Open Habitats under Threat in Mountainous, Mediterranean Landscapes: Land Abandonment Consequences in the Vegetation Cover of the Thessalian Part of Mt Agrafa (Central Greece)

Round 1

Reviewer 1 Report

The presented manuscript (ms) deals with the study of land abandonment effects in land use/land cover (LULC) changes in a Mediterranean landscape of Central Greece. The topic is very interesting, and the research tools used are relevant to the topic. The trend of grassland abandonment mainly in favor of woody vegetation also occurs in other parts of the European Mediterranean region and is considered to be a great environmental and cultural problem (in the sense of traditional pastoral landscape degradation). 

One major issue though in the ms is that the authors are not actually study grassland evolution but practically open habitat evolution.  As they state in lines 179 to 191 of the ms they combine in the term "Grassland" any type of herbaceous vegetation including agricultural lands (arable lands and fallow lands). The reason for this was that they considered that the separations of agricultural areas from grasslands was not possible (not safe as they state) in the black and white photos of 1945 and 1996. This statement based on my opinion is not correct. Yes, it is difficult to distinguish arable lands from grasslands, but it is also possible. There are a lot of papers and technical guides published over the years dealing with this issue.  The authors could recognize even in the oldest BW ortophotos of 1945 the geometric pattern of arable lands vs the irregular ones of grasslands. The grayscale tone is usually also different. Furthermore, there is a very useful cartographic material existing for Greece called maps of forest vegetation and land cover for 1960 (scale 1:20000), available for their study area that the authors could used as a photo interpretation guide in their effort to separate arable lands from grasslands. Finally, if they had the financial resources, they could obtain the original aerial photos from 1945 and investigate with the help of stereoscopic classes in more detail the difference between arable lands and grasslands.

Moreover, the comment that they make in lines 188 to 190 that grasslands are areas grazed by farm animals which is agricultural activity and for that reason and we can include arable lands as grasslands is also inappropriate. According to the definition of “Natural Grasslands” from Corine Lands Cover nomenclature: CLC code 321, Natural grasslands are areas with herbaceous vegetation ….. that can be extensively grazed, but never sown nor otherwise managed by application of fertilizers, pesticides, drainage or reseeding….

The study of the transition trend between grasslands and arable lands in areas suffering from land abandonment is a very interesting aspect of LULC analysis that is missing from this paper. This transition trend many times have opposite direction than the one expected and it is always useful and interesting to be investigated.

A.               Based on the above comments the use of the term “Grasslands” in the title of the ms and in the text as a definition of the third classification unit (line 202) is inappropriate. The authors should change from the title and from their main body of text (and all theirs maps and tables) the term “Grasslands” and replace it to “Open habitat” or “Open area” or something relevant and provide a new definition in line 202 as my example below

 Their title should be also change from “Grasslands under threat in mountainous…..” to “Open habitat under threat in mountainous…..” or skip completely the first line of their title which is also a bit long.

B.                The authors after changing their title and text they should adjust their finding especially in their discussion section. That practically means that they should compare their findings to similar ones from unified results of arable (agricultural) lands and grasslands from similar research papers from Mediterranean countries and Greece.

The first paragraph in the conclusion sections of the ms (lines 703 to 710) is written in the correct way that I am suggesting for the entire ms. Here the only correction could be (line 703 to 704) from …open habitat (i.e., grassland and other types of agriculture land) to …open habitat (i.e., grassland and agriculture land)

C.               They could include more relevant papers from Greece that are related to their topic as the followings:

-       Zomeni, M.; Tzanopoulos, J.; Pantis, J.D. Historical analysis of landscape change using remote sensing techniques: An explanatory tool for agricultural transformation in Greek rural areas. Landsc Urban Plan 2008, 86, 38–46.   https://doi.org/10.1016/j.landurbplan.2007.12.006     

-        Mallinis, G.; Emmanoloudis, D.; Giannakopoulos, V.; Maris, F.; Koutsias, N. Mapping and interpreting historical land cover/land use changes in a Natura 2000 site using earth observational data: The case of Nestos delta, Greece. Appl. Geogr. 2011, 31, 312-320, doi: https://doi.org/10.1016/j.apgeog.2010.07.002

-     Sidiropoulou, A.; Karatassiou, M.; Galidaki, G.; Sklavou, P. Landscape Pattern Changes in Response to Transhumance Abandonment on Mountain Vermio (North Greece). Sustainability 2015, 7, 15652–15673, https://doi.org/10.3390/su71115652

-      Chouvardas, D.; Karatassiou, M.; Tsioras, P.; Tsividis, I.; Palaiochorinos, S. Spatiotemporal Changes (1945–2020) in a Grazed Landscape of Northern Greece, in Relation to Socioeconomic Changes. Land 2022, 11, 1987. https://doi.org/10.3390/land11111987  

From the above paper they could obtain useful information about land abandonment issues in relation to agriculture and grazing activities and also enhance their discussion in the landscape metrics analysis.

