Climate and Land Use/Land Cover Changes within the Sota Catchment (Benin, West Africa)

Round 1

Reviewer 1 Report

This research studies rainfall and temperature changes in the Sota catchment (West Africa) at annual and seasonal scale, for a  period of 60 years. At the same time, LULC changes were studied for years 1990, 2005, and 2020. Authors have done an extensive work.

But, the main issue I found is that the manuscript itself is too large. It must be shorten in order to motivate readers, and also to guide the reading.

Materials and Methods: try to be brief, and do not miss the relevant things. For example, it is not mentioned why specific years 1990, 2005 and 2020 where chosen for studying the land use. Also, there is not much detail regarding the characteristic of the established LULC classes.

From my point of view, the Results section must be rewritten. Trying to shorten some sentences, and also eliminating some Figures and Tables.

Below there are some specific comments:

- Abstract: explain why the chosen climate period is different from the chosen land use change period. Include the location of the Sota catchment and its area.

-line 37: Some keywords are already in the Title

-Table 3, Figure 2: I believe it is not necessary all this detail

-Line 280-345: this explanation must be shorten. It has too many details.

-Figure 3 and Figure 4: could be joined. So it will help readers to locate in the catchment.

-Table 16 and Figure 13: brings almost the same information

Author Response

Reviewer 1

1-This research studies rainfall and temperature changes in the Sota catchment (West Africa) at annual and seasonal scale, for a period of 60 years. At the same time, LULC changes were studied for years 1990, 2005, and 2020. Authors have done an extensive work. But, the main issue I found is that the manuscript itself is too large. It must be shorten in order to motivate readers, and also to guide the reading.

Author response: The manuscript was shortened as suggested by the reviewer. Some figures  (Figure 2, Figure 3, Figure 8, Figure 11) and tables (Table 3, Table 13, Table 16) have been deleted.

2-Materials and Methods: try to be brief, and do not miss the relevant things. For example, it is not mentioned why specific years 1990, 2005, and 2020 were chosen for studying land use. Also, there is not much detail regarding the characteristic of the established LULC classes.

Author response:  The reason why specific years 1990, 2005 and 2020 were chosen for studying the land use, has been added (Page 6, section 2.2.3). In fact, similarly to the climate data period, we also wanted to analyze land use from 1960 to 2020. However, satellite images are not available in the region in the 1960s and even in the 1970s making it difficult to make a LULC classification in these years. We could have used the Global LULC data over Africa available in 1975 but they are less accurate since their image resolution is 2km. The satellite images are observable in the region from 1986 with a resolution of 30m. Therefore, we proceeded to a LULC classification over the most recent period from 1990 to 2020. We found it better to split the study period into a period of 15 years to have 3 dates (1990-2005-2020) with similar intervals, and especially to have the possibility to better identify and analyze the changes.

The established LULC classes are the commonly classes identified in the region and in the whole Benin in general. Their description are mentioned in the following references (Koumassi et al. 2014; Zakari et al. 2015).

3. 3. From my point of view, the Results section must be rewritten. Trying to shorten some sentences, and also eliminating some Figures and Tables.

 Author response: Some sentences are shortened and some figures and tables have been deleted

-Below there are some specific comments:

Abstract: explain why the chosen climate period is different from the chosen land use change period. Include the location of the Sota catchment and its area.

 Author response: The reason why the chosen climate period is different from the chosen land use change period has been explained in the abstract. The area and the location of the Sota catchment have been also included in the abstract as suggested by the reviewer.

 -line 37: Some keywords are already in the Title

Author response: The keywords section is revised to take into account the reviewer’s comment 

-Table 3, Figure 2: I believe it is not necessary all this detail

Author response: Table 3, and Figure 2 have been deleted

-Line 280-345: this explanation must be shorten. It has too many details.

Author response: This part has been shortened as suggested by the reviewer

-Figure 3 and Figure 4: could be joined. So it will help readers to locate in the catchment.

