Submit to this Journal Review for this Journal Propose a Special Issue

Article Menu

Share Help Cite Discuss in SciProfiles

Open AccessArticle

Peer-Review Record

Exploring Short-Term Climate Change Effects on Rangelands and Broad-Leaved Forests by Free Satellite Data in Aosta Valley (Northwest Italy)

Climate 2021, 9(3), 47; https://doi.org/10.3390/cli9030047

by Tommaso Orusa^1,*

and Enrico Borgogno Mondino²

Reviewer 1: Anonymous

Reviewer 2: Anonymous

Reviewer 3: Anonymous

Climate 2021, 9(3), 47; https://doi.org/10.3390/cli9030047

Submission received: 26 January 2021 / Revised: 26 February 2021 / Accepted: 12 March 2021 / Published: 15 March 2021

(This article belongs to the Special Issue The Interaction of Climate Change with Landscape and Environment)

Round 1

Reviewer 1 Report

The authors have already made substantial amendments to previously submitted paper, which goes in the right direction. however, I still hope the authors could make some additional clarifications to my previous comments.

First, the authors added supplementary analysis in the results and observed some interesting trends in PMs. Such finding is well worth to be mentioned in the abstract, otherwise, the readers will not receive nothing critical except for the datasets used in this study. Therefore, the authors are recommended to simplify the datasets of this study, and at least how PMs changed over the past two decades and its cause.

Second, I have noticed that the authors applied an arbitrary value of 0.5 to interpret the date of SOS and EOS from NDVI seasonal curve (see lines: 351-355). The threshold of 0.5 might be problematic if the MAXVI falls below 0.5. Further, the authors said the derived PMs are very close to the MODIS phenology product (MCD12Q2, based on pairs of piecewise logistic functions, see Zhang et al., 2003, RSE). In theory, the NDVI value at the date of SOS and EOS is usually much lower than 0.5. I would like the authors provide a supplementary figure showing the relationships between PMs and MODIS phenology product.

Author Response

Response to Reviewer 1 Comments

We would like to thank reviewers for their appropriate comments and helpful suggestions that have been carefully considered. Majority of provided suggestions highlighted gaps in the text and were really useful to improve, we hope, paper quality.

In particular, the synthesis of reviewers’ comments suggested a deep revision in paper organization and harmonization. Consequently, you will find some structural changes aimed at simplifying paper reading and make content more effective.

All comments were carefully evaluated and for the most of them corrections and integrations have been provided.

Point 1: The authors have already made substantial amendments to previously submitted paper, which goes in the right direction. however, I still hope the authors could make some additional clarifications to my previous comments.

Response 1: The referee is right. Taking into account his/her precious observation the following modification has been apported to the abstract to clarify the activity made. The dataset adopted had no received too much variation because they have been used to retrieved the following patterns: PMs, ET, snowpack time persistence to soil, air temperature anomaly and precipitation. Anyway, an explanation in the abstract has been done in order to avoid misunderstanding and following your wisely advice as follow:

“Satellite remote sensing is a power tool for long-term monitoring of vegetation. This work, with reference to a regional case study, investigates remote sensing potentialities for describing annual phenology of rangelands and broad-leaved forest at landscape level with the aim of detecting eventual effects of climate change in the Alpine region of the Aosta Valley (NW Italy). A first analysis was aimed at estimating phenological metrics (PMs) from satellite images time series and testing the presence of trends along time. A further investigation concerned evapotranspiration from vegetation (ET) and its variation along years. Additionally, in both the cases the following meteorological patterns were considered: air temperature anomalies, precipitation trends and timing of yearly seasonal snow melt. The analysis was based on time series (TS) of different MODIS collections datasets together with CHIRPS collection obtained through Google Earth Engine. Ground weather stations data from the Centro Funzionale VdA ranging from 2000 to 2019 were used. In particular, the MOD13Q1 v.6, MOD16A2 and MOD10A1 v.6 collections were used to derive PMs, ET and Snow Cover maps. The SRTM (Shuttle Radar Topography Mission) DTM (Digital Terrain Model) was also used to describe local topography while the CORINE Land Cover map was adopted to investigate land use classes. PMs and ET and snow cover melting proved to have significantly changed their values in the last 20 years, with a continuous progressive trend. Averagely in the area, rangelands and broad-leaved forests, showed that the length of season is getting longer with a general advance of SOS (Start of the Season) and a delay in EOS (End of the Season). With reference to ET, significant increasing trends were generally observed. Water requirement from vegetation appeared to have averagely risen up of about 0.05 Kg·m^-2 (about 0.5%) per year in the period 2000-2019, for a total increase of about 1 Kg·m^-2 in 20 years (corresponding to a percentage difference in water requirement from vegetation of about 8%). This aspect can be particularly relevant in the bottom of the central valley where the precipitations have shown a statistically significant decreasing trend in the period 2000-2019 (conversely, no significant variation was found in the whole territory). Additionally, the snowpack timing persistence showed a general reduction trend. This fact, together with an increase of air temperature, could be one of the main reasons that explain PMs trends. Results encourage the adoption of remote sensing to monitor climate change effects on alpine vegetation, with particular focus on the relationship between phenology and other abiotic factors, like snowpack timing and temperature, supporting the idea that, presently, an effective technology transfer is now possible to dynamically support management of mountain vegetation. Specifically, remote sensing-derived information can help breeders, agronomists, foresters and also policy-makers to better manage alpine ecosystems by adapting their practices to better face climate change effects”.

(Please see lines 13 – 50)

Point 2: Second, I have noticed that the authors applied an arbitrary value of 0.5 to interpret the date of SOS and EOS from NDVI seasonal curve (see lines: 351-355). The threshold of 0.5 might be problematic if the MAXVI falls below 0.5. Further, the authors said the derived PMs are very close to the MODIS phenology product (MCD12Q2, based on pairs of piecewise logistic functions, see Zhang et al., 2003, RSE). In theory, the NDVI value at the date of SOS and EOS is usually much lower than 0.5. I would like the authors provide a supplementary figure showing the relationships between PMs and MODIS phenology product.

Response 2: The referee suggestion is correct but we have considered that the threshold of 0.5 as relative and not absolute (please see the better explanation given). This was done in order to evaluate the active phenological season in general with the aim to remove noise and to perform a correct analysis. In the TIMESAT algorithms we have considered the start of season the time in which for the following weeks vegetation is boosting and so for further weeks the NDVI is equal or more than 0.5 normally the start (is indicated as the 15% less this relative threshold given of 0. 5 with a progressive and constant increasing trend please see:

Tan, Bin, et al. "An enhanced TIMESAT algorithm for estimating vegetation phenology metrics from MODIS data." IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 4.2 (2010): 361-371.
Stanimirova, Radost, et al. "An empirical assessment of the MODIS land cover dynamics and TIMESAT land surface phenology algorithms." Remote Sensing 11.19 (2019): 2201.

This threshold was adopted not to put the Start but to check if the phenological season is active or not. We have better specified this into the paper and a general figure have been added as request.

This is the explanation “With reference to de-trended/filtered NTS profiles PMs were extracted using the simple relative thresholding approach. The approach assumes that SOS and EOS correspond to the date (DOY) when the NDVI value reach a quote (0.5 for this work) of the range [0.2-MAXVI] during the starting and ending phenological phases, respectively. LOS map was computed by grid differencing from EOS and SOS maps.” (Please see lines 392 – 403)

“Finally, a validation of the PMs retrieved on TIMESAT was performed. To do this, the MODIS product MCD12Q2 in Google Earth Engine was used. It is worth to remind that the MODIS product PMs are obtained starting from EVI (Enhanced Vegetation Index) and from the product GreenUp, Senescence, EVI_Amplitude and other to consider SOS, EOS and LOS respectively. Here it is reported the relative trends between SOS, EOS and LOS (fig. 7)”

(Please see lines 561 – 571)

Author Response File: Author Response.docx

Reviewer 2 Report

Investigating the effects of climate change on rangelands and forests using low-cost tools is an important study, which could have important implications for policy development on future rangeland land and forest management. The topic looks interesting; however, it seems that the authors are unclear on the goals and/or objectives of this research. For example, the topic says they are looking into the effects of climate change on rangelands and forests using satellite data; But the abstract says they are studying the potentials of using remote sensing tools for describing annual phenology of rangelands and forests – so, studying effects of climate change and potentials of remote sensing are two different things. Authors need to stay focused all the way to the conclusions of the manuscript. This manuscript is rather very long and seems that authors kept writing without an appropriate organization of ideas/story.

Abstract comprehension is poor, and sentences or ideas need articulation. Conclusions are overly descriptive – major conclusions or implications should be enlisted with clear take-home messages. It became hard to pick a clear take-home message from conclusions.

The introduction is overly detailed; however, it is not even briefly mentioned what authors mean in their study. What components of climate change they studied should be included in one of the earlier sentences of the first or 2^nd paragraph? This is quite a new experience that about the whole page is one paragraph – readers would feel lost. This paragraph could be broken into 2-3 paragraphs which would be connected with each other with an organized flow of information.

Fig. 4. Graphs are not legible with extremely small fonts, so are the following figures.