D.               Finally, the authors could include a bit more technical details about the primary source of data as resolution or scale of their cartographic material. Also, they could prepare a table including all the necessary details about the biophysical and socioeconomic variables that they used as independent variable in their RF model instead of using so much text notes (lines 252 to 280).

As a final note, I believe that the revised manual will be suitable for publication in Land magazine.

Author Response

We would like to cordially thank the reviewer for his/her constructive comments. We tried to follow most of the reviewer’s suggestions. Please read below our response to each of the reviewer’s comments (they are given in italics below each comment). 

Response to the Reviewer’s 1 comments

One major issue though in the ms is that the authors are not actually study grassland evolution but practically open habitat evolution.  As they state in lines 179 to 191 of the ms they combine in the term "Grassland" any type of herbaceous vegetation including agricultural lands (arable lands and fallow lands). The reason for this was that they considered that the separations of agricultural areas from grasslands was not possible (not safe as they state) in the black and white photos of 1945 and 1996. This statement based on my opinion is not correct. Yes, it is difficult to distinguish arable lands from grasslands, but it is also possible. There are a lot of papers and technical guides published over the years dealing with this issue.  The authors could recognize even in the oldest BW ortophotos of 1945 the geometric pattern of arable lands vs the irregular ones of grasslands. The grayscale tone is usually also different. Furthermore, there is a very useful cartographic material existing for Greece called maps of forest vegetation and land cover for 1960 (scale 1:20000), available for their study area that the authors could used as a photo interpretation guide in their effort to separate arable lands from grasslands. Finally, if they had the financial resources, they could obtain the original aerial photos from 1945 and investigate with the help of stereoscopic classes in more detail the difference between arable lands and grasslands.

Response:

The reviewer is right in that whatever is referred to as grassland in our manuscript includes both semi-natural grasslands as well as agricultural lands (arable and fallow land). We explained this in detail in our manuscript. As the reviewer states, the distinction between farmland and grassland is difficult but still possible. Our initial aim was to distinguish between grassland and farmland but trying to do so we faced the problem of finding a continuous cline between actively cultivated land, fallow land and grassland. Fallow land in many cases looks like grassland in the orthophotos. Agricultural land in the area, many times, was left fallow for few years (sometimes even for more years because of temporary migration or change of occupation of local people) and then started to be cultivated again. Overgrazed grasslands, on the other hand, look like in some cases as cultivated fields. For these reasons as well as for those mentioned in the manuscript we considered that our interpretation regarding the occurrence of farmland or grassland would be subjective and thus biased. That’s why we compromised by unifying agricultural land and grassland. Nevertheless, we followed the suggestion of the reviewer at this issue by renaming grassland term all over the manuscript with the term open habitat (please see below).

Moreover, the comment that they make in lines 188 to 190 that grasslands are areas grazed by farm animals which is agricultural activity and for that reason and we can include arable lands as grasslands is also inappropriate. According to the definition of “Natural Grasslands” from Corine Lands Cover nomenclature: CLC code 321, Natural grasslands are areas with herbaceous vegetation ….. that can be extensively grazed, but never sown nor otherwise managed by application of fertilizers, pesticides, drainage or reseeding….

Response:

We removed this argument concerning the unification of farmland and grassland in one class, based on the above comment of the reviewer.

The study of the transition trend between grasslands and arable lands in areas suffering from land abandonment is a very interesting aspect of LULC analysis that is missing from this paper. This transition trend many times have opposite direction than the one expected and it is always useful and interesting to be investigated.

Response:

Indeed, the transition trend between grasslands and arable lands in areas suffering from land abandonment is interesting. We report in the manuscript that for the most recent studied period (2015) almost all farmland was observed to be abandoned. This indicates that farmland followed a corresponding decreasing trend with that of grassland. Nevertheless, we cannot discuss more on the trends of farmlands as we unified them with grasslands.

1. Based on the above comments the use of the term “Grasslands” in the title of the ms and in the text as a definition of the third classification unit (line 202) is inappropriate. The authors should change from the title and from their main body of text (and all theirs maps and tables) the term “Grasslands” and replace it to “Open habitat” or “Open area” or something relevant and provide a new definition in line 202 as my example below

Grasslands or arable lands with canopy cover of trees and shrubs <20%

 Their title should be also change from “Grasslands under threat in mountainous…..” to “Open habitat under threat in mountainous…..” or skip completely the first line of their title which is also a bit long.