Author response:  In the reviewing process, Figure 3 has been finally deleted from the manuscript.  

-Table 16 and Figure 13: brings almost the same information

Author response: We only kept Figure 13 and deleted Table 16

References

1. Zakari et al., “Variabilité hydropluviométrique et dynamique de l’occupation des terres dans le bassin de la Sota à l’exutoire de Coubéri au Bénin (Afrique de l’Ouest),” Int. J. Innov. Appl. Stud., vol. 13, no. 2, pp. 235–250, 2015, [Online]. Available: http://www.ijias.issr-journals.org/.
2. H. Koumassi, “RISQUES HYDROCLIMATIQUES ET VULNERABILITES DES ECOSYSTEMES DANS LE BASSIN VERSANT DE LA SOTA A L ’ EXUTOIRE DE COUBERI. These de Doctorat,” Univ. d’Abomey-Calavi, 2014, [Online]. Available: https://hal.archives-ouvertes.fr/tel-01572602.

Reviewer 2 Report

This study is focused on assessment of climate and Landcover changes over the Sota catchment. The rainfall for a period of 1960 – 2019 have been assessed. The landuse changes have been correlated with the rainfall patterns. The breakpoint analyses based on the Tmax, and Tmin time series have not been conducted. Based on these results, critical steps such as adapting and mitigating to climate and land cover changes over the Sota catchment have been identified and discussed.

The authors should have included the departure analysis to understand the rainfall variation especially the effect of landuse.

Estimation of LST would have helped more to understand the variations in rainfall

Language is good as a whole. 

Author Response

1. This study is focused on assessment of climate and Landcover changes over the Sota catchment. The rainfall for a period of 1960 - 2019 have been assessed. The landuse changes have been correlated with the rainfall patterns. The breakpoint analyses based on the Tmax, and Tmin time series have not been conducted. Based on these results, critical steps such as adapting and mitigating to climate and land cover changes over the Sota catchment have been identified and discussed. The authors should have included the departure analysis to understand the rainfall variation especially the effect of landuse. Estimation of LST would have helped more to understand the variations in rainfall

Author response: We agree with the reviewer that departure analysis will help to understand the rainfall variation. The rainfall departure and SPI are ideal indicators of dry or wet conditions for a given time over a specified area. We tried to perform the departure analysis and observed that the results are almost the same as those of the SPI analysis showing dry years in the 1970s and 1980s as in West Africa. As we were requested to reduce the content and the length of the paper we didn’t include it. We thank the reviewer for this remark. LST datasets in the region are not available over the whole study period while we need climate data from 1960. We consider this as a limitation of the study and suggest it for further studies on page 25.

Reviewer 3 Report

This paper examines the Sota Catchment in the Republic of Benin, West Africa, in terms of climate change and land cover change. Sixty years of rainfall and temperature changes were analyzed by combining meteorological station data with grid data between 1960 and 2019; land cover changes were also assessed for 1990, 2005 and 2020. Climate and land cover changes are key for regional or global management, and the paper provides a detailed study of these two components. However, I suggest that the article would be changed and resubmitted with a major revision, because the article is more like a technical report rather than a scientific paper.

(1)    The analysis of the article is still at a relatively superficial level, and should be suggestive or meaningful, e.g., is there a link between the current regional climate change and land use cover change? Is it indicative? In particular, what is indicative of hydrological processes?

(2)    Please reduce large number of figures and leaving key figures, also, many figures do not meet the standards of a scientific paper. Please add trend equation in Figure 3, Figure 9; the horizontal coordinates of Many figures are not clearly visible. Line markers in the text overlap with parts of the table!

This paper examines the Sota Catchment in the Republic of Benin, West Africa, in terms of climate change and land cover change. Sixty years of rainfall and temperature changes were analyzed by combining meteorological station data with grid data between 1960 and 2019; land cover changes were also assessed for 1990, 2005 and 2020. Climate and land cover changes are key for regional or global management, and the paper provides a detailed study of these two components. However, I suggest that the article would be changed and resubmitted with a major revision, because the article is more like a technical report rather than a scientific paper.

