The roster came from Uzbekistan’s implementation of a settlement-review process launched by Presidential Resolution PP-120 in July 2005 and consolidated in Cabinet Resolution No. 68 in March 2009. The process allowed rural settlements with at least 2,000 residents and specified functional characteristics to be proposed as urban-type settlements. The later resolution recorded decisions made by regional councils and the Jokargy Kenes of Karakalpakstan, directed statistical authorities to adjust urban and rural indicators, and attached planning obligations. It did not establish one decision date for every place, and the public record lacks the local decision dates and the set of eligible settlements that were not proposed.

Researchers combined the legal list with public gazetteers, the harmonized nighttime-light series, WorldPop population grids, the World Settlement Footprint, or WSF, and GHSL built-up surfaces. Fixed buffers around settlement centers were used to prevent later legal boundary changes from being counted as physical growth. The 1-km GHS-SMOD cell at each coordinate supplied the primary point classification, while alternative 1-, 2- and 5-km buffers and population-free built-up thresholds tested whether the result depended on spatial scale or classification rules. This design is descriptive rather than a natural experiment: it compares observed trajectories, but does not identify a causal effect of reclassification.

Geocoding covered 750 of the 965 listed settlements, or 77.7% of the roster. Coverage was uneven: it ranged from 60.7% to 100% across 12 regions but was only 23.5% in Khorezm. The observed 78.0% urban-domain share therefore has a district-clustered 95% interval of 73.3% to 82.3%, while a full-roster lower bound falls to 60.6% if every ungeocoded settlement was rural in 2005. A scenario that assigned ungeocoded places their district’s observed share produced 79.5%, but that assumption cannot be tested with the available data. The authors retained both the observed estimate and the conservative bound because unresolved coordinates could select either more urban or more rural settlements.

The three trajectories were also geographically clustered. In Fergana Region, 96.3% of the 135 geocoded reclassified settlements were already in the urban domain in 2005, compared with 39.3% in Jizzakh and 36.4% in Karakalpakstan. Even within the 585 baseline-urban points, physical form varied substantially: 263 were suburban or peri-urban cells, 75 were semi-dense urban clusters, 154 were dense urban clusters and 93 were urban centres. Population and built-area differences reinforced that heterogeneity. Median 2005 modeled population within 2 km was 6,141 in baseline-urban settlements, 3,286 among points that later crossed the urban threshold and 1,885 among points that remained rural.

Spatial scale mattered as much as the underlying geography. One-quarter-rule readings within 1- and 2-km buffers agreed with the point classification for 97.1% and 93.9% of settlements, respectively, although the high agreement at 1 km was partly mechanical because the buffer contained only a few cells. Agreement fell to 70.1% under a 2-km majority rule and 44.9% under a 5-km majority rule, where surrounding landscape increasingly dominated the classification. Population-free WSF thresholds also weakened the majority finding. At least 5% of the 2-km buffer was built up at 91.5% of geocoded settlements, but the share fell to 83.9% at 10% coverage, 61.1% at 20% and 38.3% at the 10th percentile of established urban-form benchmarks.

The contrasts with nearby rural settlements initially favored the reclassified group on modeled population, but that advantage disappeared after accounting for baseline differences. Reclassified sites had 0.031 more log points of WorldPop growth in the unstratified comparison, with a 95% bootstrap interval of 0.020 to 0.046. Yet the initial pairs were markedly imbalanced: 78.0% of reclassified sites, compared with 35.8% of rural benchmarks, were already in the GHS-SMOD urban domain. When comparisons were restricted to pairs within the same baseline domain, the WorldPop difference was 0.019 log points in the rural stratum, with an interval spanning zero, and −0.001 in the urban stratum. The unadjusted population advantage was therefore largely associated with different starting positions.

Nighttime-light growth remained lower among reclassified settlements after baseline adjustment, while differences in the two built-up measures were generally close to zero. In the covariate-matched subsample, 387 pairs with complete light and GHSL endpoints had standardized baseline differences below 0.10. Reclassified settlements showed 0.042 fewer log points of nighttime-light growth than their matched rural benchmarks, with a 95% interval of −0.074 to −0.010. Their WorldPop difference was −0.013, their WSF built-area difference was −0.002 and their GHSL built-area difference was 0.007; all three intervals included zero. A population-free stratification produced the same broad result: among pairs whose members both had at least 20% built-up coverage in 2005, no outcome showed a consistent difference.

Clustered resampling supported the matched nighttime-light result without changing the main conclusion. Drawing whole districts rather than settlements widened some intervals, but the matched light difference remained negative: −0.042 log points, with a district-clustered 95% interval of −0.082 to −0.003. The result was also stable across buffer sizes, geocoding tiers, match-distance limits and leave-one-region-out checks. Yet these comparisons remain descriptive. The rural group is a geographic benchmark, not a counterfactual population of settlements that were eligible for reclassification but were not proposed. Matching improves comparability on observed 2005 characteristics; it cannot erase the legal selection process itself.

Representative cases illustrate why one label conceals divergent histories. Yangikadam remained a suburban-class point from 2005 to 2020 as modeled population within 2 km rose from 6,141 to 9,799, while nighttime lights remained nearly flat. Jamatak crossed from a rural cluster to a suburban cell between 2005 and 2020, with modeled population increasing from 3,176 to 5,371 but lights declining from 153 to 138 by 2024. Balandchayla remained in a rural-class cell at both dates despite modeled population more than doubling, from 2,257 to 4,971. These cases are compatible with confirmation of existing concentrations, transitional upgrading or persistent separation between legal status and physical form, but the available evidence cannot determine which mechanism operated in each case.

The implication is methodological as much as geographic. Administrative status is necessary for questions about jurisdiction, planning obligations and legal recognition; it is not a complete proxy for compact settlement form or economic structural change. Population grids and built-up imagery can characterize morphology, but nighttime lights remain an indirect activity proxy, and sectoral employment or production data are required to establish changes in livelihoods or the movement of labor out of agriculture. For international urban monitoring, the study’s authors recommend that researchers identify the administrative classification rule and its vintage, report the spatial scale and thresholds used for morphological measures, and avoid presenting a change in legally defined urban population as physical growth. For Uzbekistan, the 965-place reform instead offers a concrete demonstration that administrative urbanization can pass through a highly varied settlement landscape.

Subject of Research: Administrative reclassification and multidimensional measures of settlement urbanization in Uzbekistan

Article Title: Administrative reclassification and the measurement of urbanization: multi-source evidence from 965 settlements in Uzbekistan

Article References: Zhang, L., Kozik, K., He, Y., & Huang, Y. (2026). Administrative reclassification and the measurement of urbanization: multi-source evidence from 965 settlements in Uzbekistan. Frontiers in Sustainable Cities, 8, Article 1971818. https://doi.org/10.3389/frsc.2026.1971818

Image Credits: AI Generated

DOI: 10.3389/frsc.2026.1971818

Keywords: Uzbekistan, urbanization, administrative reclassification, settlement morphology, remote sensing, nighttime lights, WorldPop, GHS-SMOD, built-up area, urban measurement, Administrative, reclassification