Response:

We followed reviewer’s suggestion. In the title, text and figures of the manuscript, as well as in the supplementary file we have replaced “Grassland” with “Open habitat”. We redefined this class in Table 1 as suggested by reviewer (Open habitat: Grasslands or arable lands with canopy cover of trees and shrubs <20%.)

2. Comment:

The authors after changing their title and text they should adjust their finding especially in their discussion section. That practically means that they should compare their findings to similar ones from unified results of arable (agricultural) lands and grasslands from similar research papers from Mediterranean countries and Greece.

Response:

We followed reviewer’s suggestion. The comparison of our results with those of corresponding studies has been updated. However, as we had already based our comparisons in the discussion on a unified class of grassland and agricultural land, only minor modifications were needed. Additionally, in some other studies (e.g. Oikonomakis & Ganatsas, 2012) a similar unification of grassland and agricultural land in one land class was followed.

The first paragraph in the conclusion sections of the ms (lines 703 to 710) is written in the correct way that I am suggesting for the entire ms. Here the only correction could be (line 703 to 704) from …open habitat (i.e., grassland and other types of agriculture land) to …open habitat (i.e., grassland and agriculture land)

Response:

We followed reviewer’s suggestion.

3. Comment:

They could include more relevant papers from Greece that are related to their topic as the followings:

 -       Zomeni, M.; Tzanopoulos, J.; Pantis, J.D. Historical analysis of landscape change using remote sensing techniques: An explanatory tool for agricultural transformation in Greek rural areas. Landsc Urban Plan 2008, 86, 38–46.   https://doi.org/10.1016/j.landurbplan.2007.12.006    

-        Mallinis, G.; Emmanoloudis, D.; Giannakopoulos, V.; Maris, F.; Koutsias, N. Mapping and interpreting historical land cover/land use changes in a Natura 2000 site using earth observational data: The case of Nestos delta, Greece. Appl. Geogr. 2011, 31, 312-320, doi: https://doi.org/10.1016/j.apgeog.2010.07.002

-     Sidiropoulou, A.; Karatassiou, M.; Galidaki, G.; Sklavou, P. Landscape Pattern Changes in Response to Transhumance Abandonment on Mountain Vermio (North Greece). Sustainability 2015, 7, 15652–15673, https://doi.org/10.3390/su71115652

-      Chouvardas, D.; Karatassiou, M.; Tsioras, P.; Tsividis, I.; Palaiochorinos, S. Spatiotemporal Changes (1945–2020) in a Grazed Landscape of Northern Greece, in Relation to Socioeconomic Changes. Land 2022, 11, 1987. https://doi.org/10.3390/land11111987 

Response:

We added the above studies to the discussion chapter. Specifically, we cite and discuss these studies in the lines 653-655, 669-672, 612-614 and 711-713.

4. Comment:

Finally, the authors could include a bit more technical details about the primary source of data as resolution or scale of their cartographic material. Also, they could prepare a table including all the necessary details about the biophysical and socioeconomic variables that they used as independent variable in their RF model instead of using so much text notes (lines 252 to 280).

Response:

We added technical details of the orthophotos we used, such as scale and color range (lines 166-173).

We added a table (Table S1) in the supplementary with the explanatory variables used in the model.

Reviewer 2 Report

In this work land abandonment drivers and effects were explored using orthoimages and random forest classifier. The work is very interesting and worth publishing. Nevertheless, some problems in the data description and methods adopted are present. The main problem concerns the replicability of current method. In fact, many processing steps and classification rules were subjectively set. Therefore, I suggest a major revision of current manuscript.

Specific comments:

1.       Line 112.  Please better specify CSb and Csa meanings.

2.       L119. Please adopt the correct nomenclature rules (i.e., add author who first discovered that species).

3.       L166. Please specify: geometric resolution (pixel size), the nominal scale of orthoimages, and if the orthophotos are true colors or false colors and which spectral range they explored (e.g. visible or vis+ NIR).

4.       L169-171. Please specify the parameters set to perform segmentation.

5.       L172-173. Specify the photointerpretation scale.

6.       L178-179. How did you measure by only photointerpretation the canopy cover? This is a very subjectively assessment and can compromise the reliability of your method. In fact, without an objectively measure of canopy cover different researchers can obtain different results and interpretations....

7.       L233. How did you define "gross" gain/loss. Gross is something subjectively. Please provide more details.

8.       L251. A distribution may be normal, binomial, exponential, asymmetric but no linear... linear may be referred to variables’ separability or some regressions/classification models.... Please change this sentence.

