EVALUATION OF COTTON YIELD MONITORS

 The foundation of precision agriculture is based on the ability to continuously monitor yield at harvest. Yield monitoring combined with the ability to establish a geographic reference for these data allows the producer to construct yields maps and track field performance from year to year. Two commercially available cotton yield monitors (Micro‑Trak inc. and Agriplan inc.) were tested for three years (1997-99) at research centers and growers' fields. In year 2000, the AgLeader (Figure 1) and Agriplan cotton yield-monitors were evaluated for their spatial accuracy. Field trials in 1997 and 1998 showed that the Micro-Trak and Agriplan units would accurately predict cotton weight when they remain clean and free from obstruction.  Results also showed that sensors to become dirty and obstructed very quickly under normal field harvesting conditions. As the sensing units became dirty, measurement errors reached up to 107%.  Even with cleaning before each load, errors greater than 5% were evident in data. Therefore, a positive air pressure mounting technique was developed by Agricultural and Biological Engineers at Clemson University to completely enclose the existing sensor effectively sealing it from environmental contamination.  The AirBox was pressurized by the picker fan, which forced air across the sensor eyes.

A John Deere model 9900 2-row picker was utilized for testing at the Edisto Research and Education Center, Blackville, South Carolina. Each picker chute was equipped with the latest versions of the both Agriplan and Micro-Trak sensors using the AirBox mounting technique (Figure 2). The yield monitors were tested in two 10-acre fields at the Edisto REC and a 15-acre field on a grower farm. The yield monitors were cleaned and calibrated prior to data collection. Sensors for both Micro-Trak and Agriplan stayed clean during the test.  Measurement errors for Agriplan unit were less than 5% in all three fields. Six yield measurements out of 25 had errors in excess of 5% for the Micro-Track system.  The positive air pressure and isolation from dirt, dust, and lint contamination kept the sensing units clean over 25 harvested loads.

During 1999, Micro-Trak sensors with Air-Box mounting systems (Figure 2)were evaluated on the Bozard Farms in Cameron, South Carolina.  Sensing units were installed on two chutes of a Case-IH 5-row picker model # 2055. The trial began with a cleaning and calibration of the yield monitor.  Eight basket loads of cotton were monitored in a 50-acre field.  Average error for the loads was less than 1%. However, error ranged from –12% to 7%. Since the sensing units stayed clean during the test, the variation in yield measurements could be because only two out of five rows of cotton were monitored.  In addition, sensor calibration was based on a single basket load. Calibrating the sensors utilizing several baskets loads could reduce measurement errors.

The Agriplan system was more accurate in detecting cotton yield levels than was the Micro-Track system especially in smaller fields. In 1999, the Agriplan Inc. adopted the positive air pressure technology to keep the sensing units clean and currently the system is commercially available. Yield maps were successfully produced with both yield-monitoring systems using global positioning systems (GPS) and geographic information systems (GIS).

In year 2000, four sensors of the new AgLeader Technology cotton yield-monitor and four sensors of the Agriplan Inc. were installed on both the front and rear chutes of a John Deere spindle-picker. Two pickers were used, one for plot work and one for large scale testing. Ground speed and fan speed sensors were installed on both pickers. These sensors are requirements for the AgLeader yield monitor.

Tests were conducted to evaluate the performance of these cotton yield monitors under grower’s field environments, and to determine the spatial accuracy of the individual yield data. For spatial accuracy test, 160 plots ranging from 35 to 60 ft were utilized.  Yield data were recorded in one-second intervals, which included 8 to 14 data points per test plot. Seed cotton yields were measured using sacking attachment and the yield monitors from 2 rows plots. The mean error from first field (64 plots) was 0.85% for AgLeader and 1.7% for Agriplan.  Measurement errors ranged from –7.6 to 7.4% for AgLeader and from –16.9 to 18% for Agriplan. 70% of the test runs for AgLeader and 65% for Agriplan had errors less than 5%. In another field (96 plots), 80% of the data for AgLeader had errors less than 5% with mean error of 0.4% ranging from –8 to 9.3%.  Errors for Agriplan in the same field ranged from –20 to 34% with only 30% of runs with less than 5% error. The AgLeader system was more accurate in detecting individual yield data points within a small area than was the Agriplan system.

In addition, the yield monitors were tested in three 15-acre fields at the Edisto REC and a 15-acre field on a growers’ field, near Elko, SC, to evaluate the performance of the yield monitors under large field environment.  A Trimble Ag132 GPS receiver was connected to the yield monitors for developing yield maps to determine spatial variability of cotton yield within a given area. The yield monitors were calibrated prior to data collection using four basket loads. The sensors were inspected routinely during the harvest and were cleaned as needed.

In growers’ field, all measurement errors for both yield-mapping systems were less than 5%. Errors for the AgLeader ranged from –2.9 to 2.8%. Yield maps were successfully produced with both yield monitors using global positioning systems (GPS) and geographic information systems (GIS). The yield maps were similar for this field.

Similar results were obtained in other fields. Measurements errors in both fields for the AgLeader unit were less than 5%. For the Agriplan system, 84% of the basket loads had error less than 5%. The measurement errors ranged from 6.8 to 6.9%. The flush mounting techniques for the AgLeader sensors eliminated the need for manual cleaning. The Agriplan yield monitor required sensor cleaning on daily bases. With proper maintenance both sensors can accurately predict seed cotton yields within an acceptable level. The AgLeader system is more users friendly than the Agriplan mapping system.

Figure 1. The Micro-Trak, Agriplan, and AgLeader cotton yield monitors.

Figure 2. AirBox enclosure mounted on a John Deere picker and AirBox enclosure surrounding Micro-Trak sensor on a Case IH picker (Note air supply lines from fan to AirBox).