Use of Remote Sensing Imagery for Improving Crop Management Decisions
Crop Insights written by Bob Gunzenhauser1 and John Shanahan2
- Remote sensing is collecting reflected light information from objects like crop canopies using remote platforms such as satellites, aircraft or ground-based platforms.
- In a 2013 pilot program, DuPont Pioneer is providing remote sensing imagery services to growers through Pioneer® Field360™ services.
- In-season imagery from RapidEye is provided by Satshot, a national distributor.
- This imagery can be displayed from a mobile device such as an iPad® or other tablet and can be used for directed field scouting.
- Images can be used to develop management zone-directed soil sampling schemes, validating hybrid tests or evaluating other agronomic practices on your farm.
Remote sensing is defined as collecting information about objects (e.g., soil or crop surfaces) from remote platforms like satellites, aircraft or ground-based booms. This practice involves the collection and analysis of reflected light and is a potentially important source of data for making site-specific crop management decisions. Remote sensing tools can provide information not only about spatial variability within fields (Figure 1) but also about changes in crop conditions throughout the growing season (NRC, 1997). Obtaining time-based information that is spatially distributed using remote sensing is often less difficult and less expensive than using other site-specific crop management tools. In addition, sampling intensity is virtually unlimited, unlike gridded soil sampling or similar practices (Moran et al., 1997).
Figure 1. Aerial color-infrared image depicting spatial variation in crop vigor for several fields. Images courtesy of Cornerstone Mapping (cornerstonemapping.com).
There are 2 main types of remote sensing tools: passive and active. For passive systems, reflected sunlight is the source of energy measured. Examples of passive systems include satellite sensors, camera film or digital cameras placed on aircraft. Active sensors also measure reflected light but instead of relying on natural sunlight, emit their own source of light. The use of several commercially available active crop sensors for improving corn nitrogen (N) management was discussed in a previous Crop Insights (Shanahan, 2010). This article provides an overview of passive remote sensing concepts, examples of image providers, and a discussion of how these tools can be used to improve crop management. Some of these tools are currently available through Pioneer Field360 services.
Remote Sensing Basics
The electromagnetic (EM) radiation emitted by the sun and measured with passive remote sensing devises comprises a whole spectrum of wavelengths, ranging from short-wave gamma rays to long-wave radio frequencies. For most commercial remote sensing applications, the visible (440-690 nm) and near infrared (NIR) (760-900 nm) regions of the EM spectrum are typically utilized. Objects like crop and soil surfaces absorb or reflect various EM wavelengths due to their unique physical and chemical properties. For example, Figure 2 and Figure 3 illustrate how chlorophyll present in corn leaves absorbs more blue (450-520 nm) and red (630-680 nm) and less green (520-600 nm) light. This results in higher reflectance in the green band and is why plants appear green to the human eye. Compared to visible light, plants absorb much less NIR light (Figure 3). In other words, plants reflect more light in NIR wavelengths, with percent NIR reflectance increasing as crop biomass increases.
Figure 2. Corn leaf cross section depicting the interaction of EM radiation of various wavelengths with different leaf anatomical components.
Figure 3. Reflectance spectrum for corn plants receiving 4 rates (0, 50, 100 and 150 lb/acre) of N fertilizer. These plants differ in “greenness,” which can be measured as differences in reflectance in the visible spectrum.
The reflectance characteristics of visible and NIR light from crop canopies form the basis for the development of numerous vegetative indices. One of the first indices developed was the Normalized Difference Vegetation Index (NDVI), which is calculated using light reflectance in the red and NIR bands (Tucker, 1979). The formula for calculating NDVI is as follows: NDVI = (NIR - Red) / (NIR + Red).
Values for NDVI range from -1.0 to +1.0. In typical remote sensing operations, output ranges from 0.1 to 0.9, with values ranging from 0.1 to 0.2 for soil surfaces and 0.2 to 1.0 for crop canopies (NDVI values increase as both crop biomass and greenness increase). Differences in spatial patterns of biomass observed in imagery are often associated with overall plant health and spatial patterns in grain yield (Figure 4), especially for images acquired during mid to late grain-filling stages for corn (Shanahan et al., 2001).
Figure 4. Near infrared (NIR) aerial image and yield map show similar spatial patterns. Images taken by Cornerstone Mapping and courtesy of USDA-ARS.
