Land Cover Change + Development Forecasting Charlotte MSA · Spatial modeling workflow
This project builds a spatial forecasting workflow to understand where development is likely to occur in the Charlotte metropolitan region, based on land cover change patterns and neighborhood context. The goal is planning-oriented: translate remote sensing + census + infrastructure signals into an interpretable model and map of likely near-term change.
I set up an end-to-end pipeline that converts “messy” regional datasets into a consistent spatial unit (a grid), then builds and evaluates a model to estimate where development pressure concentrates. The deliverable is not just a model output, but a planning-readable explanation of what variables matter, where the model performs well, and what it implies for regional growth.
The workflow standardizes the region into a fishnet grid and assembles predictors at that grid scale: land cover composition, distance to highways, and neighborhood effects using a spatial lag of development. Population was also translated to the grid using areal-weighted interpolation so that demographic context could be compared consistently across space.
A key design choice here is interpretability: each feature maps back to a planning logic (access, adjacency, existing intensity, and developable land). That makes the final prediction map easier to explain and critique, instead of feeling like a black box.
Transportation access is a major part of the story in Charlotte: highways structure regional connectivity, while transit expansion is debated as growth accelerates. In the analysis, highway vectors were brought into the same spatial frame and used to generate accessibility features (e.g., distance signals) that can be tested against observed change patterns.
The model outputs a spatial pattern of expected development that concentrates strongly in Mecklenburg County, aligning with where major employment and infrastructure anchors sit. This creates a practical planning takeaway: some parts of the region behave like “growth magnets,” while other areas show weaker signals unless land-use transition happens first.:contentReference[oaicite:8]{index=8}
The workflow also surfaced an important land-cover insight: a large share of observed change occurred in “other undeveloped” categories, which may hide environmentally sensitive landscapes (including wetlands). That matters because it changes how “developable” land should be interpreted and where extra safeguards may be needed.
- Built a grid-based dataset that aligns census, infrastructure, and land cover into one spatial unit.
- Implemented areal-weighted interpolation to translate tract population into fishnet estimates.
- Engineered spatial predictors (distance/access + neighborhood lag effects) and explored their relationships.
- Generated prediction maps and evaluation visuals to communicate performance and spatial patterning.
- Converted technical outputs into planning recommendations grounded in the model’s limits and strengths.
Based on the forecasting pattern and the regional context, the write-up argues for a phased approach to transit investment (starting with high-signal connections) and for TOD-supportive financing tools (e.g., station-area value capture). It also flags wetlands risk as a likely “hidden cost” of development change, recommending tighter local protection where state protections have weakened.
