Résumé
Studying modern pollen rain in Middle Asia is crucial for understanding past climate and vegetation changes. This study presents the first dataset of pollen surface samples and vegetation plots from Tajikistan and Uzbekistan, known as the Tajikistan and Uzbekistan Surface Data Base (TUSDB), to enhance our understanding of past climate and vegetation changes in Arid Central Asia (ACA). Multivariate analysis methods, including TWINSPAN and CONISS, are used to review the primary vegetation types in Central Asia and assess pollen distribution across these types. Linear relationships, Davis indices, and R-values are applied to evaluate vegetation representativeness based on pollen abundances. Redundancy Analysis is used to compare pollen and climate parameters, and the reliability of TUSDB for local climate reconstructions is tested using transfer functions and machine learning approaches.The TWINSPAN analysis identifies vegetation types consistent with the Uzbek classification, such as desert, steppe, and cryophilous open woodlands. Nine vegetation subtypes are revealed by the pollen samples, with key contributions from trees like Juniperus spp. and Juglans regia, and non-arboreal plants like Cyperaceae and Poaceae. The study also highlights biases in pollen representation, with some tree taxa overrepresented and certain vegetation-important taxa underrepresented, which can be corrected using pollen R-values.Finally, functional trait aggregation for past pollen sequences shows similarities between modern pollen and vegetation plots. Climate reconstructions validated through transfer functions highlight the complementary role of local and global calibration datasets. These findings confirm the reliability of pollen signals for inferring Holocene climate and vegetation trends, enabling future Uzbek pollen-based reconstructions.