top of page

WSU 2020 Digital Hackathon, pt. 2

Updated: Nov 6, 2020

Our collaboration in WSU's first Ag-entric Hackathon - read the backstory in pt. 1.

Continue reading for your chance to win a fashionable gift! The first 3 US-based readers to correctly decode the secret message in the D1G1TAL AgATH0N image win! Think you have the answer? Go to our CONTACT page and fill out the form with your answer & mailing address!


Growers are constantly faced with challenges. And millennial students are all about finding the answers. In early October of 2020, Washington State University, Microsoft,, & Cascadia Innovation Corridor co-sponsored the 2020 D1G1TAL AgATHON, an agriculture-focused hackathon at Washington State University. During this 5-day virtual event, the sponsors split 50+ students into 7 cross-discipline teams where they had the opportunity to apply their education & learnings to focus on data-driven agriculture scenarios leveraging Microsoft Azure. The teams were asked to select one of four different ag use cases:

  • Land use classification and monitoring using open-source satellite imagery

  • Phenomics data-based crop improvement

  • Automated digitization of rural roads & paths in Ethiopia

  • Improving grower outcomes based on Smart Orchard data

Notably, the event was kicked off by both WSU's Dean of the College of Agriculture, Human, & Natural Resource Sciences - Dr André-Denis Girard Wright, and the Dean of the Voiland College of Engineering & Architecture - Dr Mary Rezac , along with other leaders at WSU, Microsoft,, & University of British Columbia. As you'd expect, there was a boatload of energy & innovation behind this event...take a peek at our 3-minute recap video:


In the remainder of this blog post, we'll focus on the 3 teams that chose to focus around the data collected by for our Smart Orchard Project , with the aim to provide growers with insights that help better manage a variety of their practices.

For further context about the topics we'd encourage you to take the time to read these past posts: Imagery From Above, Soil & Irrigation Sensor Partners, "Using LiDAR to See in 3D", Agtech dealing with pests and disease.

Teams 1, 2, & 7: Data-driven Orchard Management

Team 1: Crop Load Estimation, Soil Moisture, and Canopy Vigor Relations

By- Uddhav Bhattarai, Chongyuan Zhang, Ben-Min Chang, Madhulika Gurazada

This team had the opportunity to work on two different implications of data driven orchard management where they chose to focus on the crop load estimation in apple trees using deep learning and an evaluation of soil moisture and canopy vigor relation.

Crop Load Estimation:

The industry's motivation and need for crop load estimation is the fact that this data is crucial in the various cultivars of different phases in fruit development (eg. budding and harvest). During the green fruit and budding period, crop load estimations help the grower determine how much and where thinning is needed. Later on when the fruit is getting ready to be picked, harvest-ready fruit crop load estimations are crucial for estimation of labor requirements, harvesting equipment needed, storage planning, streamline harvesting tactics, transportation logistics, & of course financial planning. Growers are increasingly interested in the benefits of a digitized crop estimation approach due to economics (instead of laborious amounts of time to count, all it takes is an image), accuracy (fast and affective in counting apples in the overall orchard), and less dependence on specialty labor.

Soil Moisture and Canopy Vigor Relation:

The importance of understanding and evaluating soil moisture and canopy vigor relationships is because these two factors strongly correlate with orchard productivity, fruit quality, optimized water usage and associated costs, and even attribute to the ability to optimize irrigation scheduling. Throughout Team 1's journey of assessing the data provided from the Smart Orchard, they were able to create an imaging processing algorithm that allows users to upload images with soil moisture and canopy vigor data that is then processed through an algorithm that calculates vegetation indices (VIs), creates a mask to separate the canopy, it selects soil moisture locations, and then extracts VIs from individual locations.

Original Orchard Block Image From Drone
NDVI Image
Image Identifying Only Trees With Fruit


Team 2: Drone Imagery Identifying Crop Health

By- Mathew Yourek, Ramesh Sahni, Femi Peter Alege, Ganeshram Krishnamoorthy

Throughout many of's posts we've discussed the various implications of ground based and aerial sensors providing growers with timely and spatially distributed data that monitors crop health. From soil moisture and nutrient probes, land-based vehicles with LiDAR, all the way to drones with multi-spectral imagery cameras. These kinds of sensors provide data pertaining to water management, crop health mapping, fertilizer management, disease identification, weed detection, plant energy and many other variables that all factor into the plants health and yield potential. Team 2 took a strong interest into this subject matter in order to find out how drone imagery can be used to identify areas of crop stress that may need special attention by the grower (e.g. additional nutrients, protection against disease, irrigation changes).

In order to accomplish their goals, they decided to focus on the Enhanced Vegetation Indexes (EVI) due to it having improved sensitivity in high biomass conditions. Leaf Area Index (LAI) is used to estimate foliage cover to forecast crop growth and yield. In the end, LAI assigns a quantifiable value to the amount of vegetation on the ground. It is the leaf area per unit ground area as seen when looking down on the vegetation. This team has then approximated LAI from the EVI using special equations that are crop and site dependent.

In their approach they decided to divide the field into 16 sampling regions in order to get a more accurate and precise calculation of all the variables. They then planned on calculating the mean of each parameter (NDRE, LAI, temperature from each grid) within each grid. All of this information would then allow them to identify all the possible areas of stress within the orchard block.

This shows that drone-based thermal imaging can be used to see in-field soil moisture variability and can be a tool for proper irrigation management.

After creating multiple maps pertaining to each parameter above, the team concluded that there was indeed a correlation between the amount of heat in each block and the amount of soil moisture. The thermal band shows the greater amount of variability showing that high leaf temperature was an indication of water stress. If the temperature was higher around that section of plants, then the average soil moisture in that section was lower. This then would help inform the grower to implement uniform irrigation management practices in order to maintain uniform moisture availability for the crop. Whereas, before conducting an analysis like this one, the grower would just be applying the same amount of water they normally do during those fluctuations in the heat instead of knowing exactly that block 12 was on average, maintaining a lower soil moisture content than block 3.