摘要：Increasing global demand of vegetable oils and biofuelsresults in significant oil palm expansion in southeastern Asia, predominatelyin Malaysia and Indonesia. The land conversion to oil palm plantations has posedrisks to deforestation (50 % of the oil palm was taken from forest during1990–2005; Koh and Wilcove, 2008), loss of biodiversity and greenhousegas emission over the past decades. Quantifying the consequences of oil palmexpansion requires fine-scale and frequently updated datasets of land coverdynamics. Previous studies focused on total changes for a multi-yearinterval without identifying the exact time of conversion, causinguncertainty in the timing of carbon emission estimates from land coverchange. Using Advanced Land Observing Satellite (ALOS) Phased Array typeL-band Synthetic Aperture Radar (PALSAR), ALOS-2 PALSAR-2 and ModerateResolution Imaging Spectroradiometer (MODIS) datasets, we produced an annualoil palm area dataset (AOPD) at 100 m resolution in Malaysia andIndonesia from 2001 to 2016. We first mapped the oil palm extent usingPALSAR and PALSAR-2data for 2007–2010 and 2015–2016 and then applied adisturbance and recovery algorithm (Breaks For Additive Season and Trend – BFAST) to detect land cover changetime points using MODIS data during the years without PALSAR data (2011–2014and 2001–2006). The new oil palm land cover maps are assessed to have anaccuracy of 86.61 % in the mapping step (2007–2010 and 2015–2016). Duringthe intervening years when MODIS data are used, 75.74 % of the detected changetime matched the timing of actual conversion using Google Earth andLandsat images. The AOPD revealed spatiotemporal oil palm dynamicsevery year and shows that plantations expanded from 2.59 to 6.39×106 ha andfrom 3.00 to 12.66×106 ha in Malaysia and Indonesia, respectively (i.e. a netincrease of 146.60 % and 322.46 %) between 2001 and 2016. The highertrends from our dataset are consistent with those from the nationalinventories, with limited annual average difference in Malaysia (0.2×106 ha)and Indonesia (−0.17×106 ha). We highlight the capability of combiningmultiple-resolution radar and optical satellite datasets in annualplantation mapping to a large extent by using image classification andstatistical boundary-based change detection to achieve long time series. Theconsistent characterization of oil palm dynamics can be further used indownstream applications. The annual oil palm plantation maps from 2001 to2016 at 100 m resolution are published in the Tagged Image File Format withgeoreferencing information (GeoTIFF) athttps://doi.org/10.5281/zenodo.3467071 (Xu et al., 2019).
