Inaccurate Evapotranspiration

[kaisersoze]

Hello,

 

I live in San Diego and I find that the ISY calculated evapotranspiration is inaccurate for me, usually either too low on hot days and too high on overcast days.

 

I came to this conclusion based on CIMIS data (http://wwwcimis.water.ca.gov/cimis/welcome.jsp) and an independent Davis weather station in San Diego (http://www.stationpabloco.com/). These two sources had nearly identical daily evapotranspiration numbers over the last week when we had 2 extremes of weather: 4/14-4/15 almost complete cloud cover with some occasional drizzles and 4/16-4/19 with no clouds and highs in the 80's. For the days of complete cloud cover, the CIMIS reported ETo 0.04 and 0.02 and the sunny days were 0.13, 0.21, 0.24, 0.24 (stationpabloco had nearly identical values).

 

The ISY calculated evapotranspiration values were 0.09 and 0.07 for the drizzly days and 0.13-0.15 for the sunny warm days: too high for sunless rainy days and too low for sunny hot days. My weatherbug station was pointed at the weatherbug backyard station data posted by stationpabloco (5 minute polling interval), so data should be identical. I tried a different weatherbug station before and found it was inaccurate, thats why I switched to stationpabloco last week because it calculates ETo on its website and has weatherbug backyard station data available to my ISY which should theoretically be identical.

 

I reviewed the ISY evapotranspiration algorithm. Does the algorithm use a static historical monthly average solar radiation in its calculation? Would that explain why the calculated ETo for sunless rainy days is too high (actual solar radiation much lower due to cloud cover than the historical monthly average), and the ETo for unusually sunny/hot days is too low (actual solar radiation on sunny hot day much higher than the historical monthly average)?

 

Any suggestions for me to try to make the calculation more accurate?

[Michel Kohanim]

Hello kaisersoze,

 

Thanks so very much for the feedback.

 

Last week WB had problem with backyard stations and that caused major problems in ET calculations since ISY was not getting correct data points. Based on this event, IM suggested that we allow selection of algorithms either Penman-Montieth or Hargreaves-Saman. PM method is much more error prone in case of WB data not being there. HS is much more resilient.

 

Ra is calculated on a daily basis and applied to ETo for that day.

 

With kind regards,

Michel

[IndyMike]

Hi kaisersoze,

 

Thank you for providing the address for the WB backyard station. This particular station has feeds for both Weatherbug and WeatherUnderground. WeatherUnderground is useful since it provides historical data.

 

The "classic" Penman-Monteith calculation for ETo requires measured Solar Radiation (Rs). Weatherbug does not support Rs data in their API. As a result, the ISY is using an approximation based on elevation, temperature, and location (coastal or inland). The approximation works very well on average. As you've noted, it has a tendency to over estimate on very cloudy days. Fortunately, these are low ETo days and the errors should be on the order of hundreths of an inch.

 

It should perform very well on sunny days - in contrast to your observations. This is my reason for posting. I would have expected your sunny day calculations to perform better than you have indicated.

 

Using the WeatherUnderground interface for your local station, I pulled the historical data for the month of April. The table below shows what I believe the ISY should have calculated on a daily basis. I agree with your cloudy day numbers (4/14 - 4/15) of 0.07 to 0.09 ETo. I do not agree with your sunny day numbers (4/18 - 4/20) of 0.13 to 0.15. Something is wrong here, a Weatherbug sampling error.

 

 

To assess the calculation accuracy, I pulled CIMIS data for your area. CIMIS provides all the necessary data inputs for the calculation, and maintains historical calculated ETo. The Escondido appeared to be the best match for your location.

 

The following shows the CIMIS data and ISY predicted calculations for the current week. It also shows the ISY calculation error for the week based on the CIMIS "PM ETo" calculation. Remember that CIMIS is using measured Solar Radiation, whereas the ISY is using an approximation.

 

Given Identical input data, it would appear that you should be using a setting of "interior" in your climate settings with a resulting error of 0.058" ETo (4.7% error). Surprisingly, the Hargreaves-Samani equation is performing better still (implementation is in process). I would not expect this to be the case over a longer period of time.

 

 

 

So, why are you seeing different values? Possibilities:

1) The ISY P-M calculation relies on averages for Windspeed, Dewpoint, and Temperature. These are averaged over the course of the day. While High/Low temperatures are used, these can be acquired in a single "snapshot" at the end of the day. Communication dropouts will cause errors in the data averages.

