The Bronson Files, Dataset 8, Field 113, 2016

Dr. Kevin Bronson provides this dataset representing the first of three consecutive years of cotton and nitrogen management experimentation in Field 113. Included, is an intermediate analysis mega-table of correlated and calculated parameters, laboratory analysis results generated during the experimentation, plus high-resolution plot level intermediate data analysis tables of SAS process output, as well as the complete raw data sensor recorded logger outputs. Experimental design and operational details of research conducted are contained in related published articles, however a further description of the measured data signals and commentary is herein offered. Typical nitrogen fertilizer was delivered as liquid UAN 32-0-0 fertilizer with a density of 11.1 pounds per gallon, which contains 3.5 pounds of nitrogen per gallon. Notably, subsequent 2017 and 2018 experimentation years include a large volume of depleted nitrogen-15 isotope recovery tracing. GeoScoutX logging of CropCircle active optical reflectance sensing data - The primary component of this dataset is the Holland Scientific (HS) CropCircle ACS-470 generated reflectance numbers. Which as derived here, consists of raw active optical band-pass values digitized onboard the sensor product. Data was delivered as sequential serialized text output including the associated GPS information. Typically, this product examples a production agriculture support technology, enabling an efficient precision application of nitrogen fertilizer. However, we used this optical reflectance sensor technology to investigate plant agronomic biology, as the ACS-470 is a unique performance product, being not only rugged and reliable, but illumination active and filter customizable. Individualized ACS-470 sensor detector behavior and subsequent index calculation influence can be understood through analysis of white-panel and other known target measurements. When a sensor is held 120 cm from and flush facing to a titanium dioxide white painted panel, a normalized unity value of 1.0 can be set for each detector. To generate this dataset, we used a Holland Scientific SC-1 device and set the 1.0 unity value (field normalization) on each sensor individually, before each data collection, and without the use of any channel gain boost. The SC-1 field normalization device allows a communications connection to a Windows PC machine, where company provided sensor control software enables the necessary sensor normalization routine, and a real-time view of streaming sensor data. Noting that this type of raw value active proximal multi-spectral reflectance data may be perceived as inherently “noisy”; however basic analytical description consistently resolves a biological patterning, and more advanced statistical analysis is suggested to achieve discovery. Sources of polychromatic reflectance are inherent in the environment; and can be influenced by surface features like wax or water, or presence of crystal mineralization; varying bi-directional reflectance in the proximal space is a model reality and directed energy emission reflection sampling is expected to support physical understanding of the underling passive environmental system. We consider these CropCircle raw detector returns to be more “instant” in generation, and “less-filtered” electronically while onboard the “black-box” device, than are other reflectance products which produce vegetation indices that are averages of multiple detector samples. Soil in view of the sensor does decrease the raw detection amplitude of the target color returned and can add a soil reflection signal component. Yet that return accurately represents a largely two-dimensional cover and intensity signal of the target material present within the field view. It does not represent a reflection of the plant material solely, because it can contain additional features in the view. Expect NDVI values greater than 0.1 when sensing plants and saturating more around 0.8, rather than the typical 0.9 of passive NDVI; because the active light source does not transmit energy to penetrate perhaps past LAI 2.1, which is less than what is expected with a solar induced passive reflectance sensor. However, the focus of the active sensor scan is orientated on the uppermost expanded canopy leaves, and those leaves are normally positioned to intercept the major of incoming solar energy. Active energy sensors are easier to direct, where this capture method targets a consistent sensor height of 1 m above the average canopy height, and a roaming travel speed maintained around 1.5 mph, with the sensors parallel to earth in a nadir view. Holland Scientific 5Hz CropCircle active optical reflectance ACS-470 sensors that were measured on the GeoScoutX digital propriety serial data logger, have a stable output format as defined by firmware version. Different numbers of csv data files were generated based on field operations. Raw data files include the inserted null value placeholder -9999. CropCircle sensors supplied data in a lined format, where variables were repeated for each sensor creating a discrete data row for each individual sensor measurement instance. Hamby rig active optical reflectance data was generated by Holland Scientific CropCircle ACS-470 sensors, which were numbered 1, 2, 4 and 5, where sensors 1 and 5 had band-pass filters centered at nanometer frequency 550, 670, and 530, while sensors 2 and 4 had filters 590, 800, and 730, each for their respective R1, R2, and R3 raw detector data channels. The placement of the filters was determined as a generic optimization, where the longer wave filter was put in the middle detector position, and where the tandem sensor setup was optimized for the favored NDRE on one sensor and a green frequency test configuration on the other. Although when facing forward, there is a left and right side for the two cotton rows measured and where data was tracked and processed accordingly, the two cotton rows were not considered different experimentally. Therefore