Satellite-based estimation of carbon dioxide budget in tropical peatland ecosystems

Tropical peatland ecosystems are known as large carbon (C) reservoirs and affect spatial and temporal patterns in C sinks and sources at large scales in response to climate anomalies. In this study, we developed a satellite data-based model to estimate the net biosphere exchange (NBE) in Indonesia and Malaysia by accounting for fire emissions (FE), ecosystem respiration (Re), and gross primary production (GPP). All input variables originated from satellite-based datasets, e.g., the precipitation of global satellite mapping of precipitation (GSMaP), the land surface temperature (LST) of the moderate resolution imaging spectroradiometer (MODIS), the photosynthetically active radiation of MODIS, and the burned area of MODIS fire products. First, we estimated the groundwater table (GWT) by incorporating LST and precipitation into the Keetch-Byram Drought Index (KBDI). The GWT was validated using in-situ measurements, with a root mean square error (RMSE) of 24.97 cm and an r-squared (R2) of 0.61. The daily GWT variations from 2002 to 2018 were used to estimate respiration (Re) based on a relationship between the in situ GWT and flux-tower-observed Re. Fire emissions are a large direct source of CO2 from terrestrial ecosystems into the atmosphere and were estimated by using MODIS fire products and estimated biomass. The GPP was calculated based on the MODIS GPP product after parameter calibration at site scales. As a result, averages of long-term (17 years) Re, GPP, FE, and NBE from whole peatlands in the study area (6°N-11°S, 95-141°E) were 66.71, 39.15, 1.9, and 29.46 Mt C/month, respectively. We found that the NBE from tropical peatlands in the study area was greater than zero, acting as a C source. Re and FE were influenced by El Niño, and the value of the NBE was also high in the El Niño period. In future studies, the status of peatland degradation should be clarified in detail to accurately estimate the C budget by applying appropriate algorithms of Re with delineations of types of anthropogenic impacts (e.g., drainages and fires).