MOBILE DEVICE-BASED LOCATION SERVICES ACCURACY

The objective of this study was to characterize the expected position accuracy when using popular mobile devices for location-based agricultural decision-making activities. This study utilized Android-based Nexus 7 tablets and tested the operation of the three location services available on this system in a 24-h fixed location test and a shorter duration multiple location field test. In the 24-h test, the "network" location system had a measured error of 37.19 m while reporting an accuracy of 55.56 m. The "gps" location system had a measured error of 2.57 m and a reported accuracy of 3.20 m. Multiple tests were conducted with the location system added by Google Services code because the measured error was much higher than the reported accuracy. With this system, the measured errors were 14.13, 3.4, 24.08, 14.01, and 16.15 m with reported accuracies of 3.95, 4.83, 3.99, 7.18, and 6.68 m, respectively. All the tests with the Google Services location system had much higher variability in location estimates than the "gps" location system. For both services, the high values for reported accuracy did not correspond with high values for measured error. Field testing was only performed with the Google Services and "gps" location systems as the "network" location system did not operate in the test field. Statistical analysis confirmed that the "gps" system was more accurate in this testing but the difference was not as dramatic as in the 24-h testing. The average reported accuracy level was 3.0 m in all field tests with the "gps" system and 3.9 m in all field tests with the Google Services system. The field test data were also used to estimate areas of 0.14-ha rectangular plots. Among all three tests with the "gps" system and all three tests with the Google Services system, the mean absolute area percent error varied from 4% to 7%, and in every test at least one plot was over-or underestimated by at least 10%. The error characteristics and patterns for all but the "gps" service differed significantly from the random walk pattern and/or other characteristics of GNSS locators to which precision farming engineers have become accustomed. Mobile platform creators like Apple and Google are either requiring (Apple) or strongly encouraging (Google) developers to switch to newer services that don't provide access to the underlying locating mechanism. Therefore, it is clear that careful consideration of these differences and what they may mean to location based apps in agriculture will be important. This work highlights the importance of testing any "smart" devices to determine actual location accuracy before relying on them for making agricultural decisions based on their output.