the possible two crop row variability was not considered. CropCircle raw data adjustment approaches - Three undescribed adjustment value test calculation data columns are included, appended to the original raw data tables. For each CC sensor detector, the white panel observed amplitude delta of the raw reflectance channel was used to create minor data adjustments. This calculated test value was appended to the raw data table as variables R1_adj, R2_adj and R3_adj, and example a possible raw data minor adjustment. This was the beginning period of a method advancement, in testing control based normalization adjustments to raw active optical detector data values. Generic and course post-process raw data adjustments can be made by first measuring a white panel reference at 120 cm distant, before and / or after a data collection period, which is beyond using only the SC-1 device to normalize individual sensor detectors. A deviation from the flat white reflectance typical 1.0 unity value was recorded and used to offset the detector raw radiance values. The raw data adjustment test approach was developed as an extension of the manufacturer’s normalization routine recommendation, which uses the SC-1 device or a titanium dioxide ultra-white painted custom panel. Normally, the ACS-470 detector channels would be set to read 1.0, after 30 minutes of warmup time and when connected to the SC-1 illumination reflector, or when held 120 cm away from and facing an optically flat white panel of sufficient size to fully reflect the active light signal footprint (about 30 x 100 cm). This recommended approach does work well. We normalized multiple sensors in a field condition, by using the typical two tailed white panel field-normalization approach. One by one, each sensor was connected to the SC-1 box for communications with a PC, where the sensor real-time information was viewed and a sensor normalization command given. Once placed at an appropriate height and position relative to the white panel, a sensor zero point as measured was ascribed to the sensor configuration, by first covering the active optical LEDs source and detectors and creating a black-out condition, and then immediately afterward revealing the illuminated white panel in full detector view where a second full signal measurement was made and the unity 1.0 set point value instructed to the sensor. Values streaming through the active optical sensor detectors typical range 0-2% around the unity value after field normalization, while in a natural condition measuring the course surface white control panel. Therefore successful normalization was deemed to have occurred, or was not needed, when all detectors were within 2% of the 1.0 value when using the white panel setup. It was difficult to achieve a 1% data range for all the detectors at all times, where multiple iterations of the normalization routine would not consistently yield improved results of 1% magnitude. Therefore, we simply measured the typical 0-2% raw data value difference for each detector, with the idea that a subsequent adjustment may be possible. We found that we could measure longer time periods with sensors over a white panel reference and determine optical signal features, as well as elucidate individual sensor minor behavior. Temperature change apparently induced effects on the raw detector data stream. We also recorded the sensors when connected optically to the SC-1 device reflector, in dark conditions, at various distances and angles from a target, and with many different types of target reflectors, in a temperature control room, laboratory, shop and outdoors. We noted that each detector of each sensor can exhibit unique behaviors, which underlie the customizable band-pass color filter’s effect. Some detector channels increased with increased temperature but most decreased. The raw data magnitude shift, including the filter, was typically 0.3% per degree C, yet it was variable. However these detector behaviors were stable and repeatable. Therefore the following initial test method adjustments were considered. Adjustment approach #1 was a pre-collection white panel based adjustment. This is the most common approach, where after a period of sensor warming and after the signal data stream fully stabilized, the typical 0-2% average sample difference from the unity value 1.0, was added back to the raw values of each detector, before the VI calculation. Understanding that proximity drives signal amplitude, the 1.2 m distance from panel to detector during normalization is related to the 1 m above average canopy height in the field. One color small raw detector channel possible offset bias effect can either compound or mitigate VI error, based on the second channel color detector possible offset bias. In practice, mitigation was more typical, where detectors that were to be normalized in ratio drifted in the same direction at the same time, although detector drift was usually not to the same magnitude. An expected VI calculation drift might be 0.1% per degree C. Adjustment approach #2 was a post-collection white panel adjustment. Where the average sample difference from unity, as measured reflectance from the white panel at the set height and after the field data collection, was simply added back to each of the raw detector values. Adjustment approach #3 involves one average value, or a linear interpolation of the pre (before) and post (after) field collection white panel measurements. Measuring the white panel before and after a field collection allows for improved adjustment potential, by creating a condition of bookends of control measurements around the experimental field measurement. There was no need to stop log between white panel measurement in full sun and the field collection or the sensor’s return to its white panel position. Rather one continuous data log can be useful to validate data quality features outside of the experimental or control recordings. The initial sensor warmup consideration was important, because the active light source created a