Selecting the Optimal Spectral Band for Humid‐Tropical Archaeological Prospection in Bujang Valley: A Band‐Sum Normalization of SPOT and UAV Imagery at Sungai Batu

ABSTRACT Archaeological remnants rarely exhibit distinctive structural or spectral characteristics, which makes their detection via remote sensing in tropical environments challenging due to significant fluctuations in soil composition, moisture levels, vegetation and land use. This study analyses multi‐temporal SPOT‐5/7 satellite data (2006–2017) and DJI P4‐RTK multispectral UAV imagery (2024) to (i) characterize the spectral behaviour of established archaeological proxies at the Sungai Batu (Batu River) complex in Bujang Valley, Kedah, Malaysia, and (ii) identify an optimal spectral band (OSB) for archaeological prospection in humid tropical conditions. Twenty‐five proxies encompassing ritual structures ( Candi ), ancient jetties, Syahbandar (administrative) locations and iron‐smelting remnants ( Tuyères ) were identified from excavation records and plotted as regions of interest (ROIs) on pre‐excavation SPOT‐5 imagery (2006). We derived top‐of‐atmosphere reflectance values for each band, calculated mean–variance statistics and modelled proxy reflectance relationships using linear regressions. Additionally, we propose a straightforward band‐sum modification (per‐proxy linear model) to mitigate spectral heterogeneity caused by non‐target background effects. Among the 25 proxies, the near‐infrared (NIR) had the highest raw reflectance; nevertheless, the Green band (≈0.50–0.59 μm) consistently yielded the most robust proxy‐reflectance correlations ( R > 0.95 post‐adjustment), outperforming the Blue, Red and NIR bands. The most significant improvements following the adjustment were observed in the Syahbandar and iron‐smelting proxies ( R 2 reaching 0.98–0.99), whereas the Candi and ancient jetties proxies maintained stable, monotonic reflectance curves consistent with moist soil vegetation backgrounds. The research attributes the Green‐band's advantage to vegetation structure, soil moisture and the surface contrast derived from brick materials within the context of monsoon tropical phenology. The findings endorse the Green band as the most practical operational spectral band for preliminary assessment in Sungai Batu and similar low‐relief, forested tropical locations when only three to four multispectral bands are accessible. The work ultimately suggests implications for conducting low‐cost surveys and advocates for future testing with hyperspectral sensors and machine learning classifiers.