Land surface phenology dynamics using a 41-year Landsat NDVI time series in the semi-arid agriculture-dominated Altinekin Basin, Türkiye

Land surface phenology (LSP) provides a valuable indicator for monitoring natural and human-induced vegetation dynamics in terrestrial ecosystems. Remote sensing has become a primary tool to capture LSP metrics. This study investigates long-term LSP dynamics in the semi-arid agriculture-dominated Altinekin Basin, Türkiye. Gap-filled and smoothed 8-day NDVI composite time series were generated from Landsat 5/7/8/9 satellite imagery for the 1985-2025 period using cubic spline interpolation. Annual LSP metrics were derived using the dynamic thresholds method, with a 30% seasonal amplitude threshold and a 20% peak prominence threshold. The non-parametric trend analysis showed that start of season (SOS) was delayed by 2.0 days/year in arable land, while advanced by 0.9 days/year in grassland. However, the arable land-well end of season (EOS) was postponed by 1.0 days/year in contrast to arable land EOS advanced by 1.1 days/year. This may suggest different phenological characteristics of two cropping practices. The growing season length (LOS) shortened by about 68 and 88 days from 1985 to 2025 for grassland and arable land, respectively. The peak NDVI trend in arable land was 0.009 NDVI/year, while 81% of irrigated zones were significant (p