Long-Term Series of Chlorophyll-a Concentration in Brazilian Semiarid Lakes from Modis Imagery

Round 1

Reviewer 1 Report

This manuscript estimated the Chla concentration using modis data of 21 years in Brazilian semiarid lakes. I think more analysis of the derived Chla results should be added in the revised manuscript. There are several major comments:
(1) The comparision between MOD09GA, MYD09GA between field Rrs should be added.
(2) The number of modis images used should be added.
(3) There are 5 datasets of field data in 13 lakes, why did you just apply the model to 3 largest lakes? 
(4) The optical properties of the 13 lakes may be different, you should analyze the optical properties or Chla of these lakes. 
(5) What is the spatial distribution of Chla in these lakes? 

Author Response

First of all, we thank you very much for your careful review. We addressed your comments as follows.

This manuscript estimated the Chla concentration using modis data of 21 years in Brazilian semiarid lakes. I think more analysis of the derived Chla results should be added in the revised manuscript.

We added more results on the chla time series in the section 3.4, namely historical means and standard deviations and an overview of the observed dynamics, which are further discussed in the Discussion. We also added a new result on the relationship between chla and water renewal rate (section 3.5).

There are several major comments:

(1) The comparision between MOD09GA, MYD09GA between field Rrs should be added.

It was added as the section 3.2 and Figure 2.

(2) The number of modis images used should be added.

It was added in the section 2.3: “A total of 185978 image subsets were  processed.”

(3) There are 5 datasets of field data in 13 lakes, why did you just apply the model to 3 largest lakes?

Because, as shown in Table 5, the number of successful chla retrieval was considerably higher in the three largest lakes, allowing for better analyses of the daily chla and hydrological data.

(4) The optical properties of the 13 lakes may be different, you should analyze the optical properties or Chla of these lakes.

It was added the section 3.1 on optical properties.

(5) What is the spatial distribution of Chla in these lakes?

It is not known, unfortunately. We could use satellite imagery to address that, but it would be out of the article’s scope, which is the temporal aspect of chla monitoring. To account for such uncertainty, the generation of the chla time series from satellite images was restricted to the lake’s area around the dam, which is expected to be more homogeneous and less subject to short-term variations when compared to the lotic-lentic transition area.

Reviewer 2 Report

The paper addresses the interesting and relevant subject of chlorophyll-a concentration in Brazilian semiarid lakes from Modis imagery. The paper has a nice presentation, clear to read, and well referenced. Congratulations to Dhalton Luiz Tosetto Ventura and other co-authors, especially professor Martinez. I recommend the paper to be published with considering the following modifications:

L18: 13 or 3 lakes? Please clarify.

I completely agree with the authors that the MODIS products are valuable for eutrophication management in waterbodies, especially lakes. This is the main reason for broad applications the MODIS imagery for exploring the change in chlorophyll-a over lakes in space and time. However, I would like to see more evidences of the previous applications of MODIS imagery on lakes in Introduction. This is very important especially for the large lakes, where regular monitoring of chlorophyll-a is so expensive and time consuming. For example, a successful application of MODIS imagery is on the largest lake in the world, i.e., the Caspian Sea, where a sign of progressive eutrophication was detected by archives of chlorophyll-a data obtained through the MODIS Aqua (Caspian Sea is eutrophying: The alarming message of satellite data). I strongly suggest the authors to highlight the importance of their study by referring to such valuable investigations.

L76-86: Chlorophyll-a data is archived in OceanColour with a daily interval and an acceptable spatial resolution. Why didn’t you use these archived datasets?  Would you please clarify?

Figure 1: Please show the lakes using blue color.

Section 2.1: Please give more information on the lakes (e.g., mean and max depths, possibility of lake surface freezing in cold months) as well as min, max, and mean of air temperature in the study areas.

Table 2: Large difference between mean and median reveals a non-normal distribution of your data. Did you try to make a normal distribution for your data?

Table 2: What is “ISS”? Please give its full name.

L213: Linear regression models? Please clarify.

L215: How did you fit linear, exponential and power relationships? Please give more information about the used package in R.

Figure 3: Please use the secondary vertical axis to show the change in lake storage (%).

Discussion: What do you think about the role of water residence time in eutrophic state in the lakes? Previous studies have shown that the hydraulic residence time is a key driver that controls eutrophication in lakes/reservoir. I suggest the authors to discuss your results with a focus on the long water residence times in the lakes and support the discussion by referring to relevant references. Please feel free to use the suggested published papers to support your conclusions: “Complex dynamics of water quality mixing in a warm mono-mictic reservoir”, “Hyper-nutrient enrichment status in the Sabalan Lake, Iran”, “The effects of water retention time and watershed features on the limnology of two tropical reservoirs in Brazil”, and “Temporal and depth variation of water quality due to thermal stratification in Karkheh Reservoir, Iran”.

Author Response

L18: 13 or 3 lakes? Please clarify.

We collected data in 13 lakes but generated the chla series for the 3 largest lakes, which had a chla retrieval frequency considerably higher than the remaining lakes and, as such, their data could be better analyzed in comparison to the daily lake storage data. We corrected the text.

