832U circuit. The interface between GNU Radio and the dongle uses the RTL_SDR codebase as discussed at and I invested $25 to see if I could make it work. It is not a plug and play device and it does take a little work to get it operating so I am documenting what I had to do as a blog post here.

Acquiring and installing the NooElec NESDR Mini 2

While there are a lot of DVB-T dongles available, I purchased the NooElect NESDR Mini 2  on Amazon plus a PL-259 pigtail for it. Dongles from some suppliers take a long time to ship and defective devices are reportedly common.



I am programming this project in Python and decided that the best platform for the project was Linux so I installed the radio on a Dell Vostro notebook with 2 GB RAM and a 2 GHz Intel core duo processor running a brand new installation of XUbuntu 14.04 LTS. After reading several websites it looked like most pointed to osmocomSDR. There is a lot of information there. I have extracted only what I did to get up and running. My inexperience with Linux will become evident.

I downloaded the software as a package release as described on the website and extracted the archive. I tried to build the software following the instructions at osmocomSDR but the build threw several dependency  errors. Save yourself some time and install these before you start.

Here is the build sequence using cmake:

cd rtl-sdr/
mkdir build
cd build
cmake ../
sudo make install
sudo ldconfig

cmake was not installed so I did a

sudo apt-get install cmake

and tried again. This time the error was that the libusb1.0 library was not installed. The osmocomSDR website very explicitly says this is required. In my previous experience with Linux I have used the synaptic package manager to download things like this but it is no longer installed by default. Most of this stuff could have been install from that…I have installed it now, after the fact. I downloaded the build package for libusb1.0 from SourceForge and extracted it. It threw an error in the configure step that libudev was not installed. I installed it from the Ubuntu Software center.

After that the build as described on osmocomSDR went smoothly to completion.

The installed codebase has several command line programs so I ran the first one described on the website

rtl_fm -f 96.3e6 -M wbfm -s 200000 -r 48000 – | aplay -r 48k -f S16_LE

I was greeted by the gentle hiss of a radio that was tuned to dead air. I exited and edited the command to

rtl_fm -f 95.3e6 -M wbfm -s 200000 -r 48000 – | aplay -r 48k -f S16_LE

our local FM station. This time I was greeted by the local high school girls basketball game at the some kind of state championship or something. I must admit that I was never so happy to hear a basketball game as I was this one.

Then I installed a SDR called GQRX from the Ubuntu Software Center. After fooling with the controls (mainly the receiver gain which is set to automatic by default) I got it working too…the blue LED in the lower right is the dongle.  Shown here tuned to a nearby repeater for our regional NPR station.



Finally, I downloaded Gnu Radio from the Ubuntu Software Center and prepare to build the rtl_sdr source block for it.  I ran the Gnu Radio Companion first though and was surprised to find the Osmocom rtl_sdr block already there.  I had installed Gnu Radio on another identical install of XUbuntu and there were no sources shown.

Here is a screen shot of Gnu Radio Companion ready to build a radio based on the rtl_sdr dongle.  I will describe that too once I figure out how to do it.




A New Page — Passive Tracking of Satellites using only Range Data


I have added a new page to the Earth orbiting satellites section of Murmurs from the Earth…Whispers from the Sky.  Passive Tracking of Satellites using only Range Data continues the project started several years ago to study the Doppler shift of satellite radio beacons.  The page describes a method to determine the ground track and other information about the position of a satellite as it passes over three ground stations recording the satellite radio beacon.

While I am certain that this has been done in the past and replaced with other techniques, I developed the method and Python scripts from scratch, to the extent that is possible…obviously I scavenged code and ideas from wherever I could to make a lot of the cranks that needed to turn.  I used a  very simple geometric approach.

Lacking actual data I generated four artificial and hence, internally consistent data sets using the PREDICT pass prediction software that provide the range input data and the location check data.  I was quite pleased with the initial results.

M7.3 – 67km WSW of Jiquilillo, Nicaragua


A Major earthquake, Magnitude 7.3, occurred of the coast of Nicaragua at 03:51:35 2014 October 14 UTC.  The image above shows the trace from my seismograph in the Event mode of the latest version of jAmaSeis.  The distance is computed by dragging the trace around to get the best fit to the travel time curves.  I usually cheat a little by selecting the portion of the trace starting at the actual time of the event.