Fine Characterization of Leafing Phenology to Rainfall Regime Shifts in the Brazilian Atlantic Forest by Optical and Microwave Remote Sensing

Tropical forests are crucial for the global biosphere, but their seasonal dynamics are poorly understood. Seasonally dry forests' green-up follows rainfall, while wet forests with evergreen trees show no such pattern due to sufficient water availability. Optical remote sensing methods struggle to characterize the leafing phenology in wet forests due to insensitivity to water content and cloud interference. Using radar remote sensing data, we can track the cyclical pattern of water content in plant tissues, complementing optical data and improving vegetation classifications. Over evergreen forests, a vegetation index from optical data will therefore likely lag behind microwave data due to decoupling of photosynthesis and water accumulation while seasonal forests may show a smaller lag. To test this, I extracted time series (2019-2021) of vegetation indices (EVI2 and RVI) derived from Sentinel-2 surface reflectance and Sentinel-1 microwave backscatter imagery over 1-ha sampling plots in two common vegetation types in the Atlantic Forest of Brazil: evergreen broadleaf and seasonal semi-deciduous forest in Google Earth Engine (GEE). To characterize seasonality in each forest, I modeled annual variation in canopy green-up and water accumulation as both first-order and third-order harmonic functions in GEE. I found unexpected and large asynchrony (lag) of up to four months between the seasonal green-up (based on the optical index) and water accumulation (based on the microwave index) in both seasonal semi-deciduous and evergreen broadleaf forests referenced from the start of season. Depending on the modeling method used, green-up/water accumulation lag in the evergreen broadleaf forest was actually smaller that the semi-deciduous forest by nearly a month. This suggests that introducing lag information in seasonality may not enhance the differences between forest canopy seasonality patterns already indicated by optical amplitude. Alternatively,it may reveal greater sensitivity of evergreen broadleaf forests than semi-deciduous forests to abnormal meteorological conditions such as El Niño, since I found support for significant declines in total rainfall based on IMERG rainfall estimates over the study region in the last 20 years.