Author Response

Response to Reviewer 2 Comments

All comments were carefully evaluated and for the most of them corrections and integrations have been provided.

Point 1: Investigating the effects of climate change on rangelands and forests using low-cost tools is an important study, which could have important implications for policy development on future rangeland land and forest management. The topic looks interesting; however, it seems that the authors are unclear on the goals and/or objectives of this research. For example, the topic says they are looking into the effects of climate change on rangelands and forests using satellite data; But the abstract says they are studying the potentials of using remote sensing tools for describing annual phenology of rangelands and forests – so, studying effects of climate change and potentials of remote sensing are two different things. Authors need to stay focused all the way to the conclusions of the manuscript. This manuscript is rather very long and seems that authors kept writing without an appropriate organization of ideas/story.

The introduction is overly detailed; however, it is not even briefly mentioned what authors mean in their study. What components of climate change they studied should be included in one of the earlier sentences of the first or 2nd paragraph? This is quite a new experience that about the whole page is one paragraph – readers would feel lost. This paragraph could be broken into 2-3 paragraphs which would be connected with each other with an organized flow of information.

Fig. 4. Graphs are not legible with extremely small fonts, so are the following figures.

Response 1: The referee is right. The abstract has been modified in order to avoid misunderstanding and to improve the quality of the work taking into account the suggestions. The manuscript is long because lots of analysis have been done, and reporting each step and also performing a state of art introduction, it seemed crucial for us. Anyway, a clarification as you said, was performed to specify well the research main topic. It concerns both on the study of the effects of short-term climate change on some vegetational components of the Aosta Valley and in minor part (considering the type and topic of journal) the valuable potential that remote sensing offers in monitoring the effects of climate change in particularly sensitive areas such as the Valley of Aosta for free.

Here it is reported the modified abstract: “Satellite remote sensing is a power tool for long-term monitoring of vegetation. This work, with reference to a regional case study, investigates remote sensing potentialities for describing annual phenology of rangelands and broad-leaved forest at landscape level with the aim of detecting eventual effects of climate change in the Alpine region of the Aosta Valley (NW Italy). A first analysis was aimed at estimating phenological metrics (PMs) from satellite images time series and testing the presence of trends along time. A further investigation concerned evapotranspiration from vegetation (ET) and its variation along years. Additionally, in both the cases the following meteorological patterns were considered: air temperature anomalies, precipitation trends and timing of yearly seasonal snow melt. The analysis was based on time series (TS) of different MODIS collections datasets together with CHIRPS collection obtained through Google Earth Engine. Ground weather stations data from the Centro Funzionale VdA ranging from 2000 to 2019 were used. In particular, the MOD13Q1 v.6, MOD16A2 and MOD10A1 v.6 collections were used to derive PMs, ET and Snow Cover maps. The SRTM (Shuttle Radar Topography Mission) DTM (Digital Terrain Model) was also used to describe local topography while the CORINE Land Cover map was adopted to investigate land use classes. PMs and ET and snow cover melting proved to have significantly changed their values in the last 20 years, with a continuous progressive trend. Averagely in the area, rangelands and broad-leaved forests, showed that the length of season is getting longer with a general advance of SOS (Start of the Season) and a delay in EOS (End of the Season). With reference to ET, significant increasing trends were generally observed. Water requirement from vegetation appeared to have averagely risen up of about 0.05 Kg·m^-2 (about 0.5%) per year in the period 2000-2019, for a total increase of about 1 Kg·m^-2 in 20 years (corresponding to a percentage difference in water requirement from vegetation of about 8%). This aspect can be particularly relevant in the bottom of the central valley where the precipitations have shown a statistically significant decreasing trend in the period 2000-2019 (conversely, no significant variation was found in the whole territory). Additionally, the snowpack timing persistence showed a general reduction trend. This fact, together with an increase of air temperature, could be one of the main reasons that explain PMs trends. Results encourage the adoption of remote sensing to monitor climate change effects on alpine vegetation, with particular focus on the relationship between phenology and other abiotic factors, like snowpack timing and temperature, supporting the idea that, presently, an effective technology transfer is now possible to dynamically support management of mountain vegetation. Specifically, remote sensing-derived information can help breeders, agronomists, foresters and also policy-makers to better manage alpine ecosystems by adapting their practices to better face climate change effects.” (Please see lines 13 – 50)

We agree with the reviewer about the conclusion, therefore substantial changes have been performed to the conclusion section as follow:

“Exploring short-term climate change effects on vegetation through pheno-meteorological modeling and relative trends like ET and PMs, snowpack time persistence to the ground derived from satellites EO data, with different time scales provide useful information about biosphere-atmosphere exchanges of carbon, energy, and water at both regional and global scale. This information is crucial to calibrate adaptive strategies to better face climate change effects in alpine ecosystems. It is worth to remind that mountains are among the most sensitive ecosystems to climate change, especially rising temperature, where, effects are faster to be observed than in other terrestrial habitats [99-114] and this study could be considered another one concerning on the short-term climate change effect on rangelands and broad-lived forests phenology in Aosta Valley. In particular, EO data from public and free archives of ready-to-use products like MOD13Q1 and MOD16A2 ones, proved to support well this type of analysis in spite of their reduced geometric resolution. Specifically, they proved to be useful to analyze PMs and ET and SC time trends that, if compared with the air temperature ones, could support breeders, forester and in general policy-makers, to better address management of alpine rangelands and broad-leaved woods. In this work, we showed that in Aosta Valley phenological and evapotranspiration-related processes and snowpack melting time have been dramatically changed in the last decades with general statistically significant trends. PMs, ET and Snow Cover time melting from EO Data proved to significantly changed their values in the last 20 years, with continuous progressive trends. Averagely in the area of study, rangelands and broad-leaved forests, length of season is getting longer with a general advance of SOS (Start of the Season) and a delay in EOS (End of the Season). With reference to ET, significant increasing trends were generally observed like other areas worldwide [114]. Water requirement from vegetation appeared to have averagely risen up of about 0.05 Kg·m-2 (about 0.5%) per year in the period 2000-2019, for a total increase of about 1 Kg·m-2 in 20 years (corresponding to a percentage difference in water requirement from vegetation of about 8%). This aspect can be particularly relevant in the bottom of the central valley where the precipitations have shown a statistically decreasing trend from 2000-2019 (while, no significant variation has been observed in the whole territory). Additionally, the snowpack timing persistence to the soil shows a general reduction trend through the years investigated. This pattern together with an increase in air temperature anomalies could be one of the main reasons that explain the PMs trends observed with a general boosting on the phenological season. Results encourage the adoption of remote sensing to monitor climate change effects on alpine vegetation, with particular focus on the relationship between phenology and other abiotic factors, like snowpack timing and temperature, supporting the idea that, presently, an effective technology transfer is now possible to dynamically support management of mountain vegetation. Therefore, radical changes in rangeland and broadleaved forests management are expected to minimize these effects. Future studies will be intended to explore if the extension of the yearly period of alpine pasture for livestock and the improvement of grasslands irrigation systems at higher altitude in case of water deficit could represent key actions to be considered, also nature-based solution (NBS). A correct knowledge of how some natural component answers to climate variations, is the starting point for a correct development of adaptation plans and strategies on alpine areas to face climate change effects on vegetation and water supply too, like the ones highlighted in this study.” (Please see lines 668 – 726)

Considering the introduction, we have followed the advice of the referee dividing it into paragraph to explain the target of the work. Here it is reported the second one.

“1.1 Goals and summary description

Under these premises, this work was aimed at exploring short-term climate change effects, by remotely sensed data, on phenological metrics (PMs) and evapotraspiration (ET) of rangelands and broad-leaved forests in Aosta Valley, an alpine region located in North-Western Italy. In order to analyze possible climate change effects in this region, particular attention was paid to natural and semi-natural vegetation classes: rangelands (R) and broad-leaved forests (B). Expectation was that climatic factors affecting vegetation phenology were snowpack time persistence (snow cover duration), precipitation and air temperature anomaly. Specifically referring to alpine rangelands and broad-leaved forests, only few papers in literature [10] try to relate the effects of climate change on vegetation PMs and ET with snowpack time duration, air temperature anomalies and precipitation. This is probably due to the fact that ground measures in mountain regions are difficult to be obtained with a proper spatial distribution. Moreover, only in recent years scientific community attention has been focused on mountains, thanks to their role of climate change sentinels. This work, based on freely accessible remotely sensed data, is intended to fill this gap with special concerns about Alps.

It is worth to remind that rangelands have a considerable importance for both ecosystem and economic services of this region, since livestock is one of the main sources of economic income. Moreover, rangelands have a wide spatial distribution over the territory; therefore, monitoring their phenology and physiology can certainly support comprehension of local climate and territorial dynamics. Differently, broad-leaved forests are known to be particularly sensitive to climate change being located in two very sensitive zones: the bottom of the valley (where precipitations and temperature are highly varying) and the higher altitude bands (for e.g. Alnus alnobetula (Ehrh.) K. Koch and Acer pseudoplatanus L.; Fraxinus excelsior L. etc..., where variations in snow cover time persistence and temperature can deeply affect vegetation phenology.