9.       L285-286. This is a poor choice. In fact, a variable may be important (high weight) for a specific model, but is could be a confounder variable and negatively affecting the relationships between other variables. 

Author Response

We would like to cordially thank the reviewer for his/her constructive comments. We tried to follow most of the reviewer’s suggestions. Please read below our response to each of the reviewer’s comments (they are given in italics below each comment).

Response to the Reviewer’s 2 comments

In this work land abandonment drivers and effects were explored using orthoimages and random forest classifier. The work is very interesting and worth publishing. Nevertheless, some problems in the data description and methods adopted are present. The main problem concerns the replicability of current method. In fact, many processing steps and classification rules were subjectively set. Therefore, I suggest a major revision of current manuscript.

We followed most suggestions of the reviewers so as to make clear any issue related to the data used and the methods applied in the study. However, few of the reviewer’s comments possibly were caused by misunderstanding of what is written in the manuscript (please, see our responses below).

1. Comment:

Line 112.  Please better specify CSb and Csa meanings

Response:

We added the full description of the acronyms of the climatic types according to Köppen’s climate classification (lines 111-113).

2. Comment:

L119. Please adopt the correct nomenclature rules (i.e., add author who first discovered that species).

Response:

We added authors of plant names whenever a taxon name is mentioned in the manuscript. We followed nomenclature of Euro+Med database, which is the most commonly used and accepted database of vascular plants nomenclature for European and Mediterranean area. (lines 119-122).

3. Comment:

L166. Please specify: geometric resolution (pixel size), the nominal scale of orthoimages, and if the orthophotos are true colors or false colors and which spectral range they explored (e.g. visible or vis+ NIR).

Response:

We added in the manuscript the details of the orthophotos requested by the reviewer (lines 166-173).

4. Comment:

L169-171. Please specify the parameters set to perform segmentation.

Response:

We updated the text adding the parameters used to perform map segmentation (lines 177-179).

5. Comment:

L172-173. Specify the photointerpretation scale.

Response:

We updated the text adding the photointerpretation scale (lines 182-183).

6. Comment:

L178-179. How did you measure by only photointerpretation the canopy cover? This is a very subjectively assessment and can compromise the reliability of your method. In fact, without an objectively measure of canopy cover different researchers can obtain different results and interpretations....

Response:

Initially, the software eCognition was used for image segmentation. Unfortunately, the quality of the orthophotos that are available for the study area and for the studied years, but also in general before 2005, does not allow the application of an image-based method for measuring the canopy cover. Specifically, the black and white orthophotos of 1945 and 1996 have significant differences in their brightness and contrast between the orthophotos of the same year. In the mountainous relief of the study there were shadows in many localities preventing a safe measurement of canopy cover by an image-based method.

7. Comment:

L233. How did you define "gross" gain/loss. Gross is something subjectively. Please provide more details.

Response:

Gross gain and loss were not defined by us, but they are calculated from OpenLand R package according to the methodology presented by Aldwaik and Pontius (2012).

Aldwaik, S. Z., & Pontius, R. G. (2012). Intensity analysis to unify measurements of size and stationarity of land changes by interval, category, and transition. Landscape and Urban Planning, 106(1), 103-114.

8. Comment:

L251. A distribution may be normal, binomial, exponential, asymmetric but no linear... linear may be referred to variables’ separability or some regressions/classification models.... Please change this sentence.

Response:

We corrected the sentence (lines 257-260).

9. Comment:

L285-286. This is a poor choice. In fact, a variable may be important (high weight) for a specific model, but is could be a confounder variable and negatively affecting the relationships between other variables.

Response:

At this point we are referring to the explanatory variables that are highly correlated and should be excluded from the model building anyway. The common practice that can be found in bibliography regarding the selection of the variables to be retained or deleted from model building is that the authors select the variables either randomly or based on their intuitive judgement (e.g. they retain the variables which they consider as mathematically more simple or ecologically more interpretable) (e.g. Zanela et al. 2017, Gazis and Greinert 2021, Revermann et al. 2016, Dormann et al. 2013, Kiziridis et al. 2022). Instead of selecting the variables to be retained in the model based only on our intuition we took into account also their importance in a model run. Nevertheless, our choice does not affect the modeling results as only one from the highly correlated variables is retained, while the others are removed (randomly, subjectively or by a logical rule such as that used in our study).

Zanella, L., Folkard, A. M., Blackburn, G. A., & Carvalho, L. M. (2017). How well does random forest analysis model deforestation and forest fragmentation in the Brazilian Atlantic forest?. Environmental and ecological statistics, 24, 529-549.

