Engineering an End-to-End Geospatial Prototype With NASA Earth Data

Engineering Under Hackathon Constraints

NASA Space Apps is one of the world’s largest hackathons. Teams work with open NASA datasets and have only a short time to deliver a complete solution.

ARKNova engineers treated the event as an intensive engineering sprint. Their goal was not to build an isolated model or visualization, but to deliver an end-to-end prototype that connects data ingestion, processing, and user interaction into a single, usable flow.

They worked under several constraints: limited time to design and implement the system, large volumes of satellite and climate data, and the need for a complete pipeline with a usable interface on top of the data. These constraints shaped every architectural decision.

With only a weekend available, the team needed an approach they could fully implement and validate within the timeframe. They began by exploring SAR data, running Sentinel-1 imagery through calibration and terrain correction to test feasibility.

Sentinel-1 SAR imagery (VH and VV) over the Black Sea after preprocessing, showing a confirmed oil spill visible as a dark feature in the VV channel. Created by ARKNova team.

That exploration led to two deliberate decisions. First, the team chose to move away from SAR for this project. While SAR data is powerful, turning it into a robust and generalizable solution would require deeper scientific research than the hackathon allowed. Second, they focused on the BloomWatch challenge, which emphasized engineering execution, making it possible to deliver a complete data pipeline and a usable system within the available time.

For BloomWatch, the team chose NASA’s Harmonized Landsat–Sentinel (HLS) dataset. HLS combines Landsat 8/9 and Sentinel-2 imagery into a single, consistent dataset. That consistency made time-series analysis and vegetation indices such as NDVI practical without additional reconciliation work, allowing the team to focus on extracting meaningful signals rather than cleaning data.

Harmonized Landsat-Sentinel (HLS) false-color composite showing vegetation patterns. Image credit: NASA Earthdata.

Konstantin Polin, one of the ARKNova engineers, later shared a first-person account of the team’s technical exploration — from early SAR experiments to the final DemeterEye solution — in his LinkedIn post.

DemeterEye: Turning Satellite Data Into Field-Level Signals

During the hackathon, the ARKNova team built DemeterEye — a working prototype that combines a web interface and a mobile app for field-level monitoring. DemeterEye is not a commercial product; it is a proof of concept that demonstrates how open satellite data can be transformed into usable signals under real-world constraints.

The complete project, including technical details and artifacts, is available in the NASA Space Apps Global Project Gallery.

The system focuses on phenology detection, identifying key stages of crop development such as when the growing season starts and when it reaches its peak. These signals matter in practice. They influence irrigation, pollination, and crop protection decisions, and they can be detected reliably using vegetation time-series data.

DemeterEye architecture. Created by ARKNova team.

At the core of DemeterEye is an automated pipeline for geospatial data analytics, designed to process satellite imagery, build time-series, and generate actionable signals end to end.

1. Data ingestion
   The pipeline streams HLSL30 (Landsat 8/9) and HLSS30 (Sentinel-2) scenes from NASA’s CMR STAC LPCLOUD catalog and filters imagery by cloud cover.
2. Geospatial processing
   Scenes are clipped to user-defined field boundaries. The pipeline extracts the required spectral bands to compute NDVI and related vegetation indices.
3. Time-series construction
   For each field, the system builds multi-year vegetation histories that support seasonal comparison.
4. Forecasting and anomaly detection
   A Prophet-based model supports geospatial predictive analytics by estimating when the growing season starts and reaches its peak, while flagging deviations from historical patterns.
5. Climate data integration
   Daily weather data, including temperature, precipitation, humidity, and wind, adds context and helps distinguish natural variability from meaningful anomalies.

NDVI and weather time-series chart. Created by ARKNova team.

Together, these steps turn raw satellite imagery into field-level signals that update automatically.

But the system was designed not only to process data, but to make those signals usable in practice. DemeterEye does not stop at data processing. On top of the pipeline, the team built a simple interface that makes insights actionable.

With DemeterEye, users can:

• Define and manage multiple fields

• Explore multi-year vegetation histories per field

• See estimated season start and peak timings

• Detect anomalies such as delayed green-up or early senescence

Complex processing remains in the backend. Users see interpretable signals without needing to work with satellite imagery or raw data directly.

By the end of the hackathon, the team delivered a complete system with automated processing, a web interface, and a native mobile client.

Created by ARKNova team.

That end-to-end completeness — from ingestion to user interaction — is what ultimately led to the “Best Use of Data” award.

How This Engineering Approach Scales to Business-Ready Systems

DemeterEye is a prototype, but the engineering approach behind it reflects the kind of production systems Akvelon builds for real-world use. What the ARKNova team demonstrated during the NASA Space Apps Challenge was a repeatable way to turn complex geospatial and time-series data into operational signals, under tight constraints and with end users in mind.

At its core, this approach combines large-scale data ingestion, consistent geospatial processing, time-series analysis, and forecasting models into automated pipelines that update continuously as new data arrives. Just as importantly, it exposes results through interfaces designed for decision-makers, not data specialists.

This architecture scales beyond agriculture.

• In precision agriculture, similar systems support crop health monitoring, early stress detection, and yield forecasting across large field portfolios.

• In climate and environmental analytics, they enable vegetation monitoring, land-use change detection, and sustainability reporting.

• In infrastructure and asset monitoring, teams use geospatial risk analytics to track changes around roads, pipelines, or power lines and prioritize inspections based on detected risk patterns.

• In logistics, insurance, and risk modeling, Earth observation data supports faster impact assessment and more reliable, data-backed decisions after extreme events.

Across all these domains, the challenge is rarely access to data, but rather engineering systems that consistently transform raw signals into timely, interpretable insights, and keep doing so as data volume, geography, and operational demands grow.

That is the capability Akvelon brings to production projects: building end-to-end, scalable systems that move organizations from raw data to decisions they can act on.