Additionally, together with conifers (and in particular Larix decidua Mill.) they are greatly interesting for the National Inventory of Forests and forest Carbon pools – INFC, and, at regional level, for forestry planning adaptation to climate change perspective.

With these premises and with reference to the CORINE land cover map (see forward on), the analysis was focused on pastures, natural grasslands, moors/heathlands and broad-leaved forests. To explore altitude effects on trends, the analysis was achieved at different height class. PMs/ET investigations were based on TERRA MODIS image time series (MOD13Q1 and MOD16A2 products, see forward on) ranging from 2000 to 2019. With reference to the MOD13Q1 product, a NDVI image time series was generated and used to estimate PMs: SOS (start of season), EOS (end of seasons), LOS (length of season), MAXVI (maximum vegetation index). In order to validate results concerning PMs estimates as computed by TIMESAT tool, the MCD12Q2 phenological collection was adopted.

Air temperature anomalies, snowpack timing and precipitation data were obtained from both EO datasets and ground stations.

With reference to the MOD16A2 product the correspondent ET (computed as yearly cumulated eight days observation per year) image time series was generated and used to describe trends affecting rangelands and broad-leaved forests at different altitudes. Results showed that the most of the above-mentioned metrics significantly changed in the last 20 years in this region, proving that this alpine area is very sensitive to climate change effects. Trends were finally modelled by 1^st order polynomial relationships and some operational information extracted and discussed with the aim of providing a forecasting model useful for rangeland management under the climate change conditions.” (Please see lines 141 - 199)

Finally, considering figure 4 and following the referee is right we have not changed into the paper because of the reason that if the manuscript will be accepted, we will upload the high resolutions graphs that are legible in all their parts and will be adopted by the editor.

Author Response File: Author Response.docx

Reviewer 3 Report

I appreciate the Editor to give me a chance to review an interesting and valuable paper. I found some merits in the both methodology and results. In my opinion, this paper has a good potential to be published in the journal. However, I have also some concerns on the different parts of the manuscript. If the author(s) address carefully to the comments, I’ll recommend publication of the manuscript in the journal:

In the last paragraph of the Introduction, the authors should clearly mention the weakness point of former works (identification of the gaps) and describe the novelties of the current investigation to justify us the paper deserves to be published in this journal.
Cite this recent paper (e.g., line 500) to show the importance of evapotranspiration trend analysis to support the importance of your work:

Complexity of Forces Driving Trend of Reference Evapotranspiration and Signals of Climate Change

Discuss more the variations of the height class (H1, H2, H3) mean values of PMs (SOS, EOS, LOS, MAXVI) as estimated by TIMESAT 3.3 with STL.
How can extend the results in other regions with similar/different climates?
The quality of the language needs to improve by a native English speaker for grammatically style and word use.

Author Response

Response to Reviewer 3 Comments

All comments were carefully evaluated and for the most of them corrections and integrations have been provided.

Point 1: I appreciate the Editor to give me a chance to review an interesting and valuable paper. I found some merits in the both methodology and results. In my opinion, this paper has a good potential to be published in the journal. However, I have also some concerns on the different parts of the manuscript. If the author(s) address carefully to the comments, I’ll recommend publication of the manuscript in the journal:

In the last paragraph of the Introduction, the authors should clearly mention the weakness point of former works (identification of the gaps) and describe the novelties of the current investigation to justify us the paper deserves to be published in this journal.

Cite this recent paper (e.g., line 500) to show the importance of evapotranspiration trend analysis to support the importance of your work:

Complexity of Forces Driving Trend of Reference Evapotranspiration and Signals of Climate Change.

Discuss more the variations of the height class (H1, H2, H3) mean values of PMs (SOS, EOS, LOS, MAXVI) as estimated by TIMESAT 3.3 with STL.

How can extend the results in other regions with similar/different climates?

The quality of the language needs to improve by a native English speaker for grammatically style and word use.

Response 1: Firstly, we would like to thank the referee for his/her kind words and the precious suggestion. As wisely suggested, we have mentioned the weakness point of former works (identification of the gaps) and described the novelties of the current investigation into the Introduction as follow (a new paragraph was added after the general introduction – state of art) in order to better described the aim of the present work.

“1.1 Goals and summary description

Air temperature anomalies, snowpack timing and precipitation data were obtained from both EO datasets and ground stations.

As suggested, to enhance what was stated regarding to the evapotranspiration trend analysis the following citation was added in the references (please see line 676 and 694 respectively) Valipour, Mohammad, et al. "Complexity of forces driving trend of reference evapotranspiration and signals of climate change." Atmosphere 11.10 (2020): 1081. https://doi.org/10.3390/atmos11101081

As suggested by the referee we have more discussed the variations of PMs between height class (H1, H2, H3) as estimated in TIMESAT as follow:

“Concerning on the variation between PMs in H1, H2 and H3 it can be said that: in rangelands at different altitudes, SOS had a more marked variation at high altitudes as can be seen from the gains (in table 4 A) as in the past the snow normally lasted longer, normally at the top than at the bottom of the valley. While as regards EOS there is a more marked gain as it is reasonable to expect at low altitudes than at high altitudes, since the lower temperatures, the frost and the strong temperature variations at altitude (and possible snowfalls) make the vegetative conditions less favorable. It should also be remembered that at low altitudes the rangelands represent vegetation with a marked anthropic conditioning (such as the possibility of irrigation) which does not happen at high altitudes (except in H2 but in the height of summer) which makes vegetation a more exploitable resource. long given by the ease of access given by its topography and micro-climate (at the bottom of the valley). As regards LOS, a more marked gain is observed at high altitudes due to the fact that the shorter duration of the snow and the increase in temperatures favor a more marked lengthening of the season at altitude than in the past (certainly due to the variation of SOS and EOS). Finally, MAXVI has a more marked gain at low altitudes (H1), for the simple fact that the possibility of conditioning the vegetation with irrigation and agronomic practices significantly affects its productivity and relative vigor. In the case of broad-leaved forests in all classes it can be observed a similar gain in SOS, EOS and LOS this is due to the fact that trees are less sensitive and more adapted to environmental variation. Nevertheless, it is interesting to see that in H2 at middle altitude the MAXVI gain is higher that, in H1 and H3, because probably they found more favorable ecological condition, but more studies can be done to find a good answer.”

(Please see lines: 625-644)

The same procedure can be reproduced in other mountains area or planes also with different climate worldwide (in fact the EO datasets obtained are free and have a worldwide coverage, it is only necessary to applied correct geostatistiscs techniques and perform good analysis), considering different land cover or aspects and not only altitude for e.g. starting to the creation of a entropy matrix after a standardisation of the data and the calculation of the weight of more variable considered. A general English revision was done in order to improve the quality of the work done.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Thank you for the response. The resulting ms is even more details - sometimes redundant.

I am not sure if authors care about the overall reader's interest and brevity of the manuscript, as authors have added even more details not needed. I respect that authors want to say everything they did in their research and have in their mind. It looks like more of a graduate thesis than a research article. For example, what is a summary description following goals and hypotheses? Another example, last seven lines of the abstract could be explained in 2-3 lines; but they first provided a generalized conclusion and then specific. This is an abstract, not a discussion section. The same was done in conclusions.

Author Response

Response to Reviewer 2 Comments

All comments were carefully evaluated and for the most of them corrections and integrations have been provided.

Point 1: Thank you for the response. The resulting ms is even more details - sometimes redundant.

Response 1: We agree with the referee, part of the lengthening in the two sections (abstract and conclusions) was explicitly requested by the others referees. In any case, in agreement, we followed the suggestions and shortened them hoping for a compromise as follow. Please see lines 13-47 (abstract) and see lines 640-674 (conclusion).

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Tommaso Orusa et al., in their study provided a quick look at the potentialities of remotely sensed data in describing yearly phenology in Northwest Italy and the effects of climate change onto rangelands. In details, the authors extracted long-term (ca. 20 years) phenological metrics (PMs, e.g. SOS, EOS, LOS, MAXVI) and evapotranspiration (ET) from satellite imageries over the rangelands in Aosta Valley, Italy. The 1^st order regression analysis indicated that both PMs and ETs changed significantly over the past two decades, which further suggested that remote sensing technique could become an useful tool to better manage rangelands and to face and adapt to climate change effects on vegetation in the study area. Overall, this study should become a nice case study in mountain areas, however, it suffers from several major comments were addressed (see my specific comments listed below), from which the main results might be biased and less reliable. A revise major, therefore, is recommended.

Major issues:

First, the abstract is not informative. i) The authors weighed too much on the datasets applied in this study (see lines: 18-24), while the result and implication were quite simple (ca. 4 lines). ii) Lines: 24-28 did not well answer the question raised in the beginning of abstract. For example, the capability of remote sensing technique in monitoring phenological variations (e.g. advanced, delayed or not significant?) and climatic effects on vegetation phenology. Therefore, I would like the authors to simplify the descriptions related to datasets and pay more attention to the key findings of this study.