I completely agree with the authors that the MODIS products are valuable for eutrophication management in waterbodies, especially lakes. This is the main reason for broad applications the MODIS imagery for exploring the change in chlorophyll-a over lakes in space and time. However, I would like to see more evidences of the previous applications of MODIS imagery on lakes in Introduction. This is very important especially for the large lakes, where regular monitoring of chlorophyll-a is so expensive and time consuming. For example, a successful application of MODIS imagery is on the largest lake in the world, i.e., the Caspian Sea, where a sign of progressive eutrophication was detected by archives of chlorophyll-a data obtained through the MODIS Aqua (Caspian Sea is eutrophying: The alarming message of satellite data). I strongly suggest the authors to highlight the importance of their study by referring to such valuable investigations.

We had focused on examples of application of Modis bands 1-7, but the Caspian Sea study suggested is indeed interesting, so we highlighted it in the Introduction. Thank you.

L76-86: Chlorophyll-a data is archived in OceanColour with a daily interval and an acceptable spatial resolution. Why didn’t you use these archived datasets? Would you please clarify?

As far as we are concerned, there is no chla data available in OceanColour for our study lakes. Anyhow, the spatial resolution of OceanColour products (>= 1 km) would not be appropriate for the lakes we studied. Spectral mixing and adjacency effects would be a big issue. And, importantly, the optical properties of ocean waters are very different from the eutrophic inland waters.

Figure 1: Please show the lakes using blue color.

Done.

Section 2.1: Please give more information on the lakes (e.g., mean and max depths, possibility of lake surface freezing in cold months) as well as min, max, and mean of air temperature in the study areas.

Unfortunately, there is scarce information on the lakes, but they are always warm (> 23° C) as a result of the high insolation during the whole year in the region (Barbosa, J.E. de L.; Medeiros, E.S.F.; Brasil, J.; Cordeiro, R. da S.; Crispim, M.C.B.; Silva, G.H.G. da Aquatic systems in semi-arid Brazil: limnology and management. Acta Limnol. Bras. 2012, 24, 103–118, doi:10.1590/S2179-975X2012005000030). We added such information to the study sites section as well as the mean depth to the descriptive table on the lakes.

Table 2: Large difference between mean and median reveals a non-normal distribution of your data. Did you try to make a normal distribution for your data?

No, but we tested log-transformed data in the regression models, so it may partially account for skewness in data distribution.

Table 2: What is “ISS”? Please give its full name.

It is “surface inorganic suspended solids”. The acronym was missing in the legend. We fixed it.

L213: Linear regression models? Please clarify.

Yes, linear regression models. We fixed the text.

L215: How did you fit linear, exponential and power relationships? Please give more information about the used package in R.

In fact, the text was misleading. What we did was to test log-linear and log-log models by log-transforming the data and then used the equivalent exponential and power equations in the validation step. We added that explanation to the text. Here you can find more details: https://kenbenoit.net/assets/courses/ME104/logmodels2.pdf.

Figure 3: Please use the secondary vertical axis to show the change in lake storage (%).

Done, thank you.

Discussion: What do you think about the role of water residence time in eutrophic state in the lakes? Previous studies have shown that the hydraulic residence time is a key driver that controls eutrophication in lakes/reservoir. I suggest the authors to discuss your results with a focus on the long water residence times in the lakes and support the discussion by referring to relevant references. Please feel free to use the suggested published papers to support your conclusions: “Complex dynamics of water quality mixing in a warm mono-mictic reservoir”, “Hyper-nutrient enrichment status in the Sabalan Lake, Iran”, “The effects of water retention time and watershed features on the limnology of two tropical reservoirs in Brazil”, and “Temporal and depth variation of water quality due to thermal stratification in Karkheh Reservoir, Iran”.

We had already discussed the role of the water renewal rate, which is the inverse of the residence time, as a hydraulic control on the chla, but now we added a subsection in results about the effect of the water renewal rate on the chla variation and we cited the suggested papers along the discussion. Thank you.

Round 2

Reviewer 1 Report

For this comments: "(5) What is the spatial distribution of Chla in these lakes?". I would like to see the spatial pattern of Chla in small lakes (largest 441 km2) using MODIS data. I know that for these small lake, modis could not map the distribution well. So, I think you should provide the spatial distribution of Chla, and discuss the limitation from the spatial resolution of modis in small lakes.

Author Response

For this comments: "(5) What is the spatial distribution of Chla in these lakes?". I would like to see the spatial pattern of Chla in small lakes (largest 441 km2) using MODIS data. I know that for these small lake, modis could not map the distribution well. So, I think you should provide the spatial distribution of Chla, and discuss the limitation from the spatial resolution of modis in small lakes.

We understand your point. For the large lakes, in general it should be possible to produce chlorophyll-a maps from Modis images. But it is not the case for the method we used. Man-made lakes are frequently dendritic and have islands and peninsula-like land strips, as is the case of our lake studies. With the spatial resolution of Modis, which is effectively less than the potentially 500-m at nadir, spectral mixing is inevitable. So, to deal with spectral mixed pixels and ensure more reliable chla estimation, we set our areas of interest close to the lakes’ dams (line 210), where the lakes are generally wider, with a larger water surface. Additionally, we used k-means clustering (line 215) to select the best available pixels. That means that for each image, we select only a portion of the pixels inside the defined area of interest. Then we average the reflectance in those pixels to, finally, estimate chla. This method, therefore, won’t result in chla maps. We acknowledge that limitation, but we accept it because the focus of our study was a historical evaluation of chla uncoupled to lake zonation. We are currently working with Sentinel-2 images and will soon be able to take such spatial approach, but it is a matter for a coming study.