Comparing in situ spring phenology and satellite-derived start of season at rural and urban sites in Ireland

The timing of spring phenology marks the start of the growing season and the beginning of the carbon-uptake period. In order to accurately calculate carbon budgets across different vegetation types, reliable start-of-season data over relatively large areas is necessary. However, in-situ phenological observation networks are generally restricted in their geographical extent and limited to a small number of species, typically dominant trees. Satellite remote sensing can provide spatially extensive phenological information but requires validation with in-situ observations. Here, we examined trends (1968–2016) in the timing of leaf-unfolding (LU) as well as growing-season-length (GSL), derived from in-situ observations, of a suite of deciduous trees, at three rural sites and one urban site in Ireland, all of which are within the International Phenological Gardens network. We compared in-situ trends with satellite derived phenometrics, including start of greening increase and GSL, based on two-band Enhanced Vegetation Index (EVI2) from a combined AVHRR and MODIS time-series over the time period 1982–2016. The aim of this study was to assess the effectiveness of satellite remote sensing in capturing LU and GSL as determined by in-situ observations from sites dominated by varying land cover types. The timing of LU (DOY 111) at the urban site was significantly (p < 0.000) later than all rural sites, apart from one (1982–2016). The significantly (p < 0.000) longer (2–5 weeks) GSL (217 days), at the urban site, was thus driven by delayed leaf fall (LF). Satellite derived Start of Season (SOS) was consistently earlier than LU across all sites (RMSE 25–52d; MBE −5 to −50 d) whereas EVI2-derived GSL was consistently longer than in situ data (RMSE 65–102d; MBE 45 to 96 d). Overall, the discrepancies reported here between the timing of in-situ spring phenology and satellite-derived phenometrics at these sites were possibly due to a combination of factors including differences in scale between the two methods, the small number of cloned trees being monitored, which does not appear to represent the broader landscape vegetation, the heterogeneity of the landscape and technical differences between the satellite sensors used to construct the EVI2 time-series. Addressing these issues is challenging but engaging citizen scientists to monitor local vegetation would increase the spatial coverage of the in-situ observation network while also incorporating a broader range of species some of which would undoubtedly represent native vegetation.