kai

Three weeks at the Biosphere 2, and the barley growth experiment is nearly complete. The biomass accumulation plateaued a few days prior with overall CO2 production beginning to fall.

Overall, the experiment went very well. The data harvested is solid, from start to finish. While at moments there were what felt like major hurdles to the desired steady-state environment surrounding the experiment, one comes to realize that with 15,000 to 70,000 data points (depending upon the instrument) over 12 days, a half hour of direct sunlight on a grow chamber, or a 15 minute power loss to the fans and subsequent CO2 build-up has no affect on an overall trend.

With three CO2 sensors, one at the outlet of each grow chamber and a third at the inlet to both chambers (as they are situated perpendicular to each other, sharing inlet air space, we found the daily fluctuations to match that of known CO2 correlations to temperature.

The data shows solid trends in all five of the parameters captured. While we came into the experiment principally interested in CO2, hoping to capture the photosynthetic draw-down of CO2 once the barley chlorophyll was activated, we learned that our seedbed was too thick, the underlying seeds remaining in a O2/CO2 respiration phase that kept the CO2 in the chambers higher than anticipated. Yet for the 30 minutes each day of direct sunlight (due to a gap in the shade cast by the LEO structure), we did see drastic reduction of CO2 in the data. We had considered full sunlight in this experiment, to invoke a higher CO2 draw-down, but knew it would be difficult to model the real-world weather (full sun, partial or full cloud cover, even a snow storm as occurred). What’s more, in a lunar or martian habitat, it is anticipated that all greenhouses will be located in lava tubes or buried beneath regolith to provide radiation protection for both the plants and humans that tend to them.

While we had originally been concerned for of our ability to capture the increase in biomass (plant structure + water retained), in fact, we were quite successful. The digital scales with 0.1g sensitivity proved more than ample to provide this data. In fact, through the careful reduction of the data we are able to discern the amount of water lost to evaporation and plant respiration (combined). Now, we are drying the total, final plant biomass (over 5Kg, having started with just 800g) to learn how much was water retained and how much was true plant structure built from carbon intake, as no nutrients were added at any point in time.

As often happens with scientific experiments, we learn something different than the anticipated outcome, and know better how to conduct the experiment the next time through.

The University of Arizona and Biosphere 2 have agreed to support a research project, hosted on-site at Biosphere 2, Oracle, Arizona, in order to monitor and record light, temperature, relative humidity, carbon dioxide, and biomass. The goal? —capture the non-linear functions inherent in plant growth and provide a ground-truth data set for the SIMOC agent-based model.

The abstract reads as follows …

Mathematical models of complex systems can provide baseline assumptions about the real-world. While Environmental Control and Life Support System (ECLSS) can be modeled as linear, static, and deterministic, deployed systems do not often behave as modeled for the full duration of a mission. Models of bioregenerative systems are considerably more complex and readily identified as probabilistic. Non-linear models are typically built upon differential equations and/or computer software applications designed specifically for simulation of particular real-world systems.

An agent-based model (ABM) employs the actions and interactions of individual and collective, autonomous agents such that their behavior, when allowed to unfold over a specified time, may exhibit non-linear, dynamic, and probabilistic behavior. Used extensively in finance, biology, ecology, and social sciences ABMs are a proven alternative to more traditional systems of modeling.

SIMOC (a scalable, interactive model of an off-world community) is a Python-based ABM developed as an Interplanetary Initiative pilot project at Arizona State University. In collaboration with the Biosphere 2, SIMOC is employed to model a semi-closed BLSS built upon the NASA funded Prototype Lunar Greenhouse.

SIMOC’s web-based agent library editor enables rapid design of new agents to match real-world systems. The configuration wizard and interactive dashboard provides a graphical interface with ABM readouts and a full command-line, back-end data capture for analytical and machine learning post processing.

This publication sees the results of the first application of this novel approach to modeling a real-world
BLSS in which continuous temperature, relative humidity, luminosity, and carbon dioxide data are
collected for the full duration of the experiment and then compared to the output of the ABM that has
generated the same four parameters for the same duration.

SIMOC Phase II development was completed for the ASU Interplanetary Initiative December 6, 2018.

Over the holiday break we provided live demonstrations to National Geographic Society and Biosphere 2, and moved into code clean-up in preparation for our continued work in 2019. While it was our intent to bring SIMOC to a public venue in January of this year, we are instead moving into Phase III in which we will bring the SIMOC agent-based model to a higher fidelity, integrating various plant growth patterns with day/night cycles, revising the means by which we manage power production, storage, and consumption, and at the same time, improving the Configuration Wizard guides and Dashboard panels.

Stay tuned!