Reverse Engineering the National Land Cover Database: A Machine Learning Algorithm for Replicating Land Cover Data in the Spatial and Temporal Domains
Bibliographic record
Abstract
Land cover datasets are generally produced from satellite imagery using state-of-the-art model-based classification methods while integrating large amounts of ancillary data to help improve accuracy levels. The knowledge base encapsulated in this process is a resource that could be used to produce new data of similar quality, more efficiently. Specifically, the question addressed in this dissertation is whether this richness of information could potentially be extracted from the underlying remote sensing imagery to then classify an image for a different geographic extent or a different point in time. This research developed a machine learning framework to replicate the U.S. National Land Cover Database (NLCD) from Landsat 5 TM imagery in the spatial and temporal domains. Information characterizing individual land cover classes was extracted using a maximum entropy classifier on a Landsat image to create a generalizable model for land cover data replication. This framework was then demonstrated for spatial extrapolation and temporal extension of the NLCD by applying the model to Landsat imagery for a different geographic extent and for a different point in time. The experimental setup of this dissertation used three study areas in the U.S. featuring different landscape compositions to test the stability and generalizability of this framework. Results for the spatial and temporal replication of the NLCD showed that the objective of reproducing similar levels of overall and within class accuracies could be met and demonstrated that the knowledge base encapsulated in the NLCD can effectively be extracted for replication. The algorithm proved to be generalizable to the range of landscapes represented by the three study sites and showed stability in both spatial and temporal replication. This dissertation demonstrates how such a framework could potentially extend the NLCD into Canada or Mexico, for example, and how it could be implemented to produce annual land cover data. Effective replication of the NLCD provides a valuable case study since similar land cover datasets exist in many countries and an automated method for spatial extrapolation or temporal extension of such data would benefit the scientific community and advance similar areas of research.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.011 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".