Over time agronomists can get very good at examining the most recent satellite image of a paddock with a familiar spectral index and color scale applied to extract a large amount of information very quickly. It is a tool that improves human productivity and can help them catch issues that they may not have known existed. From a typical ground level inspection, an agronomist may see only 20% of a field and even then, variability in the crop health across the field will be hard to pick up. Imagery gives them a bird’s eye view and enables them to target in field inspections. When using satellite imagery to aid with crop monitoring and more efficient scouting process I would recommend considering the following process.

1. Understand the paddock history

Many paddocks farmed today used to be several smaller paddocks. We can sometimes identify where old paddock boundaries were 10+ years after they were combined. Similarly, if a paddock was split in half for just one season, that split can often be seen in the following crops. It’s important to know this paddock history as it impacts the overall variability. A lot of this history will come to light in step two.

2. Investigate for ‘normal’ variability

You can’t skip this step. You need to take time to understand the ‘normal’ variability that exists in a paddock. After looking at several years of satellite imagery it’s possible to see common trends that persist season to season. We are lucky now to have 5+ years of Sentinel 2 data and almost 10 years of Landsat 8 onwards. There are multiple services that attempt to automate this by taking an image at peak biomass over multiple years and running statistics on each layered pixel to produce a single map which can be useful. I would recommend taking the time to do it manually especially if you have spent little time looking at imagery of the field in the past.

The variability will differ depending on the type of season, crop type and management. An unusually wet season causes waterlogging on low lying areas cause plant death and low production. These same areas may likely to be highest producing in a very dry year. Although it can differ by season type, there is almost always some consistency over the years in how variability of crop growth persists throughout different seasons.

A few tips when doing this:

• Don’t always just look for the peak of the growth cycle to identify normal variability. Often at peak biomass much of the paddock may be saturated appearing to be uniform across a large area. See if there is some imagery available a few weeks before the peak and this may give a flatter distribution of values across a paddock identifying more variability.
• Make sure you understand how the software is visualising the imagery. Some will use a static color scale where the color you see always represents the same value, and others will always stretch the color to match the data in the field. There are pros and cons to both of these but it’s important to be comparing apples with apples. A dynamic scale is OK if the range is sufficiently wide enough as well are mostly looking for the patterns here instead of absolute values.
• Try to understand what has caused there to be variability in the paddock. Other imagery sources can help here. See if you can find a bare earth satellite image or look at the high resolution image on Google Earth – does it simply follow a soil color trend? If not, that is interesting.
• What’s normal isn’t always what is acceptable. What I mean by this is, there might be a spot in a paddock that always have bad ryegrass showing that it’s ‘normal’ for this area to indicate high plant growth but it is actually detecting the weeds.
• See if the variability pattern moves through the fence line. You need imagery of the neighbouring paddocks for this. If it stops at the fence line, ask yourself why.

3. Inspect the latest image

Armed with the knowledge gained in step two when looking at the latest image, here we look at it through two lenses

1. Trying to spot anomalies. An anomaly is where the imagery bucks the normal variability – either higher or lower than normal and often in an unusual pattern. Your job is to now see if you can explain them just by using your knowledge of the field and the data in the satellite image. And then verify your assessment when in the field.

A typical Sentinel 2 or Landsat 8 with NDVI or similar spectral index applied has been known to express some of these anomalies:

• Weed infestations expressed as higher NDVI values
• Spray drift
• Waterlogging
• Hail damage
• Large areas of disease such as ascochyta blight (although if this imagery can capture this you have problems)
• Feral pig damage
• Nutrition deficiencies

1. Targeting a low, mid and high plant growth areas. Examples of why to target these different areas include:

• Desiccation decisions
• Targeting high biomass areas as early insect pressure starts here
• Harvest timing

Don’t forget to inspect closely anywhere you know a trial or accidental trial has been conducted. Accidental trials are where maybe the fertiliser bin ran out, the fungicide ran out, the planter was planting double rate etc.

4. Consider Timed difference / change maps

Inspecting how the crop has changed over a period of time is also valuable to understand the crop development. Some software allow you to create a map that shows the change on images captured on different dates. For a 10 day timed difference map it is very simple formula of [current NDVI] – [10 day old NDVI]. Positive values are increased crop health or biomass and negative values are reduced. This new piece of information provides another perspective to examine crop growth. A single image will only give you a total measurement. A difference map provides direction and amount of change. It’s common for different soil types to impact the speed at which plants will move through growth stages.

5. Plan your next trip to the paddock

Now that you have understood as much as you can about the paddock from the imagery you can visit the paddock and make sure all anomalies can be explained and you have investigated the right places to gain a more accurate representation of the crop. Understandably, the imagery doesn’t account for all issues or variability in the paddock but it puts you leaps and bounds ahead of if you had not used it at all.

Other gotachas and tips:

• If something ever looks strange always consider that the imagery could be at fault. E.g. cloud shadow is often missed and this will play havoc with NDVI values causing what looks like anomalies
• Some in crop sprays can impact leaf angle and this can alter the value. An example is Tordon 242 in wheat and barley.
• NDVI is not the only spectral index to consider. It’s worth testing out several available algorithms. Some others such as MCARI2 can give similar results to NDVI but won’t saturate to the same extent in high biomass crops.