Towards a global high-resolution inundation map derived from remote sensing imagery: African continent application
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Wetlands are recognized as valuable landscapes for their contribution to biodiversity, ecosystem services and population livelihoods. However, current global wetland inventories do not spatially represent wetland extent at a spatial and temporal resolution appropriate for conservation and management purposes. Among the best existing global inventories, the Global Lakes & Wetlands Database (GLWD; Lehner & Döll, 2004) is a static database assembled from various existing data sources that unfortunately suffers from the inconsistency among its data sources. Another, the Global Surface Water Extent Dataset (GSWED; Prigent et al. 2007; Papa et al. 2010) produced from a multi-satellite method is capable of monthly measurements but possesses a coarse spatial resolution incapable of discriminating distinct surface water bodies. Faced with the limitations of current global inventories, a new methodological approach is required to provide the improved wetland inventory needed by the research and conservation communities.This thesis investigates a methodology capable of producing a high-resolution (~ 500 m) surface water extent map by spatially downscaling the coarse resolution (~27 km) inundated area estimates of GSWED. The methodology inspired by Bwangoy et al. (2010) has a pragmatic and straight-forward design to ensure and ease its global application. The work of this thesis consists of an initial implementation and validation of the methodology across the African continent. The downscaling approach relies on the topographic and hydrographic information from the globally available HydroSHEDS data (Lehner et al., 2008) to distribute inundated area at the finer resolution to the most topographically inundation prone areas. Thirteen hydro-topographic variables were computed from HydroSHEDS and then consolidated into a single inundation probability map with the use of decision tree learners. The decision trees were trained on regional inundation maps and subsequently employed to generate a topographic probability of inundation map at high-resolution for the entire continent. The probability map is turned into an inundated/non-inundated map by splitting the probability distribution into two (inundated/non-inundated) with a defined threshold value. A threshold value is chosen for each GSWED cell to produce an inundation map replicating the inundated area estimates of GSWED within the cell at the finer resolution. To represent the maximum wetland extent at different timescales, two sets of inundated areas estimates were downscaled as high-resolution inundation maps with this MWT downscaling procedure: 1) the mean annual maximum (MAMax) estimates were calculated for each cell from the monthly estimates of GSWED between 1993 and 2004; 2) the fusion maximum (MaxFusion) was generated from a fusion of the time-series maximum (TSMax) also calculated from GSWED, with the wetland area from GLWD. The MaxFusion estimates were produced to correct some data gaps of GSWED, as well as to offer a more complete and reliable maximum wetland extent map. The MAMax and MaxFusion estimates respectively totalled 1339 and 2779 thousand km2 of wetland area across the continent; higher than most previous estimates for Africa.Validation of the spatial distribution of inundation at the finer resolution exhibited high levels of agreement against reference regional maps (Overall Accuracy ~ 92%; KIA ~ 80%). Over selected wetland study sites, comparisons of the MaxFusion downscaled map with the global land cover GLC2000 (Mayaux et al. 2004) and wetland database GLWD indicated that the downscaled map possessed slightly lower but more consistent agreement with GLC2000 than GLWD did. Regardless, the level accuracy of the tested methodology is considered satisfactory to pursue production of a first version global inundation map. Possible follow-up applications making use of the downscaled inundation maps such as a global hydro-geomorphic wetland classification.
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.
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 | 0.000 | 0.000 |
| 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 | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 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 it