Contents

(Search or Access Pages Above and Below)

Crop Agriculture

 About Landsat for Crop Applications

Related pages:  Landsat Correlation to Corn Yield    Landsat Correlation to Soybean Yield 

Field Location Relative to Landsat Scene Extent

Current and historical crop images shown below are from Landsat 5 Thematic Mapper (TM) or Landsat 7 Enhanced Thematic Mapper Plus (ETM+) satellites. Landsat imagery is free, dates back to July of 1982 (Landsat 4 TM) so historical assessments of a field can be made, and includes the near infrared (NIR) band which is important for crop yield prediction and crop condition assessment. Landsat has 30 x 30 meter pixels which equates to 4.5 pixels per acre and a 16-day revisit cycle. A Landsat image (scene) is about 185 (e-w) x 175 km (n-s) (115 x 106 miles) and has overlap areas, particularly on the east and west sides, so fields fortuitously located in an overlap area have more imagery available. Presently, Landsat 8 (LDCM; February 2013 launch) & 7 ETM+ are operational. In 2011, Landsat 5 TM, (the longest operating Earth-orbiting satellite) experienced a problem and stopped providing imagery; the satellite provided data for 27 years, 24 years more than designed for. Although Landsat 5 TM imagery has been suspended, past data is available and has many uses. Landsat 5 and 7 raw data pixels values represent digital numbers (DNs) which are calibrated radiance values that range from 1 to 255 (8-bit); radiance is the amount of radiation emitted from a surface. Landsat 8 data is 12-bit. When satellite data are discussed, reflectance (total radiation emitted from a surface / total incoming radiation that strikes a surface) is usually referred to.  Depending on the application, conversion to reflectance by atmospheric correction may or may not be necessary. 

Below is Landsat 7 ETM+ NIR imagery on 7/31/05 with missing data shown. Landsat 7 developed a problem in 2003 that causes striping of missing data; the stripings starts near the middle and get wider towards the sides.

Landsat 7 stripings of missing data

 

Below is eastern edge of Landsat 7 NIR image (same image as above)

Landsat 7 stripings of missing data widen toward east and west sides.

 

The boundaries of pixels for Landsat 4, 5, 7, and 8 are in the same location and do not change. The positional accuracy of the data can be off somewhat and is inconsistent in the amount and direction of the error, so the data within the pixel boundary may represent an area that is not precisely within the extent of the pixel. For example, a pixel may actually represent data starting 10 meters in any direction. As a result, there can be a pixel boundary that extends beyond the edge of a field but the data within the pixel represents only the field; in this case, the pixel should be included as valid data for a field. Conversely, there can be a pixel near the edge of a field that is entirely located within a field but the data within the pixel represents some of the surface outside the field; in this case, the pixel should not be included as valid data for a field. Typically, however, at the boundary of a field there are pixels (that cross the boundary) that represent a meaningful amount of surface inside and outside of the field; these pixels should not be used. As a result, there is usually not valid data at or near the edge of the field; however this is also commonly the case with yield monitor data, particular in the headlands area and the area that is void of yield data between the headlands and main body of yield points (this is shown in the Yield Map Cleaning and Mgmt.folder). In fact, valid satellite data can extend into the void near the headlands, providing information where only interpolated yield monitor data can exist. Structures such as electrical installations and corresponding shadows or tree shadows can be averaged into and affect pixel reflectance; and assessment for this should be made in order to determine which pixels are valid. Sometimes there are obvious situations where Landsat imagery needs to be moved to improve positional accuracy to apply the data to a field, but this uncommon.

For crops, Landsat can be applied in specific ways at specific growth stages for different crops for fields large enough, with the right proportions for the 30 x 30 meter pixels (good to have at least thirty pixels), and when the canopy is essentially closed. When sensing crop condition or predicting yield patterns, imagery should be used when the impact of soil background has diminished enough (this even applies to soil adjusted vegetation indices); this is, for example, at about V12 for corn. Visible bands saturate too much when the canopy closes and should not be applied (solely or in indices) for field mapping at that point. Conversely, near infrared (NIR) is usually effective at sensing crop condition when the canopy is closed (also shown above) because, unlike visible bands, NIR can see through many layers of leaves. Corn is a main exception of NIR being useful when the canopy is closed; this is due to the effect of tassels. To acquire usefully high spatial correlations between Landsat and corn yield at the field level, imagery should be from V12 to just prior to VT and should apply an index that includes NIR and visible wavelengths (MSAVI has been shown to be better than other indices for corn). For crops that do not have a prominent feature such as tassels obscuring green vegetation, NIR reflectance, solely, from later growth stages should be applied; for example, there are high correlations with soybean yield from R2 to R6.