Over the last 2 days (11/13-14) we had quite a bit of wind in Tucson. These are probably the same winds that plagued California and spread wildfires over the weekend. The winds drove the humidity way down and the temperature followed. Note the large pressure spike (it went just off the scale on Monday). This gradient must have drug the winds along with it, I suppose.
I’m curious about the luminosity spike (both visual and IR) yesterday, because it wasn’t noticeably brighter than the previous few days. Will be watching to see if this is a trend. One hypothesis is that the wind knocked down some leaves from the mulberry tree that protects my porch (and therefore, the weather data collection rig).
Here’s an interesting view at the last month. Traditionally, Tucsonans know that Tucson weather changes quickly in the fall from summer to winter. This change is always welcomed, but it also means that Thanksgiving always sneaks up on us.
The plot below shows the drama of this changeover in the last month from higher temperatures to lower. Plus, you can see the timeperiod in the middle where the change seems to have been put into motion by a storm event (see the lightning spikes in blue). After those 6 days of cloudy and stormy weather, we have slipped into a pretty predictable pattern with no clouds and gradually decreasing temperatures. Looks like something must be happening today with the high pressure and lower temperatures, though.
We’ve been getting a lot of rain lately in Tucson. Not sure if it’s a late Monsoon or early winter rains (or hurricane rain). Two days ago we had rain nearly all day. You can see all the lightning strikes in the plot below. As normal, at the scale I use, the intersection of the humidity and temperature lines signify rain and the “pointy” luminescence fills signify cloud coverage.
October is the month that Tucsonans live for, because it is when we finally see hope that our wonderful winter weather is coming. In the one month chart here you can see the hot, consistent pattern of September get interrupted (in this case by the remnants of a hurricane). This must have been the trigger, because the patterns got all confused and the result is our nice, low morning temperatures. Humidity is still swinging wildly, but the highs are lower every day (along with the temperature. The luminescence sensors are also telling a story of cloud cover (see the jagged peaks and compare to the smooth hills of September).
You can look at the timeframe from 10/1 to 10/3 to see how the remnants of a late season hurricane passing through the desert can disrupt a stable non-monsoon weather pattern.
From about 9/27 to 10/1, we see the typical non-monsoon cycle recover after the 9/19 monsoon event. The hurricane event on 10/1 throws this pattern into disarray and it still hasn’t recovered. Note the difference in the luminosity data, temperature, and pressure.
Looking at the last month of data, we could be seeing a trend towards drying up and cooling down. The length of the days is getting shorter and the light is less intense (see the Light spectrum ‘bumps’). Also, the temperature swing is smaller through the day. The pressure went up quickly after the last rain event and now is tapering off. We’d expect higher temperatures for the last couple of days but didn’t see that. The daily lows seems to be a bit lower each day. Barring unforeseen weather events, I think we’ll start seeing the range between the highs and lows get smaller each day.
A few days with hotter temperatures. Clouds were overhead for much of the weekend, but you can see one cloudless period as represented by the spike of luminosity yesterday afternoon.
We saw lots of interesting weather last week. Had lightning and rain a few days in a row. The pattern continues to emerge that (at this scale) when humidity intersects temperature it signifies rain.
Also, I now have enough light intensity data (both visible and infrared) to plot now. I don’t see any real trends with this new data other than luminosity is a proxy for cloud coverage (temperature is another one, but perhaps situations will emerge where light measurement may forecast rain). Eventually we’ll see the trend where shortening days impact temperature (please, soon!). As always, you get better chart detail by looking at the .pdf chart. Stay tuned as I continue to explain how the data collection system works and what the long term goals for the data is.
MONSOON DATA COLLECTION SYSTEM
I’ve been sharing plots that I automatically generate daily from the sensed data from my Monsoon Data Collection system that sits on my back porch. I figured that there might be interest in what this system consists of and the vision for what this data could provide.
System Bill of Materials:
- Embedded Adventures MOD-1016 Lightning Sensor ((Link)
- Embedded Adventures MOD-1022 Environmental Sensor ((Link)
- Adafruit TSL2591 High Dynamic Range Digital Light Sensor ((Link)
- Raspberry Pi 2 Model B (Link)
- Raspberry Pi 1/2 size breadboard (Link)
- Raspberry Pi Digital I/O Extender (Link)
- Wires, etc., to connect the Raspberry Pi to the I2C Sensors
The Embedded Adventures MOD-1016 is a breakout CCA based on the AS3935 lighting/storm sensor (Link). This part is interesting because it attempts to handle the more challenging aspects of calibration at the factory . This saves the experimenter from writing a lot of calibration code. The MOD-1016 can interface with I2C or SPI, but I typically choose I2C because I’m pretty familiar with the protocol on the Raspberry Pi.
This part also sends interrupts to the Raspberry Pi when something interesting happens. If it senses excess noise in the environment (for example, my RPi case fan that’s acting up), it will send the interrupt so that the RPi can raise the noise floor. It will mask what it determines to be human made disturbers that aren’t real lightning (i.e., the microwave oven!). And finally, it will send an interrupt to inform the RPi when it senses lightning. When requested by the RPi, the sensor will then deliver a unitless intensity value and an estimated distance.
The AS3935 Franklin Lightning Sensor(tm) is made by AMS and advertises detection of cloud to ground and cloud to cloud lightning within a 40km range. The part seems to do what it advertises. The only troubles I have had with it involve interrupts getting “stuck” that force my interrupt service routine to go nuts to service them. This happened once and I powered the system down for a minute, which fixed it.
- Description of other sensors
- Description of the problem being addressed
- Description of the intent for the data
- Discussion of the SW architecture of the system
- Link to GitHub repository
A bit of hot and dry for the last few days. Had a hardware problem with the system that missed some lightning during the rain days. Fixed the issue, now it’s primed to catch the next storms.