Second, the method section was not well prepared and some key information needs to be refined. i) The authors claimed that “The analysis was aimed at testing if any significant climatic trend affected phenological metrics (PMs) and ET” (see lines: 233-234). However, how to test such hypothesis was not clearly presented in the following paragraphs nor in the result section. I have noticed that the authors attempted to calculate the trend of PMs using 1^st order polynomial regression, but the coefficient and R² were not adequate to demonstrate how or whether climatic trend would affect the PMs. Moreover, were such trends significant or not? ii) The authors removed the “outliners” in the estimated PMs using the residues defined in Equations 3 and 4, please also give the rationale of such step. Because the outliners (e.g. absolute percent residual > 100%) might be resulted from many factors, e.g. the atmosphere and cloud contamination, the snow/ice coverage, climatic extremes, etc. At least, the authors should tell us how many PMs were excluded and whether such issue could make the estimated PM trends unreliable. iii) How to determine the PMs from the smoothed NDVI curve was not explicitly described in the method section. The authors are suggested to remove lines: 281-293 to the method section, they did not belong to result.

Third, I found the result section was a bit too short, less that one page? If the figures and tables were not counted. I am not very familiar with the instruction of “Climate”, but the result as well as the discussion section should always be the essential part of a success paper. In order to enrich the result section, I recommend the authors to do the following steps in the revision: i) A nice result section should use the statistical values (either from the tables or figures) to support the authors’ statement. To be specific, how did the PMs change over time? Advanced, delayed or not significant? Were these trends differed among different regions or altitudes or biomes? ii) I can hardly find any result related to the effects of climate change on PMs, although the authors mentioned they will do such analysis in the abstract (see lines 13-16).

Forth, the discussion section needs to be improved. i) Lines: 326-336, the authors listed some factors that might influence the variations of PMs, but these findings were not even reflected in the study. In other words, were these statements in line with the findings of this study? ii) Lines: 343-352, temporal trends of PMs should also be mentioned in the result section, remember that this belongs to the main finding of this study.

Minor issues:

-Lines: 47-48, this statement was not precise, because previous literatures on this topic are mounting. For example, Liu et al., 2018, https://doi.org/10.1038/s41467-017-02690-y; Jeong et al., 2011, https://doi.org/10.1111/j.1365-2486.2011.02397.x, etc.

-Lines: 77-78, the less pronounced and spatially heterogeneous trends of leaf senescence is also supported by remote sensing based studies.

-Line: 101, please check the place of reference 35.

-Figure 1, please provide the data source or reference of R1.

-The authors should clearly tell the reader how the land cover data (especially the urban cover) was used in the analysis.

-Figure 4 & Table 5, please specify whether these trends were significant or not?

Author Response

Response to Reviewer 1 Comments

All comments were carefully evaluated and for the most of them corrections and integrations have been provided.

Point 1: First, the abstract is not informative. i) The authors weighed too much on the datasets applied in this study (see lines: 18-24), while the result and implication were quite simple (ca. 4 lines). ii) Lines: 24-28 did not well answer the question raised in the beginning of abstract. For example, the capability of remote sensing technique in monitoring phenological variations (e.g. advanced, delayed or not significant?) and climatic effects on vegetation phenology. Therefore, I would like the authors to simplify the descriptions related to datasets and pay more attention to the key findings of this study.

Response 1: According to reviewer’s suggestions we improved the abstract as it follows: “Satellite remote sensing is a power tool for long-term monitoring of vegetation. This work, with reference to a regional case study, investigates remote sensing potentialities for describing annual phenology of rangeland vegetation at landscape level with the aim of detecting eventual effects of climate change in the Alpine region of the Aosta Valley (NW Italy). A first analysis was aimed at estimating phenological metrics (PMs) from satellite images time series and testing the presence of trends along time. A further investigation concerned evapotranspiration (ET) and its variation along years. The analysis was entirely based on TERRA MODIS image time series (TS) ranging from 2000 to 2019 and available for free as ready-to-use products. In particular, MOD13Q1 v.6 and MOD16A2 products from MODIS collections were used to derive PMs and ET trends, respectively. Additionally, the SRTM (Shuttle Radar Topography Mission) DTM (Digital Terrain Model) was used to describe local topography; CORINE Land Cover map was adopted to describe land use classes. PMs and ET (as derivable from EO data) proved to significantly changed their values in the last 20 years, with a continuous progressive trend. Averagely in the area, rangeland length of season is getting longer with a general advance of SOS (Start of the Season) and a delay in EOS (End of the Season). With reference to ET, a significant increasing trend was generally observed. Water requirement from rangeland appeared to have averagely increased of about 0.05 Kg·m^-2 (about 0.5%) per year in the period 2000-2019, for a total increase of about 1 Kg·m^-2 in 20 years (corresponding to a percentage difference in water requirement from vegetation of about 8%). These results encourage the adoption of remote sensing to monitor climate change effects on alpine rangelands supporting the idea that, presently, an effective technology transfer is now possible to dynamically support management of mountain vegetation. Specifically, remote sensing techniques and the obtainable information can help breeders to better manage rangelands by adapting their practices to better face climate change effects..” (please see lines 15-47)

Point 2: Second, the method section was not well prepared and some key information needs to be refined. i) The authors claimed that “The analysis was aimed at testing if any significant climatic trend affected phenological metrics (PMs) and ET” (see lines: 233-234). However, how to test such hypothesis was not clearly presented in the following paragraphs nor in the result section. I have noticed that the authors attempted to calculate the trend of PMs using 1st order polynomial regression, but the coefficient and R2 were not adequate to demonstrate how or whether climatic trend would affect the PMs. Moreover, were such trends significant or not? ii) The authors removed the “outliers” in the estimated PMs using the residues defined in Equations 3 and 4, please also give the rationale of such step. Because the outliners (e.g. absolute percent residual > 100%) might be resulted from many factors, e.g. the atmosphere and cloud contamination, the snow/ice coverage, climatic extremes, etc. At least, the authors should tell us how many PMs were excluded and whether such issue could make the estimated PM trends unreliable. iii) How to determine the PMs from the smoothed NDVI curve was not explicitly described in the method section. The authors are suggested to remove lines: 281-293 to the method section, they did not belong to result.

Response 2: Some revisions were done in order to better focus the point of significance of trends. As far as i) comment is concerned p-values of trends, that were available, but not included in the submitted paper, are now explicitly reported in table 4 B of the Results section.

With respect to ii) comment in MM section we add the following paragraph: “Data showing an absolute percent residual (e, eq. 3) > 100% were labelled as outliers and removed. A new model calibration was consequently run after removal.

please see eq3 into the paper

(3)

Scatterplots are reported in Fig 4; model coefficients and coefficient of determination (R2) and p-values (calculated by PAST software [67]) are reported in tables 4A and 4B. (please see lines 347-357 and also please see lines 301-351)

iii) Referee’s suggestion was right. Lines 281-293 were removed and included into the methodology section as follow in order to better explain as asked by the reviewer “To investigate changes affecting vegetation activity in consequence of climate change, some metrics were extracted from the above-mentioned time series (NTS and ETS). Phenological metrics (PMs) are numerical indices of vegetation activity corresponding to some key points identified along its annual growing season. Climate change proved to condition such activity shifting and reshaping past “ordinary” behavior of plants along the year. The following PMs were considered: the start of the growing season (SOS) representing the day of the year (DOY) when phenology is admitted to boost; the end of the season representing the day of the year (DOY) when phenology is admitted to stop; the length of the growing season (LOS) representing the time range (in number of days) separating EOS from SOS; the maximum of NDVI (MAXVI) representing the highest value reached by NDVI during the growing season and proved to be a good predictor of climate change effects on vegetation.

PMs were estimated by TIMESAT 3.3 with STL software [64-67] that was specifically developed to enable the monitoring of land surface processes by remotely sensed data. TIMESAT 3.3 with STL [68,69,70] iteratively fits and smooths by mathematical functions the yearly NDVI time-series, finding the best smoothed approximation of the NDVI along the year at pixel level. Once raw data have been approximated by the selected fitting function, PMs can be extracted in correspondence of singular points having a phenological meaning (e.g. EOS, SOS, LOS, MAXVI, etc.) along the local temporal profile. In this work, phenological metrics (SOS, EOS, LOS, MAXVI) were estimated at pixel level by TIMESAT 3.3 software tool processing the whole NTS. A Seasonal Trend decomposition by Loess (STL) was adopted to de-trend NTS pixel profile and removing noise. Seasonal component was refined by Savitzky-Golay filtering to reduce, but not removing, remaining local strong variations. The yearly growing season was recognized with the whole multi annual time series using the sinusoidal harmonics approach.

With reference to de-trended/filtered NTS profiles PMs were extracted using the simple thresholding approach. An arbitrary value of 0.5 was set as reference value to refer SOS and EOS to. Consequently, SOS was located at the date (DOY) when NDVI reached the 0.5 threshold value along the ascending part of the phenological yearly curve; EOS was located at the date (DOY) when NDVI reached the 0.5 threshold value along the descending part of the phenological yearly curve. LOS map was computed by grid differencing from EOS and SOS maps. MAXVI was found looking for the highest NDVI value (yearly averaged) observed between SOS and EOS.

