Basic reporting

I must stress that I am not an expert on agricultural ecosystems or crop yields and what constrains them. That said, I'm interested in global environmental issues and understand both estimates of production and diversity that form the core of this paper. I accepted the chance to review this manuscript because I have followed the senior author's work closely and admire his creative approaches to many problems.

This manuscript addresses a disarmingly simple question: can we grow more crops if we adjusted the mix to maximise productivity. The answers answer an emphatic "yes" for cereals while improvements for oilseeds are much smaller. All that said, the most interesting aspects are why countries have not optimised productivity. The authors suggest a variety of possibilities. At the risk of asking them to expand a paper that is short and to the point, it seems that a more complete examination of what limits optimal production is warranted.

1. I consider the issue of spatial scale in the next section.

2. Would an optimal production lead to greater profits for the farmers? And to what extent are allocations driven by national subsidies?

3. Crop diversity is important. I found figures 3 and 4 to be most informative. Clearly most production is either close to optimal diversity or exceeds it considerably. The USA and China, for example, would need to move towards much less diverse croplands if they were to improve cereal production. Spain would need to become a cereal crop monoculture, for example.

4. Large food producers are unlikely to wish to modify current allocations to feed other countries. I would like to see the improvements of production ranked by the net balance of food exports and imports. Could food importers avoid such dependency? And at what cost in terms of crop specialization?

Experimental design

1. My first worry is about scale. 10 x 10km is fine scale, certainly, but I'd like to be reassured that the following possibility is excluded. For such a pixel, quite possibly irrigated rice may attain the highest productivity within a small piece of that. Extrapolating such productivity across the pixel may be impossible. Just think of what happens along the Nile, for example, where one can stand with one foot in very productive crops and the other in desert. Yes, irrigated rice is more productive than rainfed wheat, but that doesn't mean one can grow rice everywhere within the pixel.
   
   
2. The largest changes proposed would be to replace maize with wheat in Central Africa — for a huge increase in production — and to reduce wheat in China, but grow more rice. The authors mention these changes (page 5), but do not further investigate why the changes haven't been made. Water may well prevent rice from replacing wheat in China, and soil nutrients (and water) may well prevent wheat from replacing maize in Africa, especially one considers the scale issues I have already mentioned.
   
   The way to investigate these possibilities is to examine a sample of pixels that seem particularly suboptimal —where, for example, rice production is high per unit area within the pixel, but only a small fraction of the pixel grows rice. If that's an irrigation issue, then the authors need to assess how large an error this causes.

Validity of the findings

See concerns expressed above.

Comments for the author

I view this as being most interesting as a way of documenting what the limitations are to increased production. The bottom line — substantial improvements — are subject to many caveats. The value of this manuscript is to list what some of them are.