Page 1, line 27: I prefer "continuous" to "numeric".  The model postulates
a quadratic dependence of the response, xylan, over intermediate values of the
factors TIME,TEMP and CONC to the levels chosen for the experiment, so
that the response surface is indeed continuous in these factors.

Pages 1-2, lines 28-29 and 1: I prefer 
   "In addition,  a modified severity index (log M0) constructed 
   as a function of all three factors, explains most of the variation
   in xylan or lignin reduction in simple linear regressions.  The
   coefficients of determination from these two models were R^2=0.89 
   and R^2=0.78, respectively." 
to   
   "In addition,  linear models  ... respectively."

In any event, where you say "parameter" as in the treatment parameter
TEMPERATURE, I am thinking experimental "factor", as in something that 
can be controlled in the designed experiment.  I know that the term
parameter is used by many authors as you've used it here, but I parameters
as unknown values that must be determined by the experiment.  I think 
the translation from greek for para-meter is "almost measure."  I'll
end my sermon here.

Page 2, line 22: [plausible]

Page 3, line 13: "[were] observed" ?

Page 3, line 24: [concerns]

Page 3, line 32: [amenable] ?

Page 5, line 20: [enzymatically] ?

Page 7, lines 1-4.  
Your description is fine.  This is how I would be inclined to describe
the modelling:

"The experimental design was crossed and complete with respect
to temperature, time and concentration.  Factorial effects models 
with main and interaction effects were fit
using PROC GLM in SAS.  Multiple comparisons among treatment means
were carried out using Tukey's procedure to control the
experimentwise error rate at 0.05 for each response variable.   In 
cases where interaction effects were significant, the SLICE option was 
used with the LSMEANS command of the GLM procedure to test for simple treatment 
effects of one factor while holding the other two factors constant."

Page 7, lines 8-9:  Again, I prefer "treatment effects of one experimental
factor ... other two treatment factors" to "treatment effects of one 
parameter..."

Page 7, line 21: n=0,1,...,[9]

Page 7, line 24-27:  I'm unclear on what validation was really done here.
I'm also not too worried about the bias in the assessment of error.  
With designed experiments, external validation is unnecessary, as there is 
no model selection involved.  In this case, we do have a little model
selection taking place when we exclude non-significant coefficients based
on the data.  I think it has been summarized perfectly in lines 24-26:
"Additionally, since the data set ... predictive ability of the models."
I suggest we might consider omitting the next sentence which begins
"Hence the same data ..."  as we  didn't really carry out any validation.

Page 8,  line 6:  [logarithm] instead of "log  value" ?

Page 9, line 5: [predominantly] [a]ffects hemicellulose with [] little 
impact [on]

Page 10, line 26: [solubilized] 

Page 13, lines 6-7: I might consider 

"There was no evidence (p>.05) of any effects of any
of the treatment factors on the lignin content of pretreated solids."

instead of

"There was no significant (p>.05) difference ..."

Page 13, line 8: [Hydrogen]

Page 13, line 9:  degradation[s]

Page 14, lines 21 and 26: "sparing" or "sparging" ?

Page 15, lines 26: "...2.2 times and 1.4  times thhe amount of acid 
insoluble material, [respectively], compared to sodium..."

Page 15, line 30: and concentration as [continous] variables

Page 16, line 15: [solubilized] 

Page 16, lines 27-31 and  page 17, lines32-33.: I am not following what is 
reported here.  The coefficient of determination, R^2 for any model is
also the squared correlation between the fitted and observed values.
Further, the "forced through zero" remark is confusing for me because
the estimated intercept will deterministically be 0 when the observed 
values are regressed on the fitted values.  I suspect the slight
disagreement between the R^2 you report after Eqn 6 (.964)and on 
line 30 (.96)is due to rounding or round-off error,  similarly for the
R^2=.924 before Eqn 7 and on line 1, page 17 (.90).  Another possibility
is that I'm missing something.

Page 18, lines 1-2,  I think the capacity of the modified severity [index]
to predict xylan solubilization is remarkable!   I  think the capacity
to predict %lignin reduction is still awfully good.  If we try to explain
why R^2 isn't closer to 100%,  we're  speculating.   One other possibility
is that the association between lignin redux and the treatments is
sufficiently complex so that the modified severity index cannot completely
characterize the association. 

Page 18, line 21: I prefer "treatment factor" over "treatment parameter."