With reference to ETS and separately for H1, H2, H3, the class yearly average ET value was computed and plotted along the years (2000-2019) at regional level, to look for trends.

The analysis was performed at pixel and year level. Consequently, estimates of all PMs were represented as raster layers mapping their spatial distribution at the considered year.

Considering both PMs and ET estimates, a first investigation was aimed at removing outliers from data. This was obtained by calibrating a 1^st order polynomial model relating PMs/ET with the years and testing the residuals (model predicted – observed).” (please see lines 305-344)

Point 3: Third, I found the result section was a bit too short, less that one page? If the figures and tables were not counted. I am not very familiar with the instruction of “Climate”, but the result as well as the discussion section should always be the essential part of a success paper. In order to enrich the result section, I recommend the authors to do the following steps in the revision: i) A nice result section should use the statistical values (either from the tables or figures) to support the authors’ statement. To be specific, how did the PMs change over time? Advanced, delayed or not significant? Were these trends differed among different regions or altitudes or biomes? ii) I can hardly find any result related to the effects of climate change on PMs, although the authors mentioned they will do such analysis in the abstract (see lines 13-16).

Response 3: The referee is right. The Results section was deeply revised accordingly

(please see lines 373-486)

Point 4: Forth, the discussion section needs to be improved. i) Lines: 326-336, the authors listed some factors that might influence the variations of PMs, but these findings were not even reflected in the study. In other words, were these statements in line with the findings of this study? ii) Lines: 343-352, temporal trends of PMs should also be mentioned in the result section, remember that this belongs to the main finding of this study.

Response 4: The referee is right. Revisions operated for MM and Results sections now can well support statements in the Discussion section that was largely revised.

(Please see lines 487-566)

Point 5: Lines: 47-48, this statement was not precise, because previous literatures on this topic are mounting. For example, Liu et al., 2018, https://doi.org/10.1038/s41467-017-02690-y; Jeong et al., 2011, https://doi.org/10.1111/j.1365-2486.2011.02397.x, etc.

Response 5: The referee is right. The following correction was done. “In the last years, many works were published. Presently, only few works in literature presenting approaches or case studies about this topic making desirable a with a significant improvement [2-4].” (Please see lines 66-67)

Point 6: Lines: 77-78, the less pronounced and spatially heterogeneous trends of leaf senescence is also supported by remote sensing based studies.

Response 6: The referee is right. The following adjustment has been done. “In contrast, observed trends towards delayed leaf senescence are less pronounced and are spatially heterogeneous also supported by remote sensing based studies [23,24].” (Please see lines 84-85)

Point 7: Line: 101, please check the place of reference 35.

Response 7: Thank you, the reference is correct.

Point 8: Figure 1, please provide the data source or reference of R1.

Response 8: The referee is right. Here it is what it has been done. “Figure 1. The study area corresponds to the whole Aosta Valley. It is located in the Northern West Alps of Italy, close to France and Switzerland. Source: Tommaso Orusa.” (Please see lines 172-173)

Point 9: The authors should clearly tell the reader how the land cover data (especially the urban cover) was used in the analysis.

Response 9: The CLC (Corine Land Cover) was used to locate rangelands in Aosta Valley. Other classes, like urban, were not considered. As indicated as it follow “As indicated in table 2 (green), this study focused on rangelands, that was assumed to correspond to the following CORINE land cover classes: pastures, natural grassland, moors and heathlands (Fig. 2).” (Please see lines 274-275).

Point 10: Figure 4 & Table 5, please specify whether these trends were significant or not?

Response 10: The referee is right. All trends wer found to be significant. Table 4B was added in the paper and fig. 4 caption revised as it follows: “Figure 4. (a,b,c) Height class (H1,H2 and H3) mean values of PMs (SOS, EOS, LOS, MAXVI) as estimated by TIMESAT 3.3 with STL. (d) Height class (H1, H2 and H3) yearly mean values of ET. Reported values correspond to the yearly mean value of the 8 days aggregated data from MOD16A2 product. In all the graphs orange points represent outliers (see equation 2). All the trends proved to be significant (see table 4B). (Please see lines 405-427).

Author Response File: Author Response.docx

Reviewer 2 Report

The manuscript “Long-term bio-meteorological Trends Analysis of 2 Rangelands by Remote Sensing: the Aosta Valley (NW 3 Italy) Case Study” reports the same content of the manuscript already published in another journal by the same authors together with colleagues from veterinary field:
Orusa, T.; Orusa, R.; Viani, A.; Carella, E.; Borgogno Mondino, E. Geomatics and EO Data to Support Wildlife Diseases Assessment at Landscape Level: A Pilot Experience to Map Infectious Keratoconjunctivitis in Chamois and Phenological Trends in Aosta Valley (NW Italy). Remote Sens. 2020, 12, 3542.
https://doi.org/10.3390/rs12213542
In this already published paper, the same methodology was applied and the same results were obtained, and further, they were applied to a veterinary case of Infectious Keratoconjunctivitis in Chamois, leading to interesting results for the scientific and veterinary community.
Comparing the two articles, in the current submission, only part of the introduction and part of the discussion differ from previous one, but not enough to justify another publication.

Author Response

Response to Reviewer 2 Comments

All comments were carefully evaluated and for the most of them corrections and integrations have been provided.

Point 1: The manuscript “Long-term bio-meteorological Trends Analysis of 2 Rangelands by Remote Sensing: the Aosta Valley (NW 3 Italy) Case Study” reports the same content of the manuscript already published in another journal by the same authors together with colleagues from veterinary field:
Orusa, T.; Orusa, R.; Viani, A.; Carella, E.; Borgogno Mondino, E. Geomatics and EO Data to Support Wildlife Diseases Assessment at Landscape Level: A Pilot Experience to Map Infectious Keratoconjunctivitis in Chamois and Phenological Trends in Aosta Valley (NW Italy). Remote Sens. 2020, 12, 3542.
https://doi.org/10.3390/rs12213542
In this already published paper, the same methodology was applied and the same results were obtained, and further, they were applied to a veterinary case of Infectious Keratoconjunctivitis in Chamois, leading to interesting results for the scientific and veterinary community.
Comparing the two articles, in the current submission, only part of the introduction and part of the discussion differ from previous one, but not enough to justify another publication.

Response 1: The work here presented is actually a peculiar declination and re-interpretation of results obtained for another paper. The study area and part of methods are the same. Nevertheless, in the previous paper (cited by the referee) the focus was on veterinary aspects of the problem (as the reviewer correctly reported); the main topic was to test relationships between environmental parameters and infectious keratoconjunctivitis. Time trends analysis was marginal and not deeply commented nor investigated. The paper we submitted to Climate starts from those results and methods to read climatic indicators that a multitemporal analysis of vegetation by satellite can generate. In other words, in the submitted paper, we focused on the effects of climate and climate change on vegetation in an Alpine region that is known to be very sensitive to it. Presented results are different, but probably we missed to emphasize this properly and so a major revison has been apported (we invite the referee to see it). Finally, English language and style has been improved.

Author Response File: Author Response.docx

Reviewer 3 Report

The authors presented a study on the use of remote sensing data sources to derive phenology metrics and evapotranspiration over the Aosta Valley for the period 2000-2019. The period corresponds to the operational life of MODIS sensor onboard the Terra satellite. The study focused on land cover classes that can be considered Rangelands.

In its present status, the study has important weaknesses which make it not recommendable for publication. Some comments and questions are presented in the following paragraphs to justify this opinion and to encourage authors to address certain aspects.

