Gains:
- Understanding that spatial measurement (slope, orientation, area) is the job of GIS and the field, and that measurements will not be requested from artificial intelligence.
- Ability to synthesize measured data with field observations and produce opportunity-constraint lists and usage zoning recommendations with artificial intelligence
- Ability to view artificial intelligence synthesis as a hypothesis and counter-read it in the field and separate materials that require engineering
Good landscape design starts with "listening" to the plot. Where does water flow from, where does the sun shine, where does the wind blow, which paths do people already use? Site analysis is the systematic examination of the physical, ecological and social characteristics of a site - it is the basis of reality on which the design will be based. GIS (Geographic Information System / GIS) is software that maps and analyzes spatial data (slope, elevation, soil, water, structure) in layers. In this unit, you will learn how to use AI as an interpretation and synthesis partner in domain analysis; but you will see why measurement and positional accuracy remain in GIS and in the field.
Layers of field analysis
Typical layers when reading a plot are:
- Topography: elevation and slope (degree of slope of the surface — determines water flow, access, feasibility).
- Aspect: the direction the slope faces; Determines the effect of insolation and wind.
- Hydrology: where water comes from and where it flows, flood and accumulation areas.
- Soil and geology: soil type, drainage, bearing capacity.
- Vegetation and ecology: existing trees, natural life, tissues to be protected.
- Human layer: current use, access, neighbourhood, views.
GIS produces these layers sparingly; AI helps relate and interpret these layers and synthesize field observation notes. Critical rule: AI cannot measure slope or produce coordinates on a map; It helps make sense of the measured data you provide and recommend priorities.
Hint: Ask the AI "what is the slope of that land?" Don't ask — he can't know that and might make it up. Instead, give the slope data you received from GIS and ask "according to this slope distribution, which areas are suitable for walking, which are suitable for sitting, and which are suitable for water retention?" Ask for comments.
Step by step: from data to interpretation
Step 1 — Collect measured data. Extract data such as slope, orientation, water flow direction from GIS or field survey. This step is human/vehicle work.
Step 2 — Note the field observation. Write down what you see on the field (wet corner, windy ridge, shaded area, worn path).
Step 3 — Have the AI synthesize. Give measured data + observation notes and ask for a comment in the form of an "opportunity and constraint map".
Step 4 — Counter-read. Compare the AI's interpretation with field reality; Correct any areas that do not fit.
Step 5 — Tie it to the design decision. Translate the opportunities/constraints resulting from the analysis into concrete design decisions (where and what).
three mini cases
Case 1 — Slope interpretation. In a hillside park project, GIS slope data varied between 5%–22%. YZ combined it with observation notes and drafted a proposal saying "The areas above 15% are terraced, below 8% are free meadows, and the rest is walking". The architect verified this on site and saved 60% time; He drew the final terrace line himself.
Case 2 — Waterlogging. There was recurring waterlogging in a schoolyard. Field notes + slope + soil data were given to YZ; YZ pointed out that the low-sloping clay corner is a natural accumulation point and can be turned into a rain garden. The engineer also made drainage calculations.
Case 3 — Fabrication caught. An intern asked the AI directly "what is the average slope of this parcel?" he asked; The AI gave a confident number. The senior architect realized this was unfounded — the AI had not seen the parcel. The measure was taken from CBS; the actual value was different. Lesson: get the positional measure from the vehicle, not the AI.
Four copyable templates
1) Opportunity-constraint synthesis:
Your role: landscape analyst. I will give you measured data and field observations. Data: slope distribution [...], orientation [...], water flow direction [...], soil [...]. Observation: [wet corner, windy ridge, shadow, existing path]. Synthesize these and create an OPPORTUNITY and CONSTRAINT list. Size fitting; Just interpret the data I give you.
2) Usage zoning recommendation:
Based on the slope and orientation data I provide, suggest dividing the plot into function zones: walking, sitting/gathering, water retention, dense vegetation, play. Write down what data you base each recommendation on. State that this is a draft and will be verified in the field.
3) Field trip checklist:
Produce a checklist of items I would need to observe and note on a site analysis trip for the [4 ha streamside park] project: topography, water, soil, vegetation, sun/shade, wind, human use, access and neighborhood.
4) Analysis-design bridge:
Translate this opportunity-constraint list into concrete design decisions: “how do I exploit this” for each opportunity, “how do I overcome/manage this” for each constraint. Also mark the items requiring engineering (drainage, retaining).
Weak prompt / Strong prompt
Weak prompt:
Analyze this plot.
The AI does not see the plot; If there is no data, he either makes general statements or makes up measurements.
Powerful prompt:
Your role: domain analyst. Measured data (from GIS): slope 4-18%, north-facing slope, water flows to the southeast, clay loam soil, median drainage. Site observation: the northern corner is constantly shaded and moist, the southern edge is open to wind, the existing plane tree row will be preserved, there is a built-in pedestrian path on the eastern edge. A list of opportunities and constraints is derived from this data; Write down the data on which you base each item. Do not make up measurements.
Given measured data + observation, AI produces a true synthesis and the "measurement fitting" constraint cuts off the hallucination.
Data source and role table
layer
Where is the data from?
Role of AI
man's role
slope/elevation
GIS, measurement
Comment
Measurement, confirmation
orientation
GIS
attribution
verification
hydrology
GIS + field
Pattern suggestion
Engineer account
soil
drilling/laboratory
making sense
test, decision
current plant
field inventory
synthesis
detection, protection
human use
field observation
Summary
observation, comment
Common mistakes
- Requesting positional measurements from the AI. Slope, area, distance are GIS/measurement work; AI can make it up.
- Skipping the field trip. No data can replace standing in the field and seeing.
- Looking at a single layer. Opportunities arise at the intersection of layers (slope + water + shadow together).
- Not comparing the interpretation with the field. AI synthesis is a hypothesis; verified in the field.
- Compressing engineering issues into comments. Drainage capacity and retaining statics require separate calculations.
In summary
In field analysis, measurement and positional accuracy are the job of GIS and the field; AI is a powerful partner in synthesizing this measured data with field observations and producing opportunity-constraint maps and zoning recommendations. Execute the process through measured data collection, field observation, AI synthesis, field counter-reading, and linking to design. Never ask the AI for measurements; Give it data and make sense of it. Synthesis is always a hypothesis to be verified in the field.
Application task
Describe slope, orientation, and water flow for a plot (either with actual GIS data or reasonable assumptions) and write a field observation note. Create an analysis with the "Opportunity-constraint synthesis" template, then turn it into at least five concrete design decisions with the "Analysis-design bridge" template. Tick the items that require engineering.
checklist
- [ ] I took the measured data from GIS/measurement, I did not adapt it to AI.
- [ ] I put my field observation note in writing.
- [ ] I built the AI synthesis with the intersection of multiple layers.
- [ ] I counter-read the synthesis with the field reality.
- [ ] I marked the items requiring engineering separately and directed them to the expert.