(1)    The analysis of the article is still at a relatively superficial level, and should be suggestive or meaningful, e.g., is there a link between the current regional climate change and land use cover change? Is it indicative? In particular, what is indicative of hydrological processes?

(2)    Please reduce large number of figures and leaving key figures, also, many figures do not meet the standards of a scientific paper. Please add trend equation in Figure 3, Figure 9; the horizontal coordinates of Many figures are not clearly visible. Line markers in the text overlap with parts of the table!

Author Response

 The analysis of the article is still at a relatively superficial level, and should be suggestive or meaningful, e.g., is there a link between the current regional climate change and land use cover change? Is it indicative? In particular, what is indicative of hydrological processes?

Author response:

We agree with the reviewer that the link between the current regional climate change and land use cover change has not been addressed in the paper. We therefore discussed this concern in section 4 on page 24.

(2)    Please reduce large number of figures and leaving key figures, also, many figures do not meet the standards of a scientific paper. Please add trend equation in Figure 3, Figure 9; the horizontal coordinates of Many figures are not clearly visible. Line markers in the text overlap with parts of the table!

Author response: Many figures (Figure 2, Figure 3, Figure 8, Figure 11) and tables (Table 3, Table 13, Table 16) have been deleted. Trend equation has been added to the previous Figure 9 which became Figure 6 on page 16. Figures clarity was improved. The content of the paper has been improved, some sentences have been reformulated, and the results part has been shortened.  

Reviewer 4 Report

The overall scope of the work is to use weather station data and gauge data and gridded data to assess changes and trends in precipitation and temperature in the Sota Basin in Benin for circa 60 years, from 1960 though 2019. In doing this the authors describe quite well the input data, the methods for data quality control and the statistical tests. All is very rigorous. The excercise is useful to produce statistics for the area of interest, however results presented in tables 9 and 10 may be a bit difficult to read.

The work also includes land cover mapping across three epocs: 1990, 2000 and 2020. The authors have acquired 4 Landsat tiles covering the Sota basin for each epoch. Furthermore the authors have gathered pesudo in-situ data from google earth,  through visual interpretation for the three epocs collecting 100 data points for the years 1990, 2000 and 150 samples for the year 2020. The authors mention that  ''points of interest were methodically established 317 using Google Earth imagery for the years 1990, 2005, and 2020, with a keen emphasis  on aligning these points with the spectral signatures characteristic of the land cover  classes observed within the Landsat images''. Additional samples were collected in the field in 2020, apparently 50 data points.  Finally the authors use the pseudo in-situ data to perform a supervised classification of the landsat mosaics, finally resulting into LC maps for 1990, 2000 and 2020, featuring 9 land cover classes. The maps are assessed for accuracy using a confusion matrix. The three maps are then compared to assess land cover change. Finally land cover statistics and land cover change statistics are provided, based on pixel counting. In my view there are several steps which could benefit from clarification and possibly improvements. These points are:

1)  It seems that the approach selected for image analysis is based on a single acquisition rather that a time series approach. Was there a specific reason for this? e.g. clouds and lack of sufficient number of available landsat images? furthermore it is not clear which features have been extracted from the images and used for the maximum likelihood classification.

2) the collection of samples from the google earth engine seems to lack an a priori survey design, based on either probababilistic or non probabilistic reasoning. The subjective nature of the method is subject to intorduce bias in both the informing of the maximum likelihood classifier (non representative sample), and most importantly on the validity of the results of the confusion matrix. It is also not clear how the sample data set was split into a subset for data for the classification excercise and a sub set for the validation. Looking at the number in the confusion matrixes, it seems that all the samples were used for the validation. 

3) When the authors compare the land cover maps to assess land cover change, they assume that the land cover maps are consistent in time in terms of accuracy (stability of accuracy, and a very low bias in area estimation). However this is not backed up in the paper. Furthemore the validation of the land cover change class is not performed.  A change between two maps could be very the result of an error in the classifier and not of a real change on ground. 