2) Weatherbug location changes - If a station is unavailable, weatherbug will automatically switch to a "close" alternative. In some areas, where the terrain changes drastically in short distances, this can be a huge source of error. Disclaimer - I do not know if weatherbug does this for "backyard" locations.

3) Poorly placed/maintained station. Stations used for general weather conditions are often not suitable for ETo calculations. The ETo calculations require proper placement and maintenance of the equipment. Note: this is a general comment - your selected PABLOCO station appears to be reporting data that is consistent with CIMIS. If you have correlated the readings between CIMIS and PABLOCO then this is an excellent choice. My hat's off the the backyard operator.

 

Communication errors are the reason that the UDI team is investigating the Hargreaves-Samani ETh calculation. Since it relies on only the max/min temperatures of the day, the data can be acquired in a single snapshot. Averaging is not required.

 

On the subject of communication errors - when I visited the Wetherbug backyard and WeatherUnderground sites for your local station, I found some interesting differences:

 

1) Weatherbug was reporting a high temperature of 81.2 degrees recorded at midnight (00:00)

2) WeatherUnderground was reporting a high temperature of 59.4 degrees.

 

This could be a simple website translation error - I don't know what the API is transmitting to the ISY. If the 81.2 degree high is being communicated, there will obviously be some significant errors.

 

Weatherbug has been having issues with the feed from the backyard stations for some time now. Is there another station that you could use?

 

Edit: I was able to connect to your local station using the ISY. It did report a high temperature of 81.2 degrees for the day. Based on the stations reports for the day, and the weather graph presented, 81.2 degrees simply isn't possible. Apparently Weatherbug is still having issues.

 

The Station master for this site provides data to the NWS/NOAA in support of the MESONET. In other words, he/she takes this stuff seriously and probably doesn't realize that the data is being misrepresented on weatherbug. You may want to notify him/her.

 

[JMC]

After reading through much of the traffic on ETo calculation differences between the PM and HS methods I decided to try the HS to see what, if any, difference I might see. For reasons I don't understand, ever since switching to the HS calculations ETo and Yesterday's Water Deficit have been 0 each day. Considering that temps have been right around triple digits with low humidity and 0 rain this doesn't make sense. I switched back to PM and will see what effect that has.

 

mike

[IndyMike]

After reading through much of the traffic on ETo calculation differences between the PM and HS methods I decided to try the HS to see what, if any, difference I might see. For reasons I don't understand, ever since switching to the HS calculations ETo and Yesterday's Water Deficit have been 0 each day. Considering that temps have been right around triple digits with low humidity and 0 rain this doesn't make sense. I switched back to PM and will see what effect that has.

 

mike

 

Hello Mike,

 

The table below is based on readings from the KBAB weather site at Beale AFB. It shows what the ISY should be calculating for the PM and HS ETo.

 

The HS method is extremely simple. It relies on Min/Max temperature alone. The only way that I can see that you could be getting an ETo of zero would be if WeatherBug was misreporting the Min or Max temperature. if Max - Min goes to zero, your ETo also goes to zero.

 

I simulated a "mis-reported" low temperature in the second table. This produces an varying error in the PM calculation. The HS calculation goes to zero.

 

Could you check the ISY irrigation display and see if high and low temps are being displayed properly?

 

IM

 

[Z3phyr]

After working well for over a year, my climate module started reporting evapotranspiration of less than 0.01 inch/day for the last couple of months. I have returned to time based irrigation so the lawn doesn't die. I've changed between the two calc methods with no discernible improvement.

 

I just switched the WeatherBug station, will see if that helps.

[JMC]

IM: Today's high/low are being displayed correctly, low was 73 and high (so far) is 97. I don't know how to get an historical period from KAUN so I'm only giving today's numbers. PM is also giving me a very reasonable ETo calculation (0.2149"/day) for today.

 

One possible answer I neglected to mention is a couple of very short duration (less than 1 sec) power outages did occur while the calculation was set to HS. Both were too short to cause the ISY to reboot but they did produce enough anomaly that I manually rebooted. It was then I noticed the ETo was 0 so I watched for change the next morning. After two days with ETo still at 0 I posted my comment.

 

mike