Satellite Imagery Providers
Publically Available Satellite Imagery: The longest running program for acquiring satellite imagery of Earth is Landsat (landsat.usgs.gov/index.php), a joint venture between the NASA and USGS agencies of the U.S. government. The first Landsat satellite launched on July 23, 1972, and the most recent, Landsat 8, launched on February 11, 2013. The instruments on the Landsat satellites have acquired millions of images that are archived in the United States and at Landsat receiving stations around the world. These images are a unique resource for agricultural applications and are free to the public. The Landsat 8 sensor measures reflectance in 8 spectral bands, including the previously mentioned visible (blue, green and red) and NIR bands. The spatial resolution for this imagery is 30 meters, and the temporal resolution (time interval between image acquisitions) is 16 days.
Commercially Available Satellite Imagery: In addition to publically available Landsat imagery, there are numerous other commercial sources of satellite images that can be purchased for agricultural applications. One provider is the German-based RapidEye Company
(www.rapideye.com) which deploys a 5-satellite constellation equipped with Multi-Spectral Imager (MSI) sensors. A unique feature of the MSI sensor is that besides the traditional visible blue (440-510 nm), green (520-590 nm), red (630-685 nm) and NIR (760-850 nm) bands, it also measures reflectance in the red edge (690-730 nm) region of the spectrum (for more information, see the RapidEye white paper at: www.rapideye.com/upload/Red_Edge_White_Paper.pdf
The uniqueness of the red edge band is its location between the red and NIR regions, a portion of the spectrum where reflectance drastically increases from the red toward the NIR plateau (Figure 3). The red band is an area where chlorophyll strongly absorbs light, and the NIR band is where the leaf cell structure produces a strong reflection (Figure 3). Therefore, slight variations in both chlorophyll content and the amount of biomass or leaf area are often better distinguished using the red edge band. In fact, several studies have shown that the red edge is more sensitive in detecting subtle differences in chlorophyll content and hence crop nitrogen status than any other spectral band (Eitiel et al., 2007 and Schlemmer et al., 2013). The imagery acquired by the MSI sensor is delivered at a spatial resolution of 5 meters and temporal resolution of 5.5 days.
Beside spectral characteristics, other important features of remote sensing systems include the spatial and temporal resolution at which imagery is delivered. “Resolution” is the term used to describe the size of the cells or pixels that make up imagery. For example, Landsat imagery is delivered at a 30-meter resolution while RapidEye images are distributed at 5 meters (Table 1). Hence, pixels making up Landsat images are 30 x 30 meters square vs. 5 x 5 meter pixels for RapidEye imagery. This means Landsat imagery contains only around 5 pixels per acre while RapidEye images possess upwards of 160 pixels per acre. Therefore, RapidEye imagery is considered to be "higher resolution," able to distinguish much finer spatial detail than can be seen with the "lower resolution" Landsat imagery. In summary, growers have a host of options for obtaining satellite imagery that is useful for improving crop management decisions, ranging from publically available Landsat imagery to commercial sources of images.
Uses of Imagery for Improving Crop Management
Five different RapidEye imagery analyses provided by Satshot are currently available in Pioneer Field360 services. These are: NIR, NDVIR (red NDVI), NDVIG (green NDVI), red edge, and NDVIRE (NDVI red edge). Examples of these analyses (Figures 5a to 5d) for an image acquired from an irrigated corn field in the western Corn Belt in mid-July are shown to illustrate how these analyses can be used as crop scouting/management tools.
NIR: The NIR analysis shown in Figure 5a represents the output from the NIR band of the MSI sensor on board the RapidEye satellites. It typically shows the greatest amount of detail or variability of all the analyses. Differences in the amount of total biomass or leaf area, canopy architecture, and soil background effects can all influence NIR reflectance. This is clearly evident in the NIR image (Figure 5a) where it can be seen that the non-irrigated areas around the center pivots exhibit dramatically lower NIR values, indicating reduced biomass levels associated with less available water. The NIR analysis also shows the change in hybrid and planting date which occurred on the east side of the field. The lower NIR values for this north-south oriented strip indicate that soil background and canopy architecture effects were dramatically different for this hybrid and planting date compared to the remainder of field.
NDVIR (Normalized Difference Vegetation Index - Red): As previously discussed, NDVIR is a simple vegetation index that involves use of the NIR and visible red bands. Where soil background or hybrids may influence biomass readings in the NIR analysis, NDVIR uses the visible red band to “normalize” these readings (Figure 5b). The resulting difference in the 2 indices is demonstrated in Figure 5 below. The NDVIR map shows a less dramatic effect of the change in planting date and hybrid on the eastern side of the field.xxx
Figure 5. RapidEye images of an irrigated corn field acquired on July 12, 2013, displayed in Pioneer Field360 studio. 4 different spectral analyses are represented: a) NIR, b) NDVIR, c) NDVIG and d) NDVIRE. Changes in color from red to green represent increasing values.
