Rainfall Revisited: Accurate Observations and Beyond

As a provider of weather analysis and forecast services to the agricultural industry, one of the most common questions I get asked is, “How accurate are your precipitation estimates compared to a gauge?” Earlier this year, I wrote a bit about the spatial variability of rainfall and some of the reasons a radar-derived area estimate won’t usually exactly match a point gauge measurement and how gauges themselves may not be representative of actual rainfall over the entire field or even a nearby field. After an abnormally wet August in my home state of Oklahoma, I felt inspired to do some real-world comparisons of measurement variability, using my own gauges and selected nearby gauges.

As a meteorologist, weather enthusiast, accuracy-freak, and an official observer for the Community Collaborative Rain, Hail, and Snow Network (CoCoRaHS), I take great care to make sure my instruments are installed and maintained properly. I have two different gauges installed less than 10’ from each other at nearly the same height. My “official” gauge is a manual Stratus RG202 cylindrical rain collector with a 4” funnel, providing measurement accuracy of 0.01”. I check this gauge daily at 7:00 a.m. local time and have been providing observations to CoCoRaHS for this location since 2005.

My second gauge was installed in late 2017 and is an automated tipping bucket that is part of a Davis Vantage Pro2 Plus weather station. It also provides measurements to the nearest 0.01” and has the advantage of being automatically recorded. The Davis station is a popular choice for agricultural applications, and they come from the factory fairly well calibrated. Having these two gauges during this growing season has given me a great opportunity to do some real-world comparisons of collocated gauges.

I computed monthly rainfall totals for the period March through August, since the warm season eliminates issues of having to deal with snow measurements. In addition to my two gauges, I obtained data from two other nearby trusted sites. A map of the sites and table of results and site description are included below.

 Source

 Total Rain (Mar-Aug 2018)

 Description

 Home 4”  21.75  Author’s official CoCoRaHS 4” cylinder gauge
 Home Davis  20.24  Author’s collocated Davis VantagePro2 station
 OK-CV-84  22.19  CoCoRaHS site 0.5 mile SE of author’s gauges
 NRMN  22.99  Oklahoma Mesonet site 1.6 miles SW of author’s location

First, it is notable that the total rainfall between my two gauges differs by 1.51”, about a 7% difference. Although you might assume the Davis has a low bias compared to my 4” gauge, there were multiple events where it recorded more rain than my 4” gauge, and while not shown here, the May monthly total was higher for the Davis bucket than my cylinder. However, it is well-documented in scientific studies that tipping buckets can undercount compared to collection gauges in heavy rain events, and we had several of those in our typical Southern Plains thunderstorm activity over the season.

For the nearby sites, station OK-CV-84 is another CoCoRaHS site using the same model 4” rain gauge as mine. The NRMN station is an Oklahoma Mesonset site with an automated tipping bucket gauge. Across all four of the gauges, the estimates range from 20.24” to 22.99”, or from about 7% less to 5% more than my 4” gauge. The two tipping buckets differed from each other by 13% and the two 4” manual gauges by about 2%. However, the two manual gauges were closer to each other (0.5 mile apart) than the tipping buckets (1.6 miles). Nevertheless, the differences are due to a combination of measurement method, calibration, siting, and location, even within a relatively small area, and if you only had this one small study to go by, you might infer there could be a 5-10% uncertainty of any measurement as being “truth” on a field, even from a properly installed and maintained gauge. These results are consistent with a very rigorous, 30-month scientific study that compared measurements from six, high-end professional gauges (three tipping buckets and three weighing buckets) installed within the same 25 meter by 25 meter plot.

I took this one step further and compared these gauge results to two other commonly used sources of data for local precipitation: grid-based analyses and nearest climate station of record. For the grid-based values, I looked at historical data queried from my company’s ClearAg weather API service as well as from the National Weather Service’s Advanced Hydrological Prediction Service (AHPS) monthly analysis data. In both cases, I used values for the grid cell (2-km for ClearAg, 4-km for AHPS) containing my home location. Rather than assuming any one of the four gauges was the “truth” and given that they are being compared to a grid cell estimate, I used the average of the four stations as “truth.”

 Source

 Total Rain (Mar-Aug 2018)

 Description

 Gauge Average  21.79  Average of the four gauges
 ClearAg™ API  21.68  ClearAg 2-km gridded estimate for author’s location
 AHPS Grid  22.82  NWS 4-km gridded analysis estimate for author’s location
 KOKC Gauge  24.52  Official NWS station of record, located 12.7 miles NW of author’s location

Both gridded analyses, which are heavily influenced by radar information, provided credible results. The ClearAg value was less than 1% different from the four-gauge average with a difference of 0.11” below. The AHPS was about 5% different, showing 1.14” more rain than the gauge average.