I would suggest to rethink the best title given to the article in order to give better hints to potential readers on the actual content of the study. The reference to ‘Long-term’ data in a journal devoted to Climate studies might create the expectations of analyses on 30+ years time series, as that is the convention to obtain climate normals (https://library.wmo.int/doc_num.php?explnum_id=4166). Likewise, the term ‘bio-meteorological’ can encompass many more variables than those addressed in this article. Of course, this is a mere suggestion.
The study is largely based on the NDVI and ET products derived from MODIS observations. The native spatial resolution of these products is increased to 30 m. This is a big leap which may have serious impact on the results. How did you do that? Was the SRTM product somehow used for that or was this product only used to identify the H1, H2 and H3 zones? According to the text in Lines 199-200, the ET and NDVI were “oversampled” to match a 30 m grid. Does it mean that the native MODIS grid was just cut into smaller pieces without altering the data? How big is the mixing of CORINE land cover units within the native MODIS cell?
As explained in the Materials and Methods section, the MODIS, the CORINE LC and the SRTM data were obtained from Google EarthEngine. All those datasets can be overlaid because they all follow the same coordinate system. Why were the data reprojected to UTM 32 N? They were already spatially compatible; why altering the data again? This only adds uncertainty to the data.
In lines 191-193 you wrote: “Pixel values for the Net Evapotranspiration (ETo ) is the sum of all 8 days within the composite period expressed in kg m^-2 8d^-1” What do you mean by ‘Net Evapotranspiration’? Is it Actual Evapotranspiration? You introduce here the abbreviation ETo (conventionally used for 'Reference' ET) which (unless I missed it) is thereafter not mentioned in the document.
In the course of the studied period (19 years) there most be many missing data in the MODIS products. What was the policy to account for missing data?
I am not sure about the adequacy of the approach to detect (and then eliminate) outliers. How to distinguish between an ‘outlier’ and an anomaly? I dry years, for instance, ET can drop considerably with respect to an average of linear regression model. If data have been processed with the Savitsky-Golay filter, implemented in TIMESAT, what can be the origin of those outliers?
Concerning the plots of Figure 4. If I understood well the LOS plot was derived by substracting the SOS values from EOS values. This seems to hold in the presented plots. However, I doubt about the values for H2 in 2000 (labeled as outlier, by the way). In that year, by visually inspecting the plots, the EOS was ~260 and the SOS was ~135; the LOS should be something around 260-135=125. The plotted LOS value was more than 2x that figure. Is this an error?
Also in the plots of Figure 4. How did you actually processed the ET values. The units presented in the plot correspond to an 8-days period. What do those values mean? The yearly average? The annual maximum? In lines 193:194 you wrote: ‘ET layers were stacked into an ET time series (ETS); no filtering was applied’. This is not really helpful to understand the processing of the 8-day ET layers.
I found a bit strange to present Equation 2 where the studied variables are predicted with a linear equation where the independent variable is the ‘Year’ (M = Gain ∙ Y + Offset). This can be confusing for readers as the ‘Year’ is a number that can be set arbitrarily (for instance, the first year of our study period can be set to 1. Or to 0.) and does not represent a physically based causal factor of the modeled variable. Readers should at least know what does ‘Y’ actually mean in this equation and that it is constrained to the values between 2000-2019.
The Conclusions section contains information that does not refer to the specific findings of this study. In fact, the same text could have been written without performing the study.

Author Response

Response to Reviewer 3 Comments

All comments were carefully evaluated and for the most of them corrections and integrations have been provided.

Point 1: I would suggest to rethink the best title given to the article in order to give better hints to potential readers on the actual content of the study. The reference to ‘Long-term’ data in a journal devoted to Climate studies might create the expectations of analyses on 30+ years time series, as that is the convention to obtain climate normals (https://library.wmo.int/doc_num.php?explnum_id=4166). Likewise, the term ‘bio-meteorological’ can encompass many more variables than those addressed in this article. Of course, this is a mere suggestion.

Response 1: The referee is right in order to avoid misunderstanding the title has been changed as follow: “Exploring Short-Term Climate Change Effects on Rangelands by Free Satellite Imagery in Aosta Valley (NW Italy) (Please see lines 1-7). Also, extensive editing of English language and style was made in the paper as suggested.

Point 2: The study is largely based on the NDVI and ET products derived from MODIS observations. The native spatial resolution of these products is increased to 30 m. This is a big leap which may have serious impact on the results. How did you do that? Was the SRTM product somehow used for that or was this product only used to identify the H1, H2 and H3 zones? According to the text in Lines 199-200, the ET and NDVI were “oversampled” to match a 30 m grid. Does it mean that the native MODIS grid was just cut into smaller pieces without altering the data? How big is the mixing of CORINE land cover units within the native MODIS cell?

Response 2: The MODIS native product have maintained their native resolution. Anyway to catch also the rangelands that have an extension minor to a MODIS pixel in that case 250 m (a very few number in the study area), they were oversampled by applying the method of Bicubic Spline Interpolation and the SRTM was used only to identify H1, H2 and H3 as the referee correctly understood. The mixing between CORINE and native cell is good but we decided to perform this oversample in order to work with datasets that geometrically were equal. We hope to give you a good answer. In order to better explain this adjustment has been adopted into the paper “MOD13Q1 v.6 and MOD16A2 v.6 layers were geometrically oversampled up to 30 m to refine area zonation by applying Bicubic Spline Interpolation method in order to have grids with the same spatial resolution.” (Please see line 221-224)

Point 3: As explained in the Materials and Methods section, the MODIS, the CORINE LC and the SRTM data were obtained from Google Earth Engine. All those datasets can be overlaid because they all follow the same coordinate system. Why were the data reprojected to UTM 32 N? They were already spatially compatible; why altering the data again? This only adds uncertainty to the data.

Response 3: The referee is right. Anyway, the choice to reprojected to ED50 UTM 32 N

was dictated by the fact that the research data will be used by various regional bodies of the Aosta Valley to strengthen the policies of mitigation and adaptation to climate change. It is worth to remind that these entities work with that system (it is the official used into Aosta Valley Autonomous region) and asked us at the writing stage if we could adopt that projection.

Point 4: In lines 191-193 you wrote: “Pixel values for the Net Evapotranspiration (ETo) is the sum of all 8 days within the composite period expressed in kg m -2 8d -1” What do you mean by ‘Net Evapotranspiration’? Is it Actual Evapotranspiration? You introduce here the abbreviation ETo (conventionally used for 'Reference' ET) which (unless I missed it) is thereafter not mentioned in the document.

Response 4: The referee is right is also known as Actual Evapotraspiration. For further information please take a look at:

https://lpdaac.usgs.gov/documents/93/MOD16_ATBD.pdf
https://lpdaac.usgs.gov/documents/494/MOD16_User_Guide_V6.pdf
https://developers.google.com/earth-engine/datasets/catalog/MODIS_006_MOD16A2#description

ETo was a mistake (thanks for your suggestion) we have put ET.

Point 5: In the course of the studied period (19 years) there most be many missing data in the MODIS products. What was the policy to account for missing data?

Response 5: An interpolation has been performed in R by adopting the packages imputeTS (https://cran.r-project.org/web/packages/imputeTS/imputeTS.pdf) other similar tools (https://mgimond.github.io/Spatial/interpolation-in-r.html and https://rdrr.io/cran/forecast/man/na.interp.html).

Anyway, it is worth to remind that in many cases Google Earth Engine give the possibility to benefit of full dataset (they do interpolation on their own by using Neural Network from TensorFlow or other methods. It depends on the type of Dataset).

Point 6: I am not sure about the adequacy of the approach to detect (and then eliminate) outliers. How to distinguish between an ‘outlier’ and an anomaly? I dry years, for instance, ET can drop considerably with respect to an average of linear regression model. If data have been processed with the Savitsky-Golay filter, implemented in TIMESAT, what can be the origin of those outliers?

Response 6: Only PMs was filtered and processed in TIMESAT not ET: In this case PMs was firstly filtered and then eq 2 was adopted: The anomaly can persist in Timesat by the threshold assigned during the filtering process. In case of ET only eq.2 was adopted but anomalies have been considered because by using eq.2 the threshold assigned was three times higher to the medium standard deviation of the reference period. Therefore, outliers can be detect and not confused with natural anomalies.

Point 7: Concerning the plots of Figure 4. If I understood well the LOS plot was derived by substracting the SOS values from EOS values. This seems to hold in the presented plots. However, I doubt about the values for H2 in 2000 (labeled as outlier, by the way). In that year, by visually inspecting the plots, the EOS was ~260 and the SOS was ~135; the LOS should be something around 260-135=125. The plotted LOS value was more than 2x that figure. Is this an error?

Response 7: The referee is absolutely right. Thanks for your notice it was a mistake the correction was made and a correct plot in now available. (Please see line 405)

Point 8: Also in the plots of Figure 4. How did you actually processed the ET values. The units presented in the plot correspond to an 8-days period. What do those values mean? The yearly average? The annual maximum? In lines 193:194 you wrote: ‘ET layers were stacked into an ET time series (ETS); no filtering was applied’. This is not really helpful to understand the processing of the 8-day ET layers.

Response 8: It is the yearly average and it is explained into Fig. 4 caption. (Please see lines 406-412)

Point 9: I found a bit strange to present Equation 2 where the studied variables are predicted with a linear equation where the independent variable is the ‘Year’ (M = Gain ∙ Y + Offset). This can be confusing for readers as the ‘Year’ is a number that can be set arbitrarily (for instance, the first year of our study period can be set to 1. Or to 0.) and does not represent a physically based causal factor of the modeled variable. Readers should at least know what does ‘Y’ actually mean in this equation and that it is constrained to the values between 2000-2019.

Response 9: The referee suggestion has been accepted and so this part was better explained as follow:”[…] where M is the generic PM/ET yearly averaged metric and Y the considered year (in this case of study the period between 2000-2019).” (Please see lines 345-346)

Point 10: The Conclusions section contains information that does not refer to the specific findings of this study. In fact, the same text could have been written without performing the study.

