Why you’re seeing a big step in cover
Seeing the farm average fall by a couple of hundred kg DM/ha overnight is unsettling. In almost every case the grass is still there—a poor-quality satellite image has slipped past Pio’s automated filters and nudged the model off course.
Why it happens
Every day Pio ingests thousands of paddock images and runs them through machine-learning quality filters that reject cloud, haze, smoke, harsh shadow and other artefacts. Very occasionally an image looks “good enough” to the filters but is still hazy. When that image feeds the biomass model it can either:
-
Depress the reflectance signal → biomass is underestimated
-
Inflate the reflectance signal → biomass is over-estimated
How to confirm it’s an image issue
-
Open your farm → Satellites panel (on your Farm Dashboard)
-
View the Scan Date and check the satellite image on the day before and (if there is another one) on the day after the jump.
- Preview the images:
- Click on the relevant date and view the satellite image on your map.
-
A clean image shows crisp paddock boundaries and clear contrast.
-
A hazy, cloudy (foggy) image looks washed-out, slightly blurred, or grey-coloured (see the GIF below for a side-by-side example).
💡 Unsure if the image is bad? Switch to another satellite overlay or send us the date—happy to confirm.
If the change in pasture cover lines up with a hazy image, you’ve found the culprit.
Hazy, cloudy and shadow vs clean satellite images
Examples of hazy satellite image (below the clean example):
Clean image:
Clean satellite image (crisp boundaries and pixel definition)
Hazy image:
Hazy satellite image (smeared boundaries and loss of pixel definition)
Cloudy image:
Cloudy satellite image!
What to do next
-
Let us know – email support@pasture.io or use in-app chat with the scan date(s).
-
We’ll remove the bad image and re-run the model for your farm (typically the same working day).
-
Your feed wedge will update automatically; no manual editing required on your end.
Can this happen again?
Rarely, yes. The quality filters block well over 99 % of poor images, but no automated system is perfect. Any time you see a cover jump or drop that defies common sense, follow the steps above or drop us a line – we’re happy to check.
Frequently asked questions
Could heavy grazing or weather cause the same drop?
Heavy grazing can drop individual paddocks, but it won’t pull the whole-farm average down by 200 kg DM/ha overnight unless every paddock was grazed. Sudden weather alone won’t move covers that far either – so a farm-wide step is almost always a data quality issue.
Does the model learn from each mistake?
Yes. Each confirmed hazy or bad image helps us refine the filters so similar images are caught automatically in future. This is an ongoing learning for Pio's machine learning image classification system.
Can’t the model just reject any image that causes a big change?
Sounds simple, but there are genuine cases when this doesn't work. For example, when a farmer sprays out the whole farm before re-sowing (i.e. kikuyu going to annual grass). So, we can’t rely on a blanket “big change” rule without losing valid data.
How can I remove a bad satellite image?
Currently there isn't a feature that allows you to remove a bad image. All you need to do is let us know in one of the support channels along with the date the image aligns with and we will manually remove the image and re run the models.
Need a hand?
If you’re unsure, just ask. Pop a message into the online chat, or email the date of the suspect scan and we’ll verify it for you – no guesswork needed. Accurate data is critical for good grazing decisions, and we’re here to keep it that way.