NDVIG (Normalized Difference Vegetation Index- Green): The NDVIG uses the same formulation as NDVIR but employs the green band instead of the red band. Research by Gitelson et al. (1996) showed that NDVIG is more sensitive than NDVIR in detecting variation in canopy greenness, especially at high vegetative levels beyond the V9 growth stage. This is evident when comparing NDVIG (Figure 5c) with NDVIR (Figure 5b) where greater differences in color and more distinct patterns can be seen in the field. Research has even shown that NDVIG imagery acquired during early to mid-grain filling could be used to improve on the spatial detail and accuracy of yield maps (Dobermann and Ping, 2004).
Red Edge and NDVIRE: These 2 analyses involve use of the red edge band unique to RapidEye imagery. The NDVIRE uses the same expression as NDVIR but substitutes the red edge band for the red band. Because NDVIRE uses the red edge band, it is much more sensitive than NDVIR or NDVIG in delineating variation in canopy greenness or chlorophyll content (Figure 5d). This can be seen when comparing NDVIRE (Figure 5d) with the NDVIR (Figure 5b) or NDVIG (Figure 5c) analyses for this field. The nitrogen stress observed in Figure 6 is more apparent in the NDVIRE than the NDVIR or NDVIG analyses.
Figure 6. Photo taken on July 12, 2013, in field shown in Figure 5d depicting nitrogen stress for 24-row strips receiving only 25 lbs/acre of preplant N (corn on right).
Because there is a close correlation between canopy chlorophyll content and nitrogen status in plants (Schlemmer et al., 2013), NDVIRE images acquired after canopy closure (around V9 to V10) could potentially be used to generate variable-rate nitrogen prescription maps for side-dress N application. In addition, it is possible to use these maps to assess the efficiency of previous fertilizer applications and to determine zones with chronic nitrogen deficiencies.
This discussion has provided an overview of remote sensing concepts, examples of RapidEye image analyses currently available in Pioneer Field360 services, and potential uses of these analyses for improving crop management. Below are the steps required to receive RapidEye imagery in Pioneer Field360 services and suggested applications for using the imagery.
- Visit with your Pioneer sales professional about enrolling fields of interest in the RapidEye imagery program.
- Once enrolled, all images acquired for a given field (along with the 5 analyses shown) will be automatically displayed in Pioneer Field360 services. These images can be displayed on screen along with other GIS layers already present (e.g., soil type, other historic imagery or yield maps) to better explain spatial variability in crop yields for the field.
- Imagery does not replace the need for crop scouting. Instead, it directs growers to areas of the field that require ground truthing.
- Images can be displayed in Pioneer Field360 Select software, a mobile version for iPad® or other tablet devices, and can be used for directed field scouting.
- Images can be used to develop management-zone-directed soil sampling schemes, validating hybrid tests or evaluating other agronomic practices on your farm.
Dobermann A. and J. L. Ping. 2004. Geostatistical integration of yield monitor data and remote sensing improves yield maps. Agron. J. 96:1588-1597.
Eitel, J. U. H., D. S. Long, P. E. Gessler, Smith, A. 2007. Using in-situ measurements to evaluate the new RapidEye satellite series for prediction of wheat nitrogen status. Internat. J. Rem. Sens. 28:1-8.
Gitelson, A.A., Y.J. Kaufman, and M.N. Merzlyak. 1996. Use of a green channel in remote sensing of global vegetation from EOS-MODIS. Remote Sens. Environ. 58:289-298.
Moran, M.S., Y. Inoue, and E.M. Barnes. 1997. Opportunities and limitations for image-based remote sensing in precision crop management. Rem. Sens. Environ. 61:319.
[NRC] National Research Council. 1997. Precision agriculture in the 21st century: Geospatial and information technologies in crop management. Rep. 59-0700-4-139. NRC, Washington, DC.
Shanahan, J.F., .J. S. Schepers, D.D. Francis, G.E. Varvel, W.W. Wilhelm, J. M. Tringe, M.R. Schlemmer, and D.J. Major. 2001. Use of remote-sensing imagery to estimate corn grain yield. Agron. J. 93:583-589.
Schlemmer. M., A., Gitelson, J. Schepers, R. Ferguson, Y. Peng, J. Shanahan, and D. Rundquist. 2013. Remote estimation of nitrogen and chlorophyll contents in maize at leaf and canopy levels. Internat. J. of Appl. Earth Obser. and Geoinformation. 25: 47-54.
Tucker, C.J. 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 8:127-150.
1 DuPont Pioneer Technical Applications Manager, Johnston, Iowa.
2 DuPont Pioneer Agronomy Research Manager, Johnston, Iowa.