Response 10: The referee is right. Therefore, the conclusion was rewritten in order to

to be adhering to what has been demonstrated in the work as follow:

“Pheno-meteorological modeling and relative trends like ET and PMs observations derived from satellites EO data, with different time scales provide useful information about biosphere-atmosphere exchanges of carbon, energy, and water at both regional and global scale. This information is crucial to calibrate adaptive strategies to better face climate change effects in alpine ecosystems. It is worth to remind that mountains are among the most sensitive ecosystems to climate change, especially rising temperature, where, effects are faster to be observed than in other terrestrial habitats [95-106]. In this work, authors showed that in Aosta Valley phenological and evapotranspiration-related processes have been changing in the last 20 years with statistically significant trends. Climate change effects may represent an important threat to mountain ecosystems and related services for local populations. Their effects on “green” (vegetation-related) water resources can have important relapses on mountains agriculture and livestock productions (such as mountains pasture activities). Therefore, radical changes in rangeland management are expected to minimize these effects. Future studies will be intended to explore if the extension of the yearly period of alpine pasture for livestock and the improvement of grasslands irrigation systems at higher altitude in case of water deficit could represent key actions to be considered. EO data from public and free archives of ready-to-use products like MOD13Q1 and MOD16A2 ones, proved to support well this type of analysis in spite of their reduced geometric resolution. Specifically, they proved to be useful to analyze PMs and ET time trends that, if compared with the air temperature ones, could support breeders, and in general policy-makers, to better address management of alpine rangelands. A correct knowledge of these patterns is the starting point for a correct development of adaptation plans and strategies on alpine areas to face climate change effects on vegetation, like the ones highlighted in this study.” (Please see lines 567-603)

Author Response File: Author Response.docx

Reviewer 4 Report

The authors analyzed phenological metrics and evapotranspiration trends of vegetation in the Aosta Valley. The authors seek to employ this method to better understand the climate change influence in Aosta Valley, which I found useful. The topic of the article is appropriate for this journal. However, I can see some problematic issues that exist in this paper.

The method for deriving phenological metrics is too simple. The fixed 0.5 threshold can cause unignorably uncertainty for phenological metrics estimation. The authors can try the percentage of amplitude as the threshold or directly use the MODIS phenology product (MCD12Q2).

The results section is very weak. The interpretation and analysis of the results need to be rewritten. Only three paragraphs were presented while two paragraphs describe the methodology (L281-L293) in the results section.

The discussion section needs to be rewritten. A large part of the discussion didn’t base on the results derived from this study (L326-L336, L356-L363).

The conclusion section needs to be rewritten. Only two sentences of the conclusion can be supported by this study (L396-L399), while other parts come from no bases (L387-L395, L400-406).

Language editing would be required for this manuscript.

The details were listed as follows:

L102-L104. Higher spatial resolution images show denser images time series can be used to improve phenological deduction? Revise.
L197. Re-projection?
L348. How large of the MAXVI variation?
L350. How about the accumulated uncertainty?
L355. Reference
L368. ET does not mean water requirement.

Author Response

Response to Reviewer 4 Comments

All comments were carefully evaluated and for the most of them corrections and integrations have been provided.

Point 1: The method for deriving phenological metrics is too simple. The fixed 0.5 threshold can cause unignorably uncertainty for phenological metrics estimation. The authors can try the percentage of amplitude as the threshold or directly use the MODIS phenology product (MCD12Q2).

Response 1: We thanks the referee for his/her suggestion, but we do not agree. The threshold was defined on the basis of previous experiences and to verify the quality of the metrics obtained they were validated comparing with the MODIS product (MODIS phenology product (MCD12Q2). It wasn’t explained into the paper properly therefore it was better explained.

“In order to test the quality of the PMs calculated from the NDVI of MOD13Q1 v.6 collection, the MCD12Q2 v.6 Land Cover Dynamics product (informally called the MODIS Global Vegetation Phenology product) was adopted to perform a validation. This dataset was downloaded from Google Earth Engine with a 500 m GSD covering the period 1^stJanuary 2001 - 31 December 2018. Missing values were interpolated or extrapolated for a maximum of one year before or after the reference time period of the collection by adopting the functions ee.interpolate and ee.Model.fromAiPlatformPredictor available in Google Earth Engine libraries. The dataset provides estimates of the timing of vegetation phenology at global scales. Additionally, it provides information related to the range and summation of the enhanced vegetation index (EVI) computed from MODIS surface reflectance data at each pixel. It identifies the onset of greenness, green-up midpoint, maturity, peak greenness, senescence, green-down midpoint, dormancy, EVI2 minimum, EVI2 amplitude, integrated EVI2 over a vegetation cycle, as well as overall and phenology metric-specific quality information. The MCD12Q2 v.6 data product is derived from time series of the 2-band Enhanced Vegetation Index (EVI2) calculated from MODIS Nadir Bidirectional Reflectance Distribution Function (BRDF)-Adjusted Reflectance (NBAR). Vegetation phenology metrics are identified for up to two detected growing cycles per year. For pixels with more than two valid vegetation cycles, the data represent the two cycles with the largest NBAR-EVI2 amplitudes [62].

All data were converted from native geographical reference systems WGS84 into the ED50 UTM 32N one.” (Please see lines 226-245)

“In order to test the quality of the PMs obtained they were validated by comparing those obtained from MODIS phenology product MCD12Q2 with an overall discrepancy between them less than 5%.” (Please see lines 334-336)

In order to support our methodology here it is some works in which it was discussed.

Zeng, Linglin, et al. "A review of vegetation phenological metrics extraction using time-series, multispectral satellite data." Remote Sensing of Environment 237 (2020): 111511.
Atzberger, Clement, et al. "Phenological metrics derived over the European continent from NDVI3g data and MODIS time series." Remote Sensing 6.1 (2014): 257-284.
Huang, Xin, et al. "The optimal threshold and vegetation index time series for retrieving crop phenology based on a modified dynamic threshold method." Remote Sensing 11.23 (2019): 2725.
White, Katharine, Jennifer Pontius, and Paul Schaberg. "Remote sensing of spring phenology in northeastern forests: A comparison of methods, field metrics and sources of uncertainty." Remote Sensing of Environment 148 (2014): 97-107.
Fontana, Fabio, et al. "Alpine grassland phenology as seen in AVHRR, VEGETATION, and MODIS NDVI time series-a comparison with in situ measurements." Sensors 8.4 (2008): 2833-2853.

Point 2: The results section is very weak. The interpretation and analysis of the results need to be rewritten. Only three paragraphs were presented while two paragraphs describe the methodology (L281-L293) in the results section.

Response 2: The referee is right. Taking into account his/her suggestion the results has been rewritten as follow: “PMs were mapped over the area as raster layers and spatially averaged with respect to H1, H2 and H3 classes (table 3) by ordinary zonal statistics.

Graphs of figures 4a, 4b, 4c show that average PMs values of H1, H2, H3 significantly changed their values in the last 20 years, with a continuous progressive trend observable for all of them. First order polynomials used to model trends where calibrated after outlier removal. Coefficients are reported in table 4 together with correspondent coefficients of determination (R2) and p-value. To operationally synthesize information about strength of changes, in table 5 yearly rates of increment/decrement of PMs/ET are reported together with total increment/decrement rate in the period 2000-2019, as derived from the calibrated models. According to eq. 3, the number of outliers was found to be lower than 4 (orange points in Fig.4) for all considered parameters (PMs and ET). After outliers removal PMs and ET trends turned to be statistically significant.

Obtained results showed that all PMs and ET significantly changed in the reference period for rangelands in Aosta Valley. In particular, a progressive linear enlargement of rangelands growing season was observed for all the considered altitude classes (H1, H2, H3).

The following interpretation of results can be given. The start of the growing season (SOS) significantly anticipated every year during the entire considered period (2000-2019). Yearly anticipation of SOS was estimated in about 2-3 days per year depending on the altitude class (table 5). Conversely, the end of the season (EOS), significantly delayed every year during the entire considered period (2000-2019). Yearly delay of EOS was estimated in about 2.5-3 days per year depending on the altitude class (table 5). Consequently, the length of the growing season (LOS) significantly increased every year during the entire considered period (2000-2019). Yearly enlargement of LOS was estimated in about 5-6 days per year depending on the altitude class (table 5). Similarly, MAXVI showed an increasing trend too, but yearly difference value was found to be not significant (see Discussion section).

Evapotraspiration (ET) significantly increased every year during the entire considered period (2000-2019). Yearly increase of ET was estimated in 0.06 Kg m^-2 (H1), 0.04 Kg m^-2 (H2) and 0.06 Kg m^-2 (H3) per year, respectively (table 5). In table 5 cumulated differences computed for the whole reference period are also reported. With respect to Aosta Valley it was observed that between 2000 and 2019 rangelands SOS anticipated of almost 2 months, EOS delayed of about 1.5 month and LOS stretched of about three months for all the considered altitude classes. A cumulated significant increase was observed for both MAXVI and ET, as well. One of the possible factors that can be responsible of this important variation can be reasonably found in the increasing of air temperature that affected both snowpack duration and vegetation phenology (whereas water is not a limiting factor). This was confirmed by the analysis of the annual anomaly of the average yearly air temperature in the period 2000-2019. Air temperature values were obtained from the automatic weather stations and aggregated with respect to the considered altitude classes (H1, H2 and H3). Air temperature class anomalies, showed an increasing and statistically significant trend, whose gain increase with the altitude class (Fig.5).” (Please see lines 373-486)

Point 3: The discussion section needs to be rewritten. A large part of the discussion didn’t base on the results derived from this study (L326-L336, L356-L363).