somewhat spherical heat artifact that originated on the LED source, and that spread across the detector physical area and throughout the rest of the sensor body. Although perhaps 90% of the heating occurs in the first 30 minutes of a calm condition warmup, a minimum full hour was operationally favored to achieve closer to 95% warming, or rather three hours warmup time to a 99% thermal effect and best condition achievement. The goal is holding a consistent sensor temperature. Furthermore, sensors that are in full sun condition will warm quicker and become warmer than those in the shade. Fully warming sensors within solar impingement and then normalizing them just before a field collection gave the best thermal condition result. Regarding thermal condition after a field collection, a measured white panel with the CC sensors in full sun gave a more accurate thermal data condition, than when the rig was returned to the shade of a garage where the temperature of the sensors would drop a few degrees rapidly lowering the sensor thermal body status from where it had been in the field. There is an environmental or weather consideration; typically we encountered ambient warming during the field collection time period, which would boost the sensor thermal value after the pre collection white panel measurement. Conversely, a cool breeze event would cause a relatively rapid sensor body temperature decrease. We were even able to measure minor solar thermal loading on the sensor body that would change with rig orientation (data not presented). Obviously a sensor in the sun will be warmer than one in the shade. A passing cloud or shade event can change sensor body surface temperature and propagate a weak thermal pulse to the detectors. We wrapped sensors in external insulation, to reject abrupt external changes in temperature and smooth the thermal change effects that did occur. Insulation apparently mitigated our transient solar loading influence and the cool breeze effect. The post collection adjustment approach worked well for instances where sensors came to their highest temperature early in the experimental collection, soon after encountering the full sun and field conditions. Adjustment approach #4 takes into consideration ambient and equipment temperature measurements to select whether the pre (before), post (after), or both control values could best adjust for mitigating minor thermal artifacts, and at which periods of the field data collections those artifacts may have occurred. Pre and Post field data collection datasets are available for the CC sensors, presented as extra csv files which depict white panel unity sampling. Data shows raw channel static reflectance signal quality. Noting that the adjustment approach is technical in nature and ancillary to the primary research investigation. Following the CC manufacturer’s original guidance was considered sufficient for typical operation. The adjustment values are offered as example only, while the primary raw data values are presented as they were originally recorded. The adjustment values were appended as part of the original GSX processing and constitute actual research process values that were considered by Dr. Bronson at the time of investigation. Therefore, use the R1, R2 and R3 variables from the HS raw data table to access the original data, rather than the white panel secondary adjusted test values. Campbell Scientific CR1000 logger data - The Hamby rig platform upgrade involves measurement on the CS logger of two crop row surface temperatures and heights, plus the inclusion of an air temperature and relative humidity measurement recorded with sensors roaming above the canopy. The GPS receiver output followed RS232 communications protocol and involved DB9 format cable connections. NEMA RMC and GGA strings were recorded at 5 Hz. The RMC and GGA UTC variables were viewed as dual time recordings and became the primary key coordinate data table values. They were parsed from their GPS strings and recorded in a redundant fashion, so if one variable was missed, the other might still be present. Sometimes the CS logger had trouble decoding every serial string transmitted. If start and stop characters were not recognized within 200 ms of program time, the string variable may have returned NAN and the memory pointer moved and or variable buffer cleared. Therefore, a more CPU compliant and stable approach of recording the GPS string term as a single variable in the data table proved better to allow more complete CS based GPS NEMA string comma separated data recordings, that could be parsed in post processing and which contained the CS recorded location data. The previous AirTemp CS variable was replaced with the variable Air, after a hydrometer sensor replaced the active TC setup. Although we lost the aspirated lower mounted sample point approach, we gained a latent energy component in the Rh, and began measuring temperature with relative humidity above the canopy and perhaps 2 m above the ground. This compared to the typical environmental sampling from the adjacent AzMet weather station. The VectorNav VN-100 was not used, because in previous operations the CS CR1000 data logger was having trouble processing all the serial information fast enough and was dropping data table lines from memory. Dr. Bronson did not want to pursue the minor accelerometer corrections possible. Rather already knowing direction of travel, the GPS course was sufficient information. Precision sensor position offset calculations from the GPS receiver were not conducted. Instead sample points were spatially associated with the GPS receiver position, which was about 0.5 - 1 m behind the sensor positions on the left and right armatures. 