Since many agricultural weather services rely on the nearest trusted climate station for long term history, the accumulation at the nearest NWS climate station, KOKC, about 13 miles away from my home, is shown. As expected, it had the largest difference from the gauge average, with 2.73” (about 13%) more rain reported.

So what are some key takeaways from this informal study? I offer these:

  • Even the best of gauge estimates have uncertainty that is greater than the measurement precision, especially once you exceed 0.5” of accumulation.
  • Gridded, virtual rain gauge estimates can provide a good estimate within the same level of error as other gauges over a representative area.
  • The collocated grid-based estimates are better than using a more distant trusted observation.

One final question to consider: Do these levels of error matter? Certainly, when you accumulate values over a long period, especially to months or years, a small percentage difference can appear quite large in terms of absolute depth. But, this may be misleading and irrelevant for the problem you are attempting to solve. I submit that the most common uses of rainfall for agronomic purposes are:

  • Ascertaining field workability issues
  • Managing irrigation to optimize plant growth (i.e., how much water is available to the plant’s roots)
  • Management of nitrogen or other products that are water soluble or that can be transported through the soil by water
  • Determining disease pressure or making spray decisions based on leaf canopy wetness

The reality is that for those applications, precision to the nearest 0.01” is unnecessary and gives a false sense of accuracy, given that a point measurement of that precision is only representative of that point, and doesn’t account for what happens when the water hits the ground or plant.

For example, imagine two extreme scenarios, both of which are a perfectly flat field. Scenario one is a field made of smooth, impenetrable glass. In this case, you can imagine that any water falling on any point on that “field” will immediately spread out across the field until it reaches an infinitely small depth. The second scenario assumes a field made of a steel “honeycomb” with vertically pointing cylinders. Any rain falling on a point in that field stays in the cylinder in which it falls, and goes nowhere horizontally.

Now, we all know that these are both unrealistic scenarios, but they both would be the easiest to model how much water depth exists at any point given a point measurement of rainfall. In reality, fields are rarely flat, nor are they uniform in their composition. So, in addition to measurement errors of a gauge at a point, the water itself may percolate to different levels, move horizontally, be intercepted by the plant canopy, and so on.

Quickly you can see how worrying about 0.01” accuracy at a point may be the least of your concerns. Additionally, it may be best to use an area-averaged estimate rather than a point method, meaning even though they don’t match a gauge, a gridded analysis of sufficient resolution may be better than a point measurement. This is one of the reasons I believe technologies such as coupled, predictive atmosphere, soil, and plant models offer the best solution for farm management decision aids over traditionally used methods such as those based on balancing rainfall and estimated evapotranspiration.

Accurate observations are important and useful, but one must also get to the root of what the real problem is, understand all the sources of uncertainty, and understand that modeling techniques offer a consistent approach to filling in gaps and getting to the desired decision much more efficiently and accurately.

