Overall a very good product, well-documented and with good validation and uncertainty analyses. A few remaining concerns:

Lines 107, 108 - “is the most biologically diverse terrestrial ecosystem on Earth”: unfortunately, many regions of the planet claim this distinction: temperate rainforests, high-elevation regions with multiple water/soil/exposure gradients, etc. Unless authors have a specific citation for this statement, or want to get into ‘greatest biodiversity’ arguments, I suggest they write ‘is one of the most’

Line 180 - “used a random forest (RF) classifier, a robust, stable and efficient machine”: many other researchers apply RF techniques (often referenced in ESSD) but they usually cite a reference, particularly if they claim ‘robust’ and ‘efficient’. Do these author choose to buttress this claim with citation(s)? If not, perhaps more caution in the statement?

Line 203 - “trees will be destroyed and transplanted”: Unless I mis-understand oil palm biology, a destroyed tree can not be subsequently transplanted? Older non-productive trees destroyed and plantation renewed by transplantation of new younger plants?

Line 226-228 - “For those pixels with less than 30 good-quality observations (4.79% in P1 and 9.64% in P2), we didn’t apply the BFAST algorithm. For the remaining area, 61.67% (P1) and 58.24% (P2) of pixels had 12 (~50%) good-quality observations annually.” I think a reviewer also asked about these numbers. The latter number of “12 (~50%) good-quality observations annually” refers to the 23 MODIS NDVI image per year (line 220)? But where does the ’30’ come from in “less than 30 good-quality”? Not from the 181 (P1) or 138 (P2) total scene because the percentage don’t match? Please clarify.

Lines 248-250 - “Considering the consistency in capturing significant changes (e.g., logging and replanting), predefined single conversion and computation volume, we used one of the commonly used change detection algorithms, BFAST, to capture the oil palm conversion time within the two periods (2011-2014 and 2001-2006).” Yes, but this sentence merely repeats what the authors wrote at the start of this paragraph. Authors should consider if they need this sentence or if in fact they intended to make a different point?

Line 280 - “break time detected from MODIS NDVI showed the same conversion year compared with the microwave satellite images.” Good important statement but readers should get a figure, an R2 value, something to validate this statement? As one reviewer requested, we need better information to certify this statement in order to build confidence in the remaining (temporal) assertions. Later (around line 295) a reader finds reference to “consistency of change time detection from different breakpoint test approaches” referring to figure S(upplement)7, but - to the extent that correspondence of the MODIS-determined time changes and the SAR-determined time changes serves as a fundamental assumption of the paper and as crucial link between two primary data sources, a reader needs clearer quantitative proof here? I see this request as entirely consistent with comments from the reviewer who questioned whether authors had demonstrated sufficient basis for their conclusions. Later (line 352) a reader learns that MODIS NDVI failed to capture ~ 10% of land use changes? Where did this number come from?

Lines 317-320: apparently authors had access to 3000 high-quality high-resolution GE images for Malaysia for the time period 2001 to 2016 (15 years) (later, more than 7000 for Indonesia). Over many regions of the globe, high-resolution GE images remain sparse, inconsistent and unreliable, particularly before 2010 or even - for some regions - 2015. Please can authors assure readers about this apparently positive availability of high-quality GE images? Figure 3 and Table 2 show only locations and cumulative numbers of images. All those locations and land-use types relied on high-res GE images? Lines 329-330 “limited samples (370, 25.07% of the total 1476 oil palm samples) with continuous high-resolution images from Google Earth” refers to exactly this problem of spare inconsistent GE images. Perhaps I have missed something about processing steps, but if so authors need to clarify for me and for other readers.

Thank you again for responsive changes!