07/19/2016 The RMC_UTC_Az variable was a local time conversion used for display as a public variable. It did not record correctly to the CS data table due to a wrong data type and returns 7999. A serial communication processing error number that may occur is -2147483648. The CS logger timestamp was parsed to a set of time variables as part of the development around database indexing and decimal time calculation testing. The millisecond logger time variable carries the highest precision time incrimination, while the GPS based UTC one second accurate satellite generated pulse provides the true time. There is a CS time stamp string error that affected intermediate files where the SAS generated csv outputs were corrupted by Excel auto formatting of the time stamp string variable, and then those changes were saved and the variable disregarded as it was not needed by Dr. Bronson for final analysis. Therefore, corrupted CS timestamps are present in the intermediate files. However note that the time stamp string is correct in the CS raw data files presented. Recording of error flags on the CS logger was done to support user feedback, where the user would be signaled by a buzzer when a data parameter was out of bounds. The user could see which of the primary metrics was out of bounds and the number of program iteration flags generated by looking at the logger LCD where real time data tables were visible. Later in the data quality control action, post collection processing of the total number of flags and the times of their occurrence were evaluated along with the data samples. Taken together as a modular data package, the GeoScoutX and Campbell Scientific logger, were connected to the same GPS receiver, and positioned with keypad interfaces and displays near the rig operator’s view. A 100 watt solar panel was attached above the operator’s roll-cage shade plate to provide 12 volt power boost during field collections. This can be useful in cases where batteries are drained unexpectedly or have reached reduced capacity near end of life. Having solar power generation supplying a roving data system is recommended to support full battery power status during the important field operations. Avenger Rig data note - The separate platform Avenger Rig’s eight CropCircle ACS-470 sensor array was used to support initial season collections while Hamby was in the farm shop. Sensor Numbers #267, 335, 303 and 301 have filters 590, 730 and 530, where sensors #217, 256, 264 and 333 have filters 550, 800 and 670, each for their respective R1, R2, and R3 designated columns in the intermediate file. Avenger Rig Honeywell sensor #3 is added for use in crop height determination on the second preliminary collection 5/23/2016. Avenger rig data processed by Dr. Bronson at the time of experimentation, is presented here as extra to the official Hamby collections, yet provides additional relevant comparative and extended numeric. In total four Avenger rig collection events, which included additional metrics, were conducted. Specific instance data notes - Known data errors include small data gaps in the intermediate files that are likely the result of a table merge error in the SAS process; the raw data is not gapped. 05/13/2016 Avenger rig CC data was utilized in lieu of the typical Hamby CC data, and is represented by an intermediate file of SAS output with a padded numeric form. At the start of the growing season, the Hamby rig needed a repair and so the Avenger platform was substituted to take the initial measurements. Avenger raw data, including additional data metrics collected are not presented. 05/23/2016 Again the Avenger rig CC data is presented, including the addition of one Avenger Honeywell ultrasonic sensor (#3), providing a crop height determination. Only the intermediate SAS output is offered, representing the data considered at the time of experimentation. 05/28/2016 The HS GeoScoutX (GSX) data log was started with the recording the typical CC sensors. Raw data presentation is started and with the start of the primary experimental data set collection. 06/01/2016 The CS data log is started alongside the GSX data, this configuration represents the first Hamby data package, although only one Honeywell ultrasonic sensor was in operation. 06/14/2016 The second Honeywell ultrasonic sensor was included in the data. This created a left and right side (as the operator is looking forward on the rig) two-plant-row ultrasonic displacement sensing setup to compliment the two-row CC reflectance. 07/12/2016 Only HS raw data is available, the CS data was in error, likely due to a table mismatch onboard the logger. If a previously created binary table is not deleted from the CF card logger memory expansion, and there is not additional room on the card for a new table as determined by the running program, then the logger will not delete the present table and instead error by not writing new data to the card memory. The CF card can be deleted and formatted onboard the logger if the operator is aware, as there is a table mismatch display error offered. Summary - Active optical proximal cotton canopy sensing spatial data and including additional related metrics are presented. Agronomic nitrogen and irrigation management related field operations are listed. Unique research experimentation intermediate analysis table is made available, along with raw data. The raw data recordings, and annotated table outputs with calculated VIs are made available. Plot polygon coordinate designations allow a re-intersection spatial analysis. Data was collected in the 2016 cotton season at Maricopa Agricultural Center, Arizona, USA. High throughput proximal plant phenotyping via electronic sampling and data processing method approach is exampled using a modified high-clearance Hamby spray-rig. Acquired data conforms to location standard methodologies of high-throughput plant phenotyping. The weekly proximal sensing data collected include the primary canopy reflectance at six wavelengths. Lint and seed yields, first open boll biomass, and nitrogen uptake was also determined. Soil profile nitrate to 1.8 m depth was determined in 30-cm increments, before planting and after harvest. Nitrous oxide emissions were determined with 1-L vented chambers (samples taken at 0, 12, and 24 minutes). Nitrous oxide was determined by gas chromatography (electron detection detector).