4) The computation of the land cover area statistics, as well as of the land cover change area statistics are based on pixel counting: this method is not adequate, becuase it does not take into account the bias of the estimator which is class specific.   

As a final recommendation, I would encourage the sharing of results in gis files, so that those interested in the methods as well as in the actual study area can access these dinamically using any gis software or API.

Author Response

1. The overall scope of the work is to use weather station data and gauge data and gridded data to assess changes and trends in precipitation and temperature in the Sota Basin in Benin for circa 60 years, from 1960 though 2019. In doing this the authors describe quite well the input data, the methods for data quality control and the statistical tests. All is very rigorous. The excercise is useful to produce statistics for the area of interest, however results presented in tables 9 and 10 may be a bit difficult to read.

Author response: The tables have been improved to be reader-friendly, following the reviewer comment.

2. The work also includes land cover mapping across three epocs: 1990, 2000 and 2020. The authors have acquired 4 Landsat tiles covering the Sota basin for each epoch. Furthermore the authors have gathered pesudo in-situ data from google earth, through visual interpretation for the three epocs collecting 100 data points for the years 1990, 2000 and 150 samples for the year 2020. The authors mention that  ''points of interest were methodically established 317 using Google Earth imagery for the years 1990, 2005, and 2020, with a keen emphasis  on aligning these points with the spectral signatures characteristic of the land cover  classes observed within the Landsat images''. Additional samples were collected in the field in 2020, apparently 50 data points.  Finally the authors use the pseudo in-situ data to perform a supervised classification of the landsat mosaics, finally resulting into LC maps for 1990, 2000 and 2020, featuring 9 land cover classes. The maps are assessed for accuracy using a confusion matrix. The three maps are then compared to assess land cover change. Finally land cover statistics and land cover change statistics are provided, based on pixel counting. In my view there are several steps which could benefit from clarification and possibly improvements. These points are:

- It seems that the approach selected for image analysis is based on a single acquisition rather that a time series approach. Was there a specific reason for this? e.g. clouds and lack of sufficient number of available landsat images? furthermore it is not clear which features have been extracted from the images and used for the maximum likelihood classification.

Author response: The choice of satellite images is based on criteria such as spatial coverage, spatial resolution, available years, and cloudiness (minimum). It appears that for some periods images are not available in the Sota region. Since the single image acquisition approach can make it possible to perform the LULC analysis, and is commonly used, and requires a lower number of images than a time series approach, we adopted this approach (Refer to page 5 section 2.2.3). Regarding the features extracted, the supervised classification was applied after defined area of interest (AOI) which is called training classes. More than one training area was used to represent a particular class. The training sites were selected in agreement with the Landsat Image, Google Earth.

- The collection of samples from the google earth engine seems to lack an a priori survey design, based on either probababilistic or non probabilistic reasoning. The subjective nature of the method is subject to intorduce bias in both the informing of the maximum likelihood classifier (non representative sample), and most importantly on the validity of the results of the confusion matrix. It is also not clear how the sample data set was split into a subset for data for the classification excercise and a sub set for the validation. Looking at the number in the confusion matrixes, it seems that all the samples were used for the validation. 

Author response: satellite image classification relies on a fusion of remotely sensed data with ground-based reference data or aerial photographs captured in close temporal proximity to the satellite pass. However, in the context of Benin, the availability of such reference data was limited, necessitating the utilization of a very high-resolution imagery sourced from various platforms within Google Earth as the reference dataset for our classification efforts. Therefore to ensure the precision and validity of the classification results, points of interest were methodically established using Google Earth imagery for the years 1990, 2000, and 2020 with a keen emphasis on aligning these points with the spectral signatures characteristic of the land cover classes observed within the Landsat images. As noted by the reviewer, a priori survey design was not carried out for the collection of samples on google earth. As a result, supplementary points of interest were meticulously gathered in the field during December 2022, specifically tailored to the year 2020, thereby reinforcing the robustness and accuracy of the classification. 

