Review report “Estimating global land system impacts of timber plantations using MAgPIE 4.3.2 ”

The manuscript presents an updated version of MAgPIE forest sector module, which includes forest age-class dynamics and two types of managed forests (timber plantations, natural forests). Usually large-scale land-use models such as MAgPIE do not include forest age-class dynamics, so this is an important contribution in the field and deserves to be published. However, a proper modelling of  forest management requires that forest-owners anticipate the future changes in harvest volumes. MAgPIE is a recursive dynamic model, which means that the model is forced to make  some simplifications concerning the future foresight. The manuscript should include more detailed discussion on these simplifications since they seem to have significant impact on the results.      

General comments:

1) p12L200-215: The expansion of timber plantation depends on the share of production that comes from plantations (η). This parameter is exogenous and extrapolated from Pöyry (1999). This means that the expansion of plantation is not endogenous in the model, but it is taken as given. According to figure A5, the model assumes that share plantation increases on average from 25%  in 2000 to 62.5 % in 2100. This issue should be made clear already in the abstract because it has significant impact on the results. For example, if the share of plantations were endogenous in the model, then an increasing demand for roundwood would increase the share of plantations relative to natural forests in EUR region. But because the share is exogenous this does not happen, and EUR region is not able adapt higher demand by intensifying their forest management (Figure A6).

2) The outcome of the optimal rotation models depends much on interest rates and usually these models  include sensitivity analysis relative to different interest rate. To avoid this complication, the rotation times could be solved by maximizing increment (f’=f(ac)/ac) instead of maximizing NPV (f’(ac)/f(ac)=r). This would also be more reasonable objective for the recursive dynamic model where all other choices are based on recursive optimization instead of intertemporal optimization.  

3)  Add some discussion about the forest age-class dynamics and optimal rotation models in the introduction.  Basically move some material from discussion to introduction.  Including forest age-class dynamics in the large-scale land-use model is the main contribution of the study, but this issue is completely ignored in the introduction.   

Specific comments:

1)  p2L26: According to FAOSTAT global roundwood demand was 3969 Mm3 in 2019 and industrial roundwood 2024 Mm3. Global roundwood demand cannot be  1683 Mm3.  

2) p2L26 Add reference or explanation for 33% (560 Mm3) plantation supply -> Pöyry (1999) extrapolation (Figure A5). This it is not data but model outcome.  

3) p2L38: Add more relevant references for high roundwood productivity of plantations relative to natural forests than FAO (2013), e.g.  IPCC (2006). Also, add some explanation why roundwood productivity is higher in plantations than managed natural forests.

4) p6L124, p6L132, p7L136: Equation should be f ’/ f=r.       

5) Rotation times for timber plantations in Figure 4 are “interesting”, but the question is how reasonable they are. For example, with 30-40 years rotation time in Russia and Europe you get only pulpwood (sawlogs require 60-100 years rotation). Moreover, it is not clear why rotation times are longer in North-America than in Europe and Russia. Is this connected to interest rates or productivity?  There is only a small difference in interest rates (Table A2) and there should not be large differences in biomass growth between these regions.  Some discussion of this should be added and eventually an update to growth curves, interest rate data and add a minimum diameter constraint for sawlogs.

6) Is rotation time for natural forests determined by the same rule than for timber plantations (equation 1). If yes, then add similar map (Figure 3) for natural forest rotation time. It would be interesting to  see the regional difference between timber plantations and natural forests rotation times. If no, then add some justification why natural forest rotation time is chosen differently than in timber plantations. Basically explain also natural forest rotation lengths in chapter 2.3.  

7) According to Figure A8 EUR region growing stock decreases close to zero in 2100, which implies that forest management is not sustainable in EUR region. Easy way to avoid this would be to add additional “sustainability” constraint on harvests (harvests ≤ α x increment where α=1 for normal forests, α> 1 for old forests and α < 1 for younger forests).  Alternatively, increase the share of plantation in EUR region. Basically take account in extrapolation of η that demand increases in the future. Interesting is also opposite development of growing stock in CHA region. This seems to be connected to the higher share of roundwood production coming from plantation in CHA region (Figure A7). Why cannot EUR region do the same as CHA region and avoid the decrease in growing stock ?   

The manuscript presents the first results on land use after the implementation of a dynamic forestry component in

MAGPIE (which is an economic rooted land use model). Prior to this, MAGPIE basically covered the agricultural and

the bioenergy sector in the model. Undoubtedly, this new forestry component represent a relevant improvement in the 

model, increasing the coherence and the completeness of its land use simulations. Given that, I accept the publication 

of the manuscript, but it is important that minor (perhaps "moderate") revisions/suggestions should be considered in a revised version 

or in a rebuttal document. 

General comments.

I understood the principle of rotation lenghts derived from the relation between interest rate 

and IGR. This generates information on rotation lenghts and it is important for calculating timber production costs. I was wondering 

whether some reference data could be added to validate the rotation lenght map. I`m not a forest specialist, but in 

principle Latin America and Australia should have the similar spatial pattern, given that that both regions are 

important eucalyptus producers. Instead, the map shows different spatial patterns. 

I believe that an analysis on the main species (and they respective rotation period) per region could be used to 

calibrate the rotation lenght map.

It is clear that the study was focused on creating a proof of concept that enables a forestry module in MAGPIE. 

However, I missed a bit the discussion on how realistic are the figures presented. 

For example, in section 3.1, it is mentioned a large increase in cropland at expense of primary forest areas (where exactly?). Are protected areas included in the analysis? What about agro-forestry (also included?)?

Perhaps some discussions around the current and future spatial uncertainties would be relevant as well.

Productivity is an important component for calculating global costs of demand driven land uses. I`m not fully aware of 

LPJML, but doing a quick research, I found that LPJML might include the productivity of natural vegetation and planted forests. These were surely incoporated in MAGPIE (right?). But yet it is not clear to me how the productivity of secondary forest came about?

Specific comments.

Line 175. If a fraction of forestry residues is recovered during the harvesting period, it is likely that there will be a potential 

decay in productivity in the forthcoming period in comparison with a plantation system that does not recover any fraction of 

residues. It was not cleat for me if that was included in the model, but it is something to be considered.

Typo.

Line 119. "optimal"