Response 3: The referee is right. The analysis section has been expanded with the intention of aligning with what is present in the discussion (aspects that previously, as rightly highlighted by the referee, had not been correctly highlighted in this study) as follow: “PMs and ET modeling and trend analysis based on EO data, with particular concern about multispectral passive remote sensing, proved to provide useful information at seasonal, inter-annual and annual time scales. This can be proficiently used to describe phenomena related to biosphere-atmosphere exchanges of carbon, energy, and water at both regional and global scales. Mountain environments are known to be very sensitive to climate change, in particular to temperature increasing [24]. Similarly, to other alpine areas, Aosta Valley is suffering from an important increase of temperature as a result of climate change, that appears to determine significant changes in the growing season of rangelands. Air temperature increasing depends on altitude. Results showed that, in the analyzed period, the average increase of the yearly average temperature was lower than 2°C, lower than 2.5°C and higher than 3°C within altitude classes H1, H2 and H3, respectively. These evidences can be directly related to the described trends of the considered phenological metrics and evapotranspiration. High altitudes are, in fact, more sensitive to rising temperatures that determine a reduced persistence of snowpack and a lengthening of the growing season. For this reason, they appear among the most severely and rapidly impacted ecosystems, being possibly impacted by any change in temperature and precipitation patterns at all scales [27]. Snow and ice are the main control parameters of the hydrological cycle, particularly of the seasonal runoff, and their variations can condition the entire geosystem (rocks, soils, vegetation, and river discharges). With climate change, water will probably become less available due to global warming and consequences will reach far beyond mountain regions [16]. Similarly, climate change is likely to increase exposure to either natural or economic hazards, all the more so because in many mountain areas, low incomes levels are higher than in lowland areas [7].

Results from this work demonstrate that phenological and evapotranspiration processes are significantly changing in the study area and that these effects are linked to rising temperatures. Changes can be modelled by linear trends confirming what other similar works already found for different areas [12-18].

In particular, in Aosta Valley, PMs and ET proved to have significantly changed their values in the last 20 years, with a continuous progressive trend observable for all of them.

As previously discussed in the analysis section, it is worth to remind that in terms of strength of changes, an average delay of EOS was observed of about 2.6 days, independently from the altitude class. Differently, SOS appears to anticipate of about 2 d/y up to 2000 m a.s.l. (H1 and H2) and of about 3 d/y at higher altitudes (H3). Consequently, LOS proved to enlarge of about 4.7 d/y at lower altitudes (H1 and H2) and of about 6.5 d/y at higher altitudes (H3). According to Borgogno et al. [80], potential accuracy of NDVI measurements is about ±0.02; consequently, estimated yearly variations of MAXVI cannot be considered singularly significant. Nevertheless, the cumulated effects along the entire explored period (2000-2019) showed that MAXVI significantly changed, since the accuracy reference value of ±0.02 was largely overcome. MAXVI variations between 2000 and 2019 appear to be positively higher at lower altitudes (about +0.09) while almost stable as altitude increases. This could be possibly explained with reference to the previously demonstrated increasing of both biomass production (MAXVI) and enlargement of the growing season, that, consequently make vegetation needing more water yearly. This can be explained admitting that at lower altitudes, in Aosta Valley, grasslands and pastures are often irrigated [81, 82]. Consequently, farmers can vary water release regimes to face climate change effects (higher temperatures, in particular) with the result of moving forage yields (that NDVI is a predictor of) to higher values. Differently, where more natural (not managed) systems are located (higher altitudes) the increasing of yearly MAXVI can be only related to glacier melting that could compensate the increasing of water requirement (as confirmed by the ET analysis) by vegetation: glaciers are, in fact, dramatically reducing in Aosta Valley. Moreover, another compensating action could come from the surrounding forest areas that have been proved to tolerate summer heatwaves. These results find strong evidences in different studies in literature [83-94].

With reference to ET graphs of figure 4d, a significant increasing trend was observed for all height classes. According to values of table 4, it can be noted that all height classes behave similarly; water requirement, in a period of 8 days, appears to averagely increase of about 0.05 Kg·m^-2 (about 0.5%) every year, for a total increase of about 1 Kg·m^-2 in 20 years (2000-2019), corresponding to a percentage difference in water requirement from vegetation of about 8%. This could be possibly explained by the increasing of biomass production (well represented by MAXVI) and by the enlargement of the growing season.

Authors are conscious that this work just focus on a semi-natural component as rangelands are. More investigations should be done with reference to other land cover classes. Nevertheless, preliminary results from this work could support breeders to better address their alpine activities like grazing and management of natural (water included) resources. A radical change is expected also by technicians and institutional subjects in their ordinary procedures, by adapting scheduling of alpine activities according to this information. It is authors’ opinion that the greatest limitation to the technology transfer of this type of approaches is a lack of “georeferenced” data from ground activities needed to calibrate and validate EO data-based deductions. A great effort is therefore required from involved players to carefully consider the possibility of structuring proper networks for data acquisition and transfer and procedures where georeferencing of ground observation is mandatory. Georeferencing of ground data is, in fact, at the basis of an effective and reliable spatial based approach like the one here proposed, where EO data (especially if available over a long-time span) play a crucial role. Consequently, we expect that future approaches should more properly rely on spatially and temporally distributed data in place of spatially aggregated and temporally distributed ones, like those adopted in this work. ” (Please see lines 487-566)

Point 4: The conclusion section needs to be rewritten. Only two sentences of the conclusion can be supported by this study (L396-L399), while other parts come from no bases (L387-L395, L400-406).

Response 4: The referee is right. The conclusion has been rewritten taking into account the revision made in the entire paper as follow: “Pheno-meteorological modeling and relative trends like ET and PMs observations derived from satellites EO data, with different time scales provide useful information about biosphere-atmosphere exchanges of carbon, energy, and water at both regional and global scale. This information is crucial to calibrate adaptive strategies to better face climate change effects in alpine ecosystems. It is worth to remind that mountains are among the most sensitive ecosystems to climate change, especially rising temperature, where, effects are faster to be observed than in other terrestrial habitats [95-106]. In this work, authors showed that in Aosta Valley phenological and evapotranspiration-related processes have been changing in the last 20 years with statistically significant trends. Climate change effects may represent an important threat to mountain ecosystems and related services for local populations. Their effects on “green” (vegetation-related) water resources can have important relapses on mountains agriculture and livestock productions (such as mountains pasture activities). Therefore, radical changes in rangeland management are expected to minimize these effects. Future studies will be intended to explore if the extension of the yearly period of alpine pasture for livestock and the improvement of grasslands irrigation systems at higher altitude in case of water deficit could represent key actions to be considered. EO data from public and free archives of ready-to-use products like MOD13Q1 and MOD16A2 ones, proved to support well this type of analysis in spite of their reduced geometric resolution. Specifically, they proved to be useful to analyze PMs and ET time trends that, if compared with the air temperature ones, could support breeders, and in general policy-makers, to better address management of alpine rangelands. A correct knowledge of these patterns is the starting point for a correct development of adaptation plans and strategies on alpine areas to face climate change effects on vegetation, like the ones highlighted in this study.” (Please see lines 567-603)

Point 5: Language editing would be required for this manuscript.

Response 5: Taking into account the referee suggestion a global English editing and revision was done.

Point 6: L102-L104. Higher spatial resolution images show denser images time series can be used to improve phenological deduction? Revise.

Response 6: Thanks for your suggestion. We did it. As follow: “Most recently, a number of studies have begun to use 10–30 m imagery, having similar spectral properties, available from Landsat and Sentinel 2 in order to improve phenological deductions at higher spatial resolution”.

(Please see line 121-124)

Point 7: L197. Re-projection?

Response 7: Thanks for your suggestion. We did it. As follow: “All data were converted from native geographical reference systems WGS84 into the ED50 UTM 32N one.” (Please see line 244-246)

Point 8: L348. How large of the MAXVI variation?

Response 8: Please see table 5.

Point 9: L350. How about the accumulated uncertainty?

Response 9: Please see table 5

Point 10: L355. Reference

Response 10: The reference has been added.

Point 11: L368. ET does not mean water requirement.

Response 11: We agree on the referee but if the temperature rise and evapotraspiration increase, it is more probable that plants need to have water availability to find optimal condition to grow up without stress factors. Anyway, this part was deleted from the discussions.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors have well addressed the comments/concerns raised during the first-round review. Thus, the revised version is recommended for acceptance and publication.

Reviewer 2 Report

The paper, in its current form, contains more inherent introduction, comments and discussion to the study area and the result presented.
However, remains the fact that the results (except the table 4 B and figure 5) have already been published elsewhere (exactly the same! https://doi.org/10.3390/rs12213542) and no newly investigations with different methods or approaches are here presented. My opinion is that one more graph is not enough to justify a paper published as new research article.

Reviewer 3 Report

Thank you for processing the suggestions and comments made in the first review. The quality of the article has improved.

Reviewer 4 Report

Sorry, I still do not see enough improvement or scientific contribution of this manuscript.

Article Menu

Exploring Short-Term Climate Change Effects on Rangelands and Broad-Leaved Forests by Free Satellite Data in Aosta Valley (Northwest Italy)

Further Information

Guidelines

MDPI Initiatives

Follow MDPI