Meteorological parameters significantly affect cotton growth and development, and selecting an optimum sowing date can improve seed cotton yield. The experiment consisted of sowing a film-covered, drip-irrigated cotton field (Gossypium hirsutum L.) on four sowing dates from April to May and took place in 2011 and 2012 at the Agrometeorological Experimental Station of Wulanwusu, which was in an arid region of north-western China. The results showed that late sowing dates produced less aboveground dry biomass, yield and water-use efficiency (WUE) than did the normal sowing dates. The yield increased with the increases of mean diurnal temperature range (DTR) from full bloom to maturity, mean temperature and sunshine hours (SH) during the whole growing season, accumulated temperature (AT) and days from squaring to anthesis, and mean temperature during the reproductive stage. The yield decreased with the increases of AT and days from sowing to emergence. However, the main effect factors of meteorological parameters were AT from squaring to anthesis, mean temperature during the whole growing season and AT from sowing to emergence. It was significant to choose the optimal planting date to improve yield. Meanwhile, yield was also affected by leaf area index (LAI), boll number per plant and gin turnout. However, the main effect factors of yield component were boll number per plant, gin turnout and boll weight. Boll number per plant suffered from mean DTR from boll setting to maturity and SH during the whole growing season. Gin turnout was affected by mean temperature during the whole growing season and mean DTR from boll setting to maturity. The relationship between yield and boll weight was insignificant. Sowing date, year and interaction (sowing date × year) all significantly affected the yield. Thus, sowing date was an important factor affecting the yield, biomass and reproductive duration and minimized the impact of temperature and duration of the reproductive growth stage. With climate change, an earlier planting date might be an efficient method of increasing yield in the future.