Agriculture Remote Sensing: Getting a Closer Look
By John Enterline

Since the early 1980s, proximal sensing-collecting spectral reflectance data close to crop plants- has allowed researchers to detect crop stress long before detection with the human eye was possible. Now on the verge of integration into widespread site-specific farming applications, this technology could enable farmers to diagnose crop deficiencies in real time, and then remedy yield-threatening problems immediately. Moreover, proximal sensing could significantly reduce input costs by allowing farmers to provide fertilizer, pesticides, and water strictly on an "as needed, where needed" basis. Walter Bausch, an agricultural engineer with the USDA's Agriculture Research Service water management research unit, has been using just such a spectral reflectance technique for over a decade to develop a nitrogen reflectance index and reflectance-based crop coefficients for corn. The outcome: an important step in real-time crop deficiency detection, and the possibility of landmark farming equipment such as an automated water/fertilizer mechanism based on field-pivot or linear move irrigator systems.
      Like its satellite and airborne observation cousins, proximal sensing characterizes reflected solar radiation from the crop canopy to uncover information about a crop's health. Walter Bausch uses Exotech (Gaithersburg, Md.) radiometers to monitor reflectance in the green (520-600nm), red (630-690nm), and near-infrared (760-900nm) portions of the spectrum-similar to the Landsat satellite Thematic Mapper wavebands TM2, TM3, and TM4. In addition, Bausch uses an Analytical Spectral Devices (Boulder, Colo.) FieldSpec spectroradiometer to monitor the continuous spectrum from 350nm to 1050nm, in order to identify additional areas within the visible near-infrared that may provide better crop health information. Customized software interprets the spectral reflectance data and assigns it Global Positioning System (GPS) coordinates, then translates the information into a map of the crop's water or nitrogen status. One significant difference between proximal sensing and satellite or airborne observation, however, is the "pixel cell" scale of the images. Simply, a pixel cell represents the smallest resolvable area in the scene being monitored. While satellite and airborne observation typically detect clear images in a pixel cell scale of from 25 to 100 square meters, proximal sensing allows the clear resolution of pixel cells as small as six inches-although one meter is more typical. At this higher spatial resolution, crops such as corn can be monitored as early as the V-6 (sixth leaf) growth stage, and in fine detail.
      Bausch began his spectral reflectance research on corn in the late 1980s, centering his research in experimental cornfields near Fort Collins, Colo. It was there that he developed a set of crop coefficients derived solely from reflected canopy radiation data. Bausch knew that abnormal climatic conditions such as an unusually cold, wet spring would cause slower plant growth. When this occurs, traditional crop coefficients, which are based on the percent of time from planting to effective cover (75 percent ground cover) and elapsed days after effective cover, would overestimate actual crop evapotranspiration (ET) and result in over-irrigation. Since reflectance-based coefficients (Kcr) are derived directly from crop conditions, Bausch determined that they would give much better assessments of crop ET.
      Over three growing seasons, he compared his reflectance-based method with two other crop coefficient methods-simulating irrigation scheduling with each, and tracking the number of irrigations and estimated crop water use. When Bausch compared the averaged Kcr results with those derived from SCHED (USDA-ARS Irrigation Scheduling Program), the number of irrigations was reduced by 15.6 percent, and estimated crop-water use was reduced by 14.5 percent. When he compared Kcr with a basal crop coefficient method (still based on effective cover dates), the differences were much smaller. However, reflectance-based irrigation often occurred three to five days earlier, or two to three days later. The difference: Bausch's results were based on actual crop conditions, so irrigations were in direct response to climatic-based plant changes.
      This research led Bausch to another development. In corn, most nitrogen is bound by the chlorophyll, a substance easily quantified by a set of wavelengths in the visible and near-infrared portions of the spectrum. Collecting spectral reflectance data throughout the 1995 and 1996 growing seasons, Bausch developed a corn Nitrogen Reflectance Index (NRI). The NRI is the ratio of a plot's near-infrared/green wavelength measurements to the same ratio in a well-nitrogen-fertilized reference plot. A reading of less than/equal to 0.95 indicates that the crop is nitrogen-deficient. When Bausch compared his experimental corn crop's NRI against its Nitrogen Sufficiency Index (NSI)-an index determined by plant samples collected with a Soil Plant Analysis Development (SPAD) chlorophyll meter-he discovered a near 1:1 relationship. However, the NRI provided a significant advantage over the NSI. In order to determine the NSI, as many as 30 SPAD samples must be taken from different plants within the crop and their nitrogen results averaged, whereas the NRI can be determined from reflectance information gathered from a number of plants at one time. Simply put, when monitoring an entire crop, collecting spectral reflectance is faster and easier.
