Clean satellite data around the world and back in time:
Part of what makes our analytics so valuable is the fact that we craft them from an unusually long record of satellite imagery. The dataset we work with comes from satellites that have been orbiting the planet for nearly two decades! This means that we can analyse more harvests, more weather cycles, more extreme events. It also means we have to work both hard and smart to keep the dataset clean and consistent: We need more tools to visualize the temporal information.
At TellusLabs, our team has built in 15 year (2003-2017) means and standard deviations for each day of the year so we can automatically fill in remaining gaps in the Kernel database. The long term mean represents the 15 year history for each location, and has substantially less missing data than the current daily observations! These metrics enable us to derive the current day's (2018) anomaly from the long term mean based on a Z-score.
Long term standard deviations are yet another new layer we are adding to the Kernel product. For each location around the globe, we can show the normal variability expected for a given day’s observation based on this long historical record. For most locations the vegetation varies little from year to year for a given day. However, for cultivated areas during times of crop planting and harvesting, variance can be very high. For example, during the same October day in Iowa, the field’s first year could have fully mature corn but the next year could have already been harvested by that date (depending on the climate conditions during the growing season).
Therefore, high values of this index are indicative of where humans are growing crops, and more specifically, when the crops are being planted or harvested and are the most different from the surrounding natural vegetation.
This is also a solution for a real-time in-filling. Instead of seeing gaps in areas where there was no observation for that date, you will instead see the long term mean for that date. This will fill most of the black holes in the maps, making Kernel a one stop shop for all of your crop insight needs. Daily anomalies for each index will be available as separate layers on the mapping page.