Regional parameterization of canopy onset models using MODIS and flowering onset data

Canopy phenological onset (CPO) is a phenomenon of start of vegetation growing season and is an important indicator for monitoring the response of vegetation to climate change. In this study, refined detection methods of CPO date were developed for three different vegetation products from the Moderate Resolution and Imaging Spectroradiometer (MODIS). The methods adopt a parameterization scheme supported by a regional observation network of cherry-blossom flowering date. Two alternative onset algorithms (i.e. chill and thermal models) were applied to develop region-scale CPO models using the MODIS-derived CPO date. The satellite-detected MODIS CPO dates were linearly correlated with the field CPO dates (r > 0.9), while the model-estimated CPO dates were generally later than the field CPO dates (mean errors, ME from −1.9 to +6.9 days). Though the thermal model algorithm predicted the CPO date slightly better (ME of +2.5 days) than the chill days model algorithm (+4.3 days), the difference was not statistically significant. In the CPO detection, three kinds of MODIS vegetation indices showed similar detection errors of −1.0, +1.7, and +0.5 days for MODIS 1 km LAI and NDVI, and 250 m NDVI, respectively. In the CPO prediction, however, the thermal and chill models parameterized with MODIS LAI showed significantly better accuracies (ME of −1.9 and +0.5 days) than those parameterized with MODIS NDVI datasets (ME from +4.0 to +6.9). These results show that the combining datasets of field flowering onset data and MODIS vegetation products provide a useful tool to improve the satellite-based detection of CPO date with less uncertainty and to parameterize region-scale CPO models where the field CPO data is less available.

► Two regional canopy phenological onset (CPO) models were developed. ► MODIS-derived CPO data were utilized for model parameterizations. ► The MODIS CPO detection algorithm was parameterized with flowering onset data. ► Combined use of MODIS and flowering data enabled regional CPO model parameterization.