Observational Grammar
A constitution for evidence derived from sensor physics, independent of external incentives
Observational Grammar (OG) is the idea that sensors — satellites, radar, spectrometers, thermal cameras — can form a language of evidence about reality that operates independently of human bias, market incentives, or bureaucratic approval chains. Just as grammar gives structure to language, OG gives structure to what instruments can claim about the physical world. It is M33's foundational concept: build systems that let reality set the table, then let markets and decisions work within those constraints, rather than the other way around.
Why It Matters
Every day, thousands of satellites observe Earth. They measure electromagnetic reflectance, radar backscatter, thermal emission, gravitational anomalies, and atmospheric chemistry. Each measurement is a claim about reality — a statement made by physics, not by a person with an agenda.
But between the sensor and the decision-maker, something happens. Data gets filtered through procurement pipelines, formatted for specific use cases, interpreted through institutional lenses, and delivered in reports that serve the needs of whoever commissioned them. The physics remains, but the direct line from observation to understanding gets tangled.
A coastal development gets approved because the regulatory studies — designed to answer a specific bureaucratic question — concluded it was compliant. Meanwhile, the satellites recorded salinity trends, erosion rates, storm frequency patterns, and groundwater changes that told a different story. The instruments saw the failure coming. The decision pipeline didn't ask them the right question.
Observational Grammar is the framework for closing that gap. It asks: what if we started with what the sensors actually see, formed a composite understanding of reality from multiple independent observations, and then let human needs, markets, and policy operate within those physical constraints?
The Grammar
A natural language grammar has rules: syntax determines how words combine into meaning, semantics assigns meaning to those combinations, and pragmatics governs how context shapes interpretation. Observational Grammar works analogously.
Syntax — The Rules of Measurement
Every sensor operates within physical constraints that determine what it can and cannot claim. A Sentinel-2 multispectral imager observes reflected sunlight in 13 spectral bands at 10-60 meter resolution. A Sentinel-1 SAR instrument emits C-band radar and measures the backscattered return, seeing through clouds but sensitive to surface roughness and moisture. A thermal sensor measures emitted heat, not reflected light.
Each instrument has a grammar of what constitutes a valid statement. A SAR sensor can say "this area has high moisture content" with confidence. It cannot say "this area is flooded" without additional context — that requires combining its observation with elevation data, historical baselines, and potentially optical confirmation. The syntax enforces honesty about what each observation can claim on its own.
Semantics — Composite Meaning
Individual sensor readings are claims. Composite sensor readings are evidence.
When a SAR instrument detects high backscatter change, a multispectral sensor shows water absorption signatures, and a terrain model confirms the area is in a floodplain, the composite observation has a confidence level that no single sensor could achieve alone. The semantic layer of OG is where independent observations are combined — not averaged or blended, but cross-validated — to form statements about reality with quantified uncertainty.
This is where Fabric operates. The data harmonization engine transforms heterogeneous sensor data into analysis-ready formats that can be composed. It doesn't just reformat data — it preserves the provenance chain so that every composite observation can trace back to the individual claims that support it.
Pragmatics — Context and Confidence
No observation is context-free. A vegetation index reading in July means something different in the Northern and Southern hemispheres. A thermal anomaly in an urban area might be a building fire or a heat island effect. The pragmatic layer of OG encodes what we know about context: geography, seasonality, historical baselines, known land cover, and the specific question being asked.
Critically, OG's pragmatic layer includes an explicit model of its own uncertainty. When confidence is low — because of cloud cover gaps, sensor degradation, or insufficient corroborating observations — the grammar says so. This is built into the architecture, not left as an afterthought. A system that cannot say "I don't know" is not an observational grammar; it's a marketing tool.
Principles
01 — Reality precedes market. Phenomena exist independent of use cases. OG begins with what sensors observe, not what customers want to hear. The earth and stars know nothing of boardrooms or go-to-market strategies.
02 — Claims require physics. Every statement in OG must trace back to a measurable signal grounded in physical law. If a sensor cannot measure it, OG cannot claim it. Inference is allowed but must be marked as inference.
03 — Composite evidence over single readings. Confidence increases with independent corroboration. A single sensor makes a claim. Multiple independent sensors make evidence. The distance from observation to truth shortens with each corroborating source.
04 — Provenance is non-negotiable. Every observation carries its history: which sensor, which calibration, which processing chain, which corrections. Without provenance, data is assertion. With provenance, data is evidence.
05 — Humility is structural. OG is designed to be wrong and know it. Sensors drift, calibrations expire, patterns evolve. The grammar includes feedback mechanisms that let it learn from its own failures and admit when quality assurance is low.
06 — The grammar expands. As new sensors launch, new spectral bands open, and new processing methods emerge, OG grows. It is a living system, not a fixed specification. It is in a constant loop of expansion and improvement.
Beyond Linear Pipelines
Traditional data pipelines in earth observation are linear: acquire → process → analyze → report → decide. Each step is a gate, and each gate introduces constraints that narrow what can be said.
OG replaces this with what might be called a geodesic node language — a topology where observations connect to each other laterally, not just sequentially. A SAR observation of soil moisture connects directly to an optical observation of vegetation stress, which connects to a weather station's rainfall record, which connects to a hydrological model. The connections are not pre-programmed for a specific use case. They emerge from the physical relationships between phenomena.
In this topology, the distance from observation to understanding is determined by latency, coverage, and resolution — not by how many bureaucratic gates stand between the sensor and the decision-maker. Facts become a result of the network's physics, not its politics.
This is ambitious, and it is explicitly aspirational. OG does not claim to deliver this today. But it provides the architectural direction: build systems where analysis-ready data from any sensor can be composed with any other, where provenance tracks every connection, and where the grammar itself evolves as understanding deepens.