The Digital Harvest Approach
We build a comprehensive digital analog, a "twin", of cultivation, modeling it's entire growth trajectory and building a quantitative relationship between all the events and inputs on a field, and yield outcomes on that field. Our yield estimation engine starts with a time-stepping simulation of crop chemistry and biomass allocation. These produce not just one estimate, but a continuum of estimates for every day within the simulation window. Thus farm managers seeking to maximize operational yield have many "control variables", but the most common one is the date of harvest. A continuous yield curve allows clear and direct evaluation of any two harvest dates.
This yield estimation (YE) is a critical first step, but is ultimately just that; a first step. In your first year with YE, you can make a change, and either enter data about the change you made, or wait for the results of that change to be observed by satellite, and see the impact that change had on yield. In your second season, we'll use the first year's calibrations to estimate the impact a field action will have before you commit to doing it. Imagine the ability to try 3 different solutions in the digital realm, and select the best cost/value option to implement.
This process is flexible, requiring nothing more than a field boundary to start. Digital Harvest's road-map to digitization starts wherever you're organization is. We serve as your digitization and data analytics partner.
It all starts with a patch of land. The digital twin represents this patch of land with a polygon. If you already maintain a GIS database of field boundaries, you already have the first piece. If you're just beginning to digitize your operation, we have the tools to help you do it.
Data about field activities is next. Planting dates and harvest dates are most important, but data about varieties, cut numbers, soil tests, tissue samples, mechanical and manual harvests, and any other data used to manage the field makes YE more robust, and more useful to you. Even if you have nothing, not even plant dates and harvest dates, we can approximate these dates from satellite data and produce yield estimates.
A variety of satellite sources produce data bout the crop canopy, soil moisture, and weather for every DH customer. If your needs are specific, or you already engage with other commercial or science grade imaging products from private satellites, manned or unmanned aircraft, our engine can ingest those too.
Monitoring data provides a pulse reading of the crop's health, scalable from a few fields to tens of thousands. Deviations from expectations can be shown.
2.1 . DATA OVERVIEW
All Monitoring data is processed to surface reflectance with industry-leading calibration and processing. Clouds are identified and removed where applicable but may obscure useful data about a field. Digital Harvest fuses data from multiple sensors to get as much cloud free data as possible and performs advanced signal processing to fill in data gaps.
Monitoring data streams available in Digital Harvests core catalog:
SAR (synthetic aperture radar) - used in agriculture and forestry as a metric of canopy volume and shape characteristics. SAR is not impacted by cloud cover.
NDVI - spectral is a classic and ubiquitous measure of crop canopy "vitality". A healthy leaf reflects some wavelengths strongly but absorbs others. As a leaf becomes stressed, it becomes less reflective of infrared light, and heats up. NDVI highlights this vegetative response in spectral data.
MSAVI (modified soil adjusted vegetation index) - Similar to NDVI, but is less influenced by signal noise from exposed soil before canopy closure, and thus more useful early in the growth cycle. Cloud shadows that are missed by cloud masking algorithms are less disruptive to MSAVI.
LAI (leaf area index) - is a crop specific calibrated index that quantifies the leaf area within the crop canopy.
Hourly/Daily Weather largely focused on temperature and rainfall conditions are used. Derived growing degree day metrics are computed by simple integration using crop specific temperature sensitivity thresholds
3. YIELD PREDICTION
Nearing harvest season, yield estimation is one of the most critical planning steps a farming or processing operation will take. Digital harvest employs a combined approach to yield prediction, that fuses process-based biophysical modeling and remotely sensed crop characteristics with an AI (Artificial intelligence) layer to handle the volume of data and regionally specific parameters.
3.1 . MODELING SUGARCANE YIELDS
Digital Harvest's flagship crop is sugarcane.
Sugarcane is somewhat unique, in that most of the above-ground biomass contains usable yield.
If you are in a position where you must select one of two fields to harvest soon to keep the mill fed. But, both fields do not appear to have reached ideal maturity, which do you chose? With DH yield modeling, you can see that field "A"s sucrose yield is expected to increase more over the next two weeks than field "B". You can confidently make the decision to harvest field "B" today, and give "A" more maturation time. Now that you're familiar with the scenario, next season you are able to act earlier. Perhaps you decide to apply chemical ripener or cut off the water supply to field B earlier, aligning the logistically desirable harvest date with a date of maximum yield. Or, perhaps you find that field C is often given more growing time than it is efficiently using, and swap fields B and C in your schedule.
Customers with >99% of their field operations represented in the Visible Harvest platform can conduct special analyses with Digital Harvest to solve for operation-wide optimization problems such as:
Replant/ratoon and rotation strategy
Infrastructure investment impact
Climate and weather risk resilience