Area Monitoring — Combining Markers into Decisions

Some examples on how it all fits together

This post is one of the series of blogs related to our work in Area Monitoring. We have decided to openly share our knowledge on this subject as we believe that discussion and comparison of approaches are required among all the groups involved in it. We would welcome any kind of feedback, ideas and lessons learned. For those willing to do it publicly, we are happy to host them at this place.

The content:

• High-Level Concept
• Data Handling
• Outlier detection
• Identifying built-up areas
• Similarity Score
• Bare Soil Marker
• Mowing Marker
• Pixel-level Mowing Marker
• Crop Type Marker
• Homogeneity Marker
• Parcel Boundary Detection
• Land Cover Classification (still to come)
• Minimum Agriculture Activity (still to come)
• Combining the Markers into Decisions (this post)
• The Challenge of Small Parcels
• The value of super resolution — real world use case
• Traffic Light System
• Expert Judgement Application
• Agricultural Activity Throughout the Year

In the previous posts, a number of the markers have been described. This article discusses how these may be used in real-life scenarios. Below we describe a few interesting features of interest (FOIs) with potentially wrong claims based on previously described markers.

Example 1

Claimed to be a permanent meadow, shown to be corn

• FOI seems not to be similar to other meadows from the neighbourhood (similarity score is 8.07). 99.8% of all FOIs with a meadow claim in the dataset have a similarity score with meadow hypothesis less than 8.07. The same target FOI has the smallest similarity score if compared with other cornfields in its neighbourhood. In this case, the similarity score is 0.48. Only 0.1% of FOIs with a meadow claim have a similarity score for corn hypothesis less than 0.48.

• The claim of the most similar FOI from the neighbourhood is corn (maize for silage). Intra-class rank is 1 indicating that target FOI is an outlier among meadows.

• A number of bare soil observations found by the marker is found to be 23, which is too high for meadows.

• Mowing marker output cannot be interpreted as mowing events since target FOI is not a meadow. The output itself is, of course, interesting since it finds two very long events at the beginning and the end of the mowing interval, which is consistent with a summer crop (the first event is an indication of sowing and emergence of the crop, while the last event indicates harvest).

• Predicted crop group is corn.

Based on the NDVI profile, the number of bare soil observations, crop model prediction, and similarity score, it is a cornfield, which is also confirmed by the digital orto-photo image from the same year.

Example 2

Claimed to be corn.

• A similarity score for corn hypothesis is very high (above 13). Based on the similarity score this FOI is most similar to summer barely.

• A claim of the most similar FOI from the neighbourhood is winter wheat. In average, distance to FOIs claimed as summer barley in the neighbourhood is the smallest. Intra-class rank is 0.97 indicating that target FOI is an outlier among cornfields.

• A number of bare soil observations is 11. Bare-soil observations are distributed over three periods.

• Two events are detected with the mowing event. Again, since this is not a meadow the events cannot be interpreted as mowing events. In this case, the first event indicates the harvest.

• Crop type model predicts winter barley.

Based on NDVI profile and information provided by the markers it is probably a summer barely or other summer cereal.

Example 3

Claimed to be winter wheat, shown to be more likely summer wheat.

• A similarity score for winter wheat hypothesis is high. Most similar to summer oat.

• A claim of the most similar FOI from the neighbourhood is potato. In average, distance to FOIs claimed as fallow land in the neighbourhood is the smallest. Intra-class rank is 0.86 indicating that target FOI is not very similar to other winter cereals.

• A number of bare soil observations is 14. One group is detected at the beginning of the year, which is inconsistent for winter wheat, but consistent for summer cereals.

• Mowing detector detects one event. Since this is not a meadow this event cannot be interpreted as mowing. It can be however as the harvest of summer oat.

• Crop type model predicts summer cereals.

Based on everything this is summer and not winter cereal.

Example 4

Claimed to be mixed fruits, shown to be a house.

• A similarity score for mixed fruits is high. Found to be most similar to vegetables (combines many different vegetable types.)

• There’s another FOI in the neighbourhood claimed to be a grass-clover mixture. Although the average distance is the smallest to FOIs claimed to be sunflower the comparison of target’s NDVI and mean NDVI of sunflower FOIs shows large discrepancies. Intra-class rank is 1.00 indicating that target FOI is an outlier among other mixed fruits.

• A number of bare soil observations is 18.

• Mowing marker detects no events.

• Claimed as mixed fruits; DOF shows it is a house.

All data indicates that this is not mixed fruits. DOF shows it is a house with a garden and a tree.

Example 5

Claimed to be a meadow, shown to be a built-up area.

• Similarity score indicates that this FOI is not similar to anything. It is most similar to Vegetables.

• Plots based on Euclidian distance support the same conclusion. Intra-class rank is 1.00 indicating that target FOI is an outlier among other meadows.

• Bare soil marker detects bare soil in most of the observations.

• Mowing marker detects no events.

DOF confirms that it is a built-up area.

Our research in this field is kindly supported, in grants and knowhow, by our cooperation in Horizon 2020 (Perceptive Sentinel, NIVA, Dione) and ESA projects (Sen4CAP).