Peer review report 1 On “Statistical Regression Models for Assessing Climate Impacts on Crop Yields - A Validation Study for Winter Wheat and Silage Maize in Germany”
Peer review report 1 On “Statistical Regression Models for Assessing Climate Impacts on Crop Yields - A Validation Study for Winter Wheat and Silage Maize in Germany”
Agricultural and Forest Meteorology 201S (2015) 684–685
Contents lists available at ScienceDirect
Agricultural and Forest Meteorology journal homepa...
Agricultural and Forest Meteorology 201S (2015) 684–685
Contents lists available at ScienceDirect
Agricultural and Forest Meteorology journal homepage: www.elsevier.com/locate/agrformet
Peer Review Report
Peer review report 1 On “Statistical Regression Models for Assessing Climate Impacts on Crop Yields - A Validation Study for Winter Wheat and Silage Maize in Germany”
1. Original Submission 1.1. Recommendation Minor Revision 2. Comments to Author The authors have presented a compelling statistical study of silage maize and winter wheat in Germany using a suite of statistical models. The article is nearly ready for publication, pending several clarifications, improvement in the language and some consideration of additional references. 3. References A similar model to the STSM was employed for US maize in Butler and Huybers, Adaptation of US Maize to temperature variations, Nature Climate Change (2013) who found a similar relationship between sensitivity and climatology as that noted here. The results of that study may be useful when considering extrapolating the results of such statistical models into novel climates. The authors are also advised that months are not always an adequate substitute for phenology data as Butler and Huybers, Variations in the sensitivity of US maize yield to extreme temperatures by region and growth phase, Environmental Research Letters (2015) indicate. In particular see their comparison of sensitivity estimates using fixed and variable phenology in the supplemental material. Finally, Mueller, et al. Closing yield gaps through nutrient and water management, Nature (2012) developed a high resolution fertilizer dataset which could potentially be used to downscale the nation level data presented in your analysis. 4. Scientific Questions 1) The authors indicate in their abstract that “The statistical yield models can capture effects, which are influenced by the climatic or the economic variables, for example, pest and disease pressure or inventory management on crop yields and thus can be used for risk assessments.” How does the model account for these variables?
DOI of published article: http://dx.doi.org/10.1016/j.agrformet.2015.10.005. 0168-1923/$ – see front matter http://dx.doi.org/10.1016/j.agrformet.2015.12.053
There is no indication in the model equations that pest or disease pressure are accounted for other than through correlation with climate variables. As such I find this statement rather misleading and suggest that it be clarified or removed. 2) in Eqn. (3) most panel models allow for county scale fixed effects. In your model the beta 0 term would then have an i subscript. The preceding paragraph also alludes to this. Is that how this model is constructed? If not how are local effects accounted for in the model? 3) Pg 12 l 293-294: What correlation was used in these calculations (Pearson, Spearman, etc) and how were the degrees of freedom determined when calculating the significance? 5. Language suggested corrections are in scare quotes “”, where empty scare quotes indicate deleting words. Pg 4 l 88: stress as “the” mostl 92: with “” Germanl 92: like the “crop used” Pg 5 l 119: or “” Germany Pg 6 l 151: used as “the” basic functionl 152: (y’ t) “to”l 153: (x’ kt) “”. Statistical Pg 8 l 198: division “is based” on Pg 9 l 215: (SRT) for “growth potential”l 217: Note that R s has not yet been defined. Pg 10 l 257: To “more rigorously check” the robustnessl 258: process “by removing four more years, in addition to the year t.” The one-yearl 264: (RESET) allows “an evaluation of”l 268: to test “for” heteroskedasticityl 269: RCMs “there” existl 270: model goodness “of fit”l 270: because it is “suitable” for REML Pg 11 l 271: The meaning of this line is not clear - perhaps .. has advantages “over” Akaike.. this would call for a citation to defend the claiml 272: the “software used” is in Pg 12 l 286: (“examples” shownl 286: validation“s”l 287: models, “a” decreasel 288: common “when” comparingl 288: decreases “by approximately” 0.38l 293: For “” both crops “there exists”l 296: on “the” federall 312: fit distribution“s” are “” depicted Pg 13 l 326: others “are” distribute“d” aroundl 338: were not “dependent”l 338: function as “a” functional form Pg 14 l 350: lower “than” 12 Pg 15 l 357: able to “satisfactorily reproduce” thel 377: can “lead” to Pg 15 l 385 - Pg 16 l 386: (2012). “” Consistent with this explanation
Pg 16 l 404-405: by the “variance differences among the variables.” Pg 17 l 418: parameters “may also” reflectl 433-434: Several “factors and factor relationships that are unknown today” might play a major role in “the” futurel 435: What do you mean by “regularly returning yield impacts”?l 437: Crossing new thresholds does not mean that the crops were insensitive to them in the past, only that they may have rarely (if ever) experienced conditions that pushed them past the threshold. This line should be clarified to reflect this. Pg 18 l 450: scale should “only be” used
685
Pg 19: Figure “Captions” and Table “Captions” Supplemental Information: Pg 1 l 6: are “digitized” froml 24: This statement is not entirely clear, but I believe you mean: .. are not “only” crucial .. or perhaps, “Statistical significance is not the only criteria for parameter selection”? Anonymous Available online 18 December 2015