Gazis, I. Z., & Greinert, J. (2021). Importance of Spatial Autocorrelation in Machine Learning Modeling of Polymetallic Nodules, Model Uncertainty and Transferability at Local Scale. Minerals, 11(11), 1172.

Revermann, R., Finckh, M., Stellmes, M., Strohbach, B. J., Frantz, D., & Oldeland, J. (2016). Linking land surface phenology and vegetation-plot databases to model terrestrial plant α-diversity of the Okavango Basin. Remote Sensing, 8(5), 370.

Dormann, C. F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., ... & Lautenbach, S. (2013). Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography, 36(1), 27-46.

Kiziridis, D.A.; Mastrogianni, A.; Pleniou, M.; Karadimou, E.; Tsiftsis, S.; Xystrakis, F.; Tsiripidis, I. Acceleration and Relocation of Abandonment in a Mediterranean Mountainous Landscape: Drivers, Consequences, and Management Implications. Land 2022, 11, 406, doi:10.3390/land11030406.

Reviewer 3 Report

The manuscript entitled “Grasslands under threat in mountainous, Mediterranean landscapes: land abandonment consequences in vegetation cover of the Thessalian part of Mt Agrafa (central Greece” is scientifically correct. Thus it is relevant for the Journal Forests MDPI after minor revision

The structure of the manuscript is well organized. The conclusions correspond to the presented results. The authors have processes a big amount of data.

I would like to draw author’s attention to some minor comments:

In Row 174 the authors have defined 6 types of land cover. It is not clear whether the definition of the types is based on some national/ european/other legislation or other document OR according to the authors’ assessment/decision. Is it possible to be specified and noted in the text?

Row 199

The first line of table 1.  the classification criterion is based on your own consideration or according to legislation  or other requirements? Is it possible to be specified and explained in the text or noted in the table?

Rows 233-236

I suggest the information should be displayed with a formula.

Row 259-260

In the sentence “Soil erosion and depth and substrate type variables were taken from [27]” –a text  is missing

Rows 274-275

Is it possible to explain the figure 6.66:  where it comes from, is it according to national regulation or … ?  I suggest that authors specify this issue.

Good luck!

Reviewer

Comments for author File: Comments.pdf

Author Response

We would like to cordially thank the reviewer for his/her constructive comments. We tried to follow most of the reviewer’s suggestions. Please read below our response to each of the reviewer’s comments (they are given in italics below each comment).

Response to the Reviewer’s 3 comments

The manuscript entitled “Grasslands under threat in mountainous, Mediterranean landscapes: land abandonment consequences in vegetation cover of the Thessalian part of Mt Agrafa (central Greece” is scientifically correct. Thus it is relevant for the Journal Forests MDPI after minor revision

The structure of the manuscript is well organized. The conclusions correspond to the presented results. The authors have processes a big amount of data.

I would like to draw author’s attention to some minor comments:

1. Comment:

In Row 174 the authors have defined 6 types of land cover. It is not clear whether the definition of the types is based on some national/ european/other legislation or other document OR according to the authors’ assessment/decision. Is it possible to be specified and noted in the text?

Response:

We updated the text and pointed out that the classes used in the manuscript were based on the CLC classes (Kosztra et al., 2017) and were slightly modified to serve better our aims (lines 183-185).

2. Comment:

Row 199

The first line of table 1.  the classification criterion is based on your own consideration or according to legislation or other requirements? Is it possible to be specified and explained in the text or noted in the table?

Response:

This comment concerns a corresponding issue to that of the previous comment 1. Please see our response to the previous comment.

3. Comment:

Rows 233-236

I suggest the information should be displayed with a formula.

Response:

Gross gain and loss are simple quantitative terms of which the definitions are given in the text. We consider that the addition of formulas would only confuse the reader and will not add something more than their simple definition. Thus, we didn’t add these formulas.

4. Comment:

Row 259-260

In the sentence “Soil erosion and depth and substrate type variables were taken from [27]” –a text  is missing

Response:

We updated the text and completed the sentence (line 269).

5. Comment:

Rows 274-275

Is it possible to explain the figure 6.66:  where it comes from, is it according to national regulation or … ?  I suggest that authors specify this issue.

Response:

We updated the text and added the source/regulation that used to calculate “livestock” size (lines 283-286).

Round 2

Reviewer 1 Report

The authors were properly revised the manuscript according to my suggestions and I consider that it can be published in its present form. Congratulation!

Reviewer 2 Report

Authors have improved the manuscript according to my comments. Therfore, i think that now this paper is ready to be published.