Interactive comment on “ Gross and net land cover changes based on plant functional types derived from the annual ESA CCI land cover maps ”

In abstract: “The annual ESA CCI land cover products can be downloaded from http://maps.elie.ucl.ac.be/CCI/viewer/download.php” This paper is focused on the derivation of PFT change estimates from the ESA CCI LC product. (“our analyses are based on the PFT maps that have been translated from the ESA CCI LC maps, rather than the original LC classes”). Only one “example of LC map and PFT map in 2000 used in this study” is made available (“can be downloaded from doi: https://doi.org/10.5281/zenodo.834229”)

The paper focuses mainly on the comparison of estimates of areas, gross and net changes of different plant functional types (maps of PFTs derived from the ESA CCI LC product) with 3 other sources : Hurtt et al. (2011), Hansen et al. (2013 and Houghton and Nassikas (2017). This comparison is useful for understanding the range of discrepancies between such datasets.

C1
The datasets and the results of the comparison are clearly expressed and well presented. However the discussion should be complemented with further issues which can explain part of the discrepancies between the CCI LC derived PFT dataset and independent datasets (see section 'Comparison with other datasets' in Specific Comments for detailed information on such issues).

Specific comments
Title: I suggest to revise the title in order to relate better to the content of paper, e.g. : "Gross and net land cover changes of the main plant functional types derived from the annual ESA CCI land cover maps (1992-2015)" Access to the core dataset of this study: In abstract: "The annual ESA CCI land cover products can be downloaded from http://maps.elie.ucl.ac.be/CCI/viewer/download.php" This paper is focused on the derivation of PFT change estimates from the ESA CCI LC product. ("our analyses are based on the PFT maps that have been translated from the ESA CCI LC maps, rather than the original LC classes"). Only one "example of LC map and PFT map in 2000 used in this study" is made available ("can be downloaded from doi: https://doi.org/10.5281/zenodo.834229") I consider that it would be more appropriate and pertinent for this paper to provide access to the full derived dataset (PFT annual maps at 0.5 × 0.5deg resolution) as main product of this study -in complement to the ESA CCI LC product which is already available through ESA and UCL web sites.
Use of FAO data: The discussion on differences in area and area changes (section 4.1. and 4.2) is interesting and covering a number of important issues, but it should be complemented by at least two further issues: A main difference between FAO FRA-2015 dataset and Hansen et al (2013) product is that FAO reports a land use definition when Hansen reports a Tree cover percentages. A major impact of such differences in definition is related to Oil Palm plantations. Oil Palm plantations are not reported as forests by FAO (considered as agricultural use) when they are mapped as dense Tree Cover by Hansen. This difference has major impacts on estimates of LC changes for countries like Indonesia. It would be useful to pay attention to this specific vegetation type and to mention the Land Cover class under which are mapped Oil Palm plantations in CCI LC product (regional class 'Tree or Scrub Cover' under 'Cropland' first level class) and to mention to which PFT it has been attributed (Forest or croplands).
It is also known in the remote sensing community that it is difficult to map and estimate forest areas in the dry tropics with medium resolution satellite (Landsat type), in particular when tree cover is below 40 %. Consequently it is at least as challenging or more difficult to map such forests from coarser resolution imagery or to estimate accurately area changes from medium or coarse resolution data. This is illustrated, reported or discussed in a number of papers including Hansen et al ( The remote sensing community, in particular scientists dealing with monitoring of REDD+ activities, has produced technical guidelines or scientific papers which report that it is more efficient and accurate to produce area estimates by combining a sample of reference dataset (sample of reference plots) with a wall to wall map, than by using only a wall to wall map. This is particularly valid for estimating Land Cover changes which are usually considered as 'rare' events. See