      Bausch's ASD FieldSpec spectroradiometer is mounted on a high-clearance tractor that has been modified for this particular application. "It's got a boom with an instrumentation platform that's just more than 10 meters in the air," says Bausch. "We can drive quickly through the field, collecting data, with no damage to the crop." In fact, his FieldSpec unit incorporates two spectrometers with wavelength ranges of 350nm to 1050nm, each equipped with a fiber optic cable. In one case, the cable snakes from the instrument out to a sensor head on the boom to detect canopy radiance, and transmits the signal to the unit within the tractor cab. In another, the cable extends from the unit inside the cab to the outside, where it extends toward the sky in order to measure incoming solar irradiance. Simultaneous measurements and intercalibration of the radiometers determine the crop's reflectance. The computer program that calculates canopy reflectance also associates the data with the GPS coordinates. The Geographic Information System (GIS) calculates water use, records the NRI, and maps spatial variability.
      The next step: Bausch's emphasis on when to apply water and fertilizer is shifting toward how much to apply. His current experiments attempt to correlate reflectance-based canopy nitrogen information with traditional soil sampling estimates of available nitrogen in the root zone. "To really save farmers money," says Bausch, "we're going to have to decide exactly how much input they need to apply. Early nitrogen deficiency detection is critical because, if it goes too far, you can't get the plant out of the deficient state, and you end up using much more nitrogen than the crop would have otherwise required." Fertilizer applied in this "when needed, as needed" fashion could significantly reduce crop nitrogen input during the growing season, and maximize yield by keeping corn plants continuously healthy. "Until the last few years, farmers would go out and apply nitrogen fertilizer across an entire field, based upon a recommendation that came from a few soil samples taken from here and there in their fields, then lumped together and analyzed," he says. "There's also grid sampling to get variable rate applications. I think most agronomists agree that this method is just too expensive. This technique would be more efficient in terms of nitrogen use and could be coupled with crop-water-use information into a single package."
      It is the concept of a "single package" that interests sensor manufacturers such as ASD. According to Brian Curtiss, ASD director of research and development, one possibility includes the development of a real-time fertilizer/water metering system. According to Curtiss, "Field-pivot or linear-move irrigation systems could be modified to support radiometer sensors and associated software. Since the pivot goes around the field 24 hours a day, it's an ideal platform to incorporate sensors for tracking spectral reflectance and weather data. Specialized software could interpret this information and trigger when and how much fertilizer or water gets distributed." Rugged and self-contained, a real-time metering system would also be freed from the timing problems associated with satellite imagery, as the frequency of those images depends on how often the satellite passes overhead.
      Walter Bausch is skeptical about the degree of modification current equipment would have to undergo before it could meet the real-time, in-the-field demands of a field-pivot system. Instead, his vision includes a tractor-based stress detection system that would enable farmers to address crop deficiencies within a 24-hour period. "As far as applying what we know about the correlation between crop health and its reflectance data, I'd like to say we'll assess the situation today, and by tomorrow there's action being taken to remedy the problem. I really think we can save inputs this way," says Bausch.
      Certainly, proximal sensing as a widespread, site-specific farming method hinges on significant equipment modifications. For example, radiometers are sensitive to sun angles and weather conditions. The fact that they rely on solar radiation to detect reflectance makes them dependant on daylight hours and sunny weather conditions. According to Curtiss, one possibility is to create software that "weeds out" the data associated with non-optimal conditions. However, equipment manufacturers still need to address the nighttime hours and extended periods of cloud cover.
      Moreover, current spectral reflectance data demands interpretation by an expert. Future agricultural systems that incorporate this technology will need to become completely automated or developed in order to provide farmers and crop consultants with information, rather than merely data. Bausch and Curtiss both agree that the current generation of crop consultants would likely administer such a system. According to Curtiss, "Going to the farm-consultant model makes a lot of sense. The difference is that, rather than doing soil samples on a quarter-acre grid once every two years, they'll be getting the same information by driving a sensor around on a tractor, or collecting it in real time."
      Most importantly, sensor manufacturers will need to rely on researchers such as Walter Bausch to refine which wavelengths are most critical in diagnosing plant health. Current research instruments cover an entire wavelength range and cost in the mid-$30,000 range. By narrowing to select wavelengths, instrumentation costs could drop to a few thousand dollars. Selecting a limited number of wavelengths would also allow manufacturers to make a more robust instrument-one able to withstand dirt, cold, heat, and rain without disorienting its optical components.
      Even with these obstacles, researchers and equipment manufacturers predict proximal sensing devices will be in the field within the next five years. "The potential for this technology to optimize a farmer's crop yield and also save money by reducing the necessary input is incredible," says Curtiss. "If you look at how agriculture has changed over the past quarter-century, you see significant fine-tuning of how and when to apply input, and the development of new fertilizers, pesticides, and seed types. Enabling the detection of plant stress at the very earliest moment-whether by satellite, airborne observation, or a land-based system-is the next step in farming evolution."

About the Author:
John Enterline is the director of sales & marketing for Analytical Spectral Devices. Appointed to the position in 1998, Mr. Enterline has over 15 years experience marketing high-tech analytical instrumentation.

Back