Leave a Reply

Data Management Stories
Sentinel-1
Data ManagementSatellites Key in Providing Agriculture’s Next Generation of Data
September 27, 2018
With a range of extreme-weather events currently playing out across the globe, from floods on the east coast of the Read More
AmericasFarmers Edge Announces New Retail-Facing Precision Ag Platform
September 25, 2018
Farmers Edge announced the launch of a new solution designed exclusively for agricultural professionals to enhance their relationships with customers Read More
Soil-Hand
Data ManagementAre You Using Your Soil to Its Full Potential?
September 14, 2018
Harvest is progressing across most parts of the U.S. and those growers who aren’t already harvesting are gearing up to Read More
Tablet Grower
Data ManagementThe Power of Predictive Analytics in Agriculture
September 5, 2018
Years ago if we would have been told computers, data, and technology would be scattered around every farm there may Read More
Trending Articles
PrecisionAg InstitutePrecisionAg Institute Whitepaper Library
October 1, 2018
Welcome to the PrecisionAg Institute Whitepaper library! Here you will find free whitepapers and case studies from the PrecisionAg Institute Partners Read More
Sensors
Sensors/IoTThe Answer to Agriculture’s Daunting Challenges – Soil Sensors
September 20, 2018
According to the United Nations, 9.6 billion people will live on planet Earth by 2050. Feeding these mouths will require Read More
Soil-Hand
Data ManagementAre You Using Your Soil to Its Full Potential?
September 14, 2018
Harvest is progressing across most parts of the U.S. and those growers who aren’t already harvesting are gearing up to Read More
Grower-Retailer
Imagery/SensingAgtech: 10 Things I Hate About You!
September 4, 2018
Before you get bent out of shape from the title, remember if you’ve read my articles before you know I Read More
Kansas State University
Industry NewsKansas State University, Topcon Form Precision Ag Research Partnership
August 30, 2018
Kansas State University and Topcon Agriculture are collaborating to develop tools and systems to advance precision agriculture and support farmers. Read More
Blockchain building block graphic
Specialty CropsIs Blockchain the Future of Food Safety?
August 24, 2018
When the Internet Protocol Suite (TCP/IP) was standardized in 1982, permitting the worldwide proliferation of interconnected networks and eventually the Read More
Latest News
Auto-Thinner
Robotics/Labor SaversPrecision Agriculture in Specialty Crops: Labor, Cost S…
October 9, 2018
Agriculture is going through an equivalent of what the computer industry went through in the 1970s and 1980s. That’s when Read More
Growing-Innovations-logo
EventsA Crazy 2018 is Why We Need Growing Innovations
October 8, 2018
If you’re here you likely have received a special Meister Media Worldwide code to get $100 off your registration to Read More
AmericasConBAP 2018 Highlights Precision Agriculture in Brazil
October 8, 2018
The Brazilian Congress of Precision Agriculture 2018 (ConBAP 2018) was held October 2-4 in Curitiba, Paraná. Held every two years, Read More
GAR-Tootelian-Sensor
Service ProvidersGar Tootelian: Putting Ag Technology to the Test
October 8, 2018
In California’s Central Valley, a diverse crop market in which Pest Control Advisors (PCAs) are the dominant players when it Read More
Spray Drift
Decision Support SoftwareHow Technology Can Help You Prevent Spray Drift
October 8, 2018
Spray drift. News articles about damage caused by spray drift seem to pop up regularly, almost as daily reminders, during Read More
UncategorizedTopcon, Kansas State Make it Official: Establish Resear…
October 4, 2018
Today, Topcon Agriculture and Kansas State University announced a comprehensive partnership, according to a news release issued by the University. Read More
AmericasIvy Tech Students, Helena Collaborate on ‘The Fie…
October 4, 2018
Students in Ivy Tech Community College’s Agriculture and Precision Agriculture programs have had an interesting opportunity this growing season to Read More
Industry NewsSDSU to Host Precision Ag Bowl, Raven Facility Grand Op…
October 3, 2018
South Dakota State is the first university in the nation to offer a four-year degree in precision agriculture. This Saturday, Read More
Industry NewsProagrica: Connectivity in the Agriculture Supply Chain…
October 2, 2018
A recent survey of US ag retail businesses from Proagrica showed that future-proofing is one of the dominant pain points Read More
Thad-Becker
Business ManagementMFA: Mainstreaming Precision Ag Services
October 1, 2018
There’s no ironclad approach to building business through a precision program offering. There are as many nuanced approaches to success Read More
PrecisionAg InstitutePrecisionAg Institute Whitepaper Library
October 1, 2018
Welcome to the PrecisionAg Institute Whitepaper library! Here you will find free whitepapers and case studies from the PrecisionAg Institute Partners Read More
corn field
Industry NewsMFA Inc. Announces Expanded Partnership with Adapt-N
October 1, 2018
MFA Inc. has announced an expanded partnership between its Precision Advantage, Crop-Trak, and Nutri-Track service platforms and the Adapt-N nitrogen Read More
IBM-Watson
Industry NewsIBM to Launch Watson Decision Platform for Agriculture
October 1, 2018
When you think about artificial intelligence (AI), you probably don’t imagine using it for a farm. IBM is bringing data Read More
Aquaoso
Decision Support SoftwareNew Software Monitors Water Risk for California Agricul…
September 28, 2018
AQUAOSO Technologies PBC, a software company that helps customers identify, understand, and monitor water supply risks has officially unveiled their Read More
Sentinel-1
Data ManagementSatellites Key in Providing Agriculture’s Next Generati…
September 27, 2018
With a range of extreme-weather events currently playing out across the globe, from floods on the east coast of the Read More
Satellite
AmericasArgentina Bets on $600m Satellite to Boost Agriculture …
September 25, 2018
Argentina is launching a new microwave imaging satellite to monitor natural disasters and soil moisture, in a long-term bid to Read More
AmericasFarmers Edge Announces New Retail-Facing Precision Ag P…
September 25, 2018
Farmers Edge announced the launch of a new solution designed exclusively for agricultural professionals to enhance their relationships with customers Read More
FIRA
EventsAgricultural Robotics Take Center Stage at FIRA 2018
September 24, 2018
The future of agricultural innovation will continue to rely heavily on collaboration and exchange of ideas among all industry players. Read More