Remote sensing and field observations

A pattern to investigate

A grower sees a pale strip in an image of a vegetable field. It may be useful evidence: the strip gives a location to investigate. It does not yet establish why the crop differs there, or which intervention will help. A remote sensor records radiation reaching the instrument; an interpretation connects that signal to a property of interest.

Remote sensing can cover places that would take considerable time to walk. Field observations add information the image cannot supply directly, such as actual plant identity, damage, growth stage and root-zone conditions. The two methods become more useful when observations refer to the same area and a comparable time.

Bands and an index

An imaging band records a selected wavelength range. Green leaves absorb much visible light and reflect near-infrared light differently. One common summary, the normalized difference vegetation index or NDVI, compares near-infrared and red reflectance: NDVI = (NIR − red)/(NIR + red). Reflectance expresses a fraction of incident radiation reflected, so this ratio has no physical unit.

For an invented observation with NIR reflectance 0.50 and red reflectance 0.10, the difference is 0.40 and the sum is 0.60. NDVI is therefore about 0.667. The diagram shows this calculation; its arrows represent mathematical inputs, not a biological treatment response. The formula requires a nonzero denominator. Real product files may store scaled values, so the product definition matters before calculation.

One value, several possible explanations

An index compresses information. Differences in leaf cover or development can change its value, while atmosphere, background and cloud conditions can affect the signal or its usability. Dense vegetation can also reduce the sensitivity of NDVI to further change. Other indices address some limitations, but changing the index does not automatically identify a cause.

Compare equivalent observations before interpreting a difference. Consider the crop, growth stage, observation date and processing method. A low value might justify checking a sparse stand or stressed area; it does not, by itself, quantify missing nitrogen, establish a disease, or prescribe irrigation. A diagnosis requires additional evidence capable of distinguishing plausible explanations.

The area behind a pixel

A pixel is an image element associated with an area, rather than necessarily one leaf or plant. In an invented square grid, a 10 m × 10 m pixel represents 100 m². A 30 m × 30 m pixel represents 900 m², nine times as much area. A narrow affected strip can be mixed with surrounding crop within the larger pixel.

Map alignment and the sampled footprint also matter. A ground observation taken beside the mapped location, or from only a small unusual corner, may not represent the pixel. Record where observations were made and what area or plants they describe. This is a question of comparable measurement support, not merely producing a sharper-looking map.

Check the signal and the ground

Quality information belongs beside the image interpretation. USGS notes that Landsat spectral indices retain the constraints of their source reflectance products, and that source quality bands provide information beyond the index itself. A coloured map alone does not tell the reader which pixels were affected by cloud or other quality problems.

Plan field checks to compare contrasting mapped areas, while also including observations that can test whether the map misses important conditions. Keep methods and dates consistent where possible. Repeated observations can distinguish a persistent pattern from a temporary one. A targeted visit can explain a suspicious patch; it is not automatically a representative survey of the whole field.

From map to decision

A useful interpretation states the observed pattern, the evidence supporting its explanation and the uncertainty that remains. If measurements contradict the image-based interpretation, investigate timing, scale, quality and the proposed relationship rather than forcing the observations to agree.

Remote sensing can support scouting and comparison. Deciding on an application needs the relevant crop and soil evidence, while testing whether an intervention improves yield or quality belongs to field-trial design. Keep the original imagery, processing description and observations together so a later reader can understand how the conclusion was reached.

Worked examples

Check your understanding

1. A low-index patch coincides with fewer plants. Does that establish nutrient deficiency?

   Show answer and reasoning

   No.

   Reduced cover is an observation; different establishment, damage and other causes remain possible without discriminating evidence.

2. A ground count is taken a month after the image. What must a comparison acknowledge?

   Show answer and reasoning

   The crop may have changed between observations.

   A disagreement may reflect timing as well as scale, quality or an incorrect interpretation.

3. A map has detailed colours but no usable quality record. Is more colour detail evidence of reliable diagnosis?

   Show answer and reasoning

   No.

   Display detail does not establish valid source measurements or the biological cause of a pattern.

Key ideas

• Remote measurements, image-derived indices and diagnoses are different steps.
• Compare timing, spatial support and processing before comparing values.
• Use field observations and quality information to test an interpretation.