- When the authors compare the land cover maps to assess land cover change, they assume that the land cover maps are consistent in time in terms of accuracy (stability of accuracy, and a very low bias in area estimation). However, this is not backed up in the paper. Furthermore, the validation of the land cover change class is not performed. A change between two maps could be very the result of an error in the classifier and not of a real change on ground. 

Author response:  The consistency of the land cover maps in time in terms of accuracy (stability of accuracy, and a very low bias in area estimation) was based on the results of the overall accuracy and the Kappa index, which we refer to on page 19  section 3.6 in the manuscript. Validation of land cover change was performed using ground truth data to authenticate the accuracy of the classification outcomes and the resultant land cover maps page 9 section 2.3.4 …  

- The computation of the land cover area statistics, as well as of the land cover change area statistics are based on pixel counting: this method is not adequate, because it does not take into account the bias of the estimator which is class specific.   

Author response: Yes we agree. Our option of choosing land cover statistics based on pixel counting is that Pixel counting is probably the most popular way to estimate class areas from satellite-derived maps and was used in many studies (Rwanga et al.2017; Waldner et al. 2017; Dossa et al. 2015; Ahononga at al. 2015). But every method has its weaknesses . There is no perfect one. We thank the reviewer for this remark and we included it as one limitation of the study on page 25 section 5.  

As a final recommendation, I would encourage the sharing of results in gis files, so that those interested in the methods as well as in the actual study area can access these dynamically using any gis software or API.

Author response: We thank the reviewer for this recommendation. It is a pleasure to share the research achievements. We will consider this very positive suggestion!

      References

1. Rwanga, S. and Ndambuki, J. (2017) Accuracy Assessment of Land Use/Land Cover Classification Using Remote Sensing and GIS. International Journal of Geosciences, 8, 611-622. doi: 10.4236/ijg.2017.84033.

2. O. S. N. Dossa, G. H. Dassou, A. C. Adomou, F. C. Ahononga, and S. Biaou, “Dynamique spatio-temporelle et vulnérabilité des unités d’occupation du sol de la Forêt Classée de Pénéssoulou de 1995 à 2015 (Bénin, Afrique de l’Ouest). Sciences de la vie, de la terre et agronomie, 9(2), Article 2. http: publication.lecames.org/index.p,” 2021.
3. C. Ahononga, N. G. Gouwakinnou, and S. Biaou, “Vulnérabilité des terres des écosystèmes du domaine soudanien au Bénin de 1995 à 2015. Bois et Forêts des Tropiques, 346, 35–50. https://doi.org/10.19182/bft2020.346.a36295,” 2020.
4. Biaou et al., “Dynamique spatio-temporelle de l’occupation du sol de la forêt classée de Ouénou-Bénou au Nord Bénin. Des Images Satellites Pour La Gestion Durable Des Territoires En Afrique, 22. https://hal.archives-ouvertes.fr/hal-02189367,” 2019.
5. Waldner, P. Defourny / International Journal of Applied Earth Observation and Geoinformation 60 (2017) 1–10

Round 2

Reviewer 1 Report

The manuscript has been improved. But, the whole text must be checked one more time in order to find:

- spelling mistakes: as "addiion" (line 1006)

- repeted phrases: as "in addition" (lines 1006 and 1007)

I did not find the reference number 66 and 67 cited in the manuscript.

Author Response

The manuscript has been improved. But, the whole text must be checked one more time in order to find:

- spelling mistakes: as "addiion" (line 1006)

Author response: The whole text has been checked again. The mistake“ addiion” has been corrected (line 1006)

- repeted phrases: as "in addition" (lines 1006 and 1007)

Author response: this mistake has been also corrected (lines 1006 and 1007)

-I did not find the reference number 66 and 67 cited in the manuscript.

Author response: the reference numbers 66 and 67 have been deleted from the manuscript.

The whole text has been checked again.

 We thank the reviewer for his relevant comments, which helped to improve the manuscript.