Abstract

The calendarization of Growing Degree Days (GDD) is a key tool for adjusting the sowing date and achieving different growth characteristics of crops. It is also a climatic factor that can be used to calibrate and predict crop production based on land surface phenology (LSP) and remote sensing data. However, most studies use only fixed calendars and do not consider the regional sowing date or surface thermal parameters for GDD accumulation, which results in high variability of GDD across the winter wheat regions. In this study, we analyzed the effect of using different calendars and thermal parameters on GDD accumulation for winter wheat. We used air temperature at 2 m (Ta) and land surface temperature (Ts) from ERA5-Land reanalysis data and WorldCereal calendars . We compared two types of calendars: fixed calendars based on January 1st (Franch et al., 2015) as the biofix date, and regional calendars based on sowing dates at administrative level (Franch et al., 2022). We filtered the data using the USDA crop data layer (CDL) and considered only pixels with at least 5% of winter wheat signal at county level. We calculated t he mean, the standard deviation and coefficient of variation of GDD at the end of season for each combination of calendar and thermal parameter. We found that using regional calendars instead of fixed ones reduced the variability of GDD accumulation in the United States. The coefficient of variation and standard deviation decreased from 26.36%, 655.10 °C with Ta and fixed calendar to 18.80%, 473.30°C with Ta and regional calendar and 18.10%, 463.73°C with Ts and WorldCereal calendar. The Ts data also showed a slightly higher mean GDD than the Ta data (2562.43°C vs 2517.85°C). For example, in Oklahoma, which had the highest winter wheat coverage (52% on average), the GDD accumulation was more stable with Ts and WorldCereal calendar (2546°C) than with Ta and fixed calendar (2968°C).We concluded that using regional calendar is essential for normalizing the GDD for winter wheat, and land surface temperature can slightly improve the normalization. This information improves the input parameter for calibrating the yield model by providing more accurate and consistent estimates of crop phenology and thermal conditions. It also reduces the uncertainties of LSP parameterization on crops at regional scales.