Combining Global Remote Sensing Products with Hydrological Modeling to Measure the Impact of Tropical Forest Loss on Water-Based Ecosystem Services
, Sarah M. Walker 2 and Raghavan Srinivasan 3Round 1
Reviewer 1 Report
My opinion is that the manuscript is interesting and of interest to a broad Forests audience. A generally well-written paper. However, before publication, the manuscript needs below minor revisions. My review consists of a couple of minor issues, which I consider mandatory to be properly addressed in the revised version.
1. section 2.1: Not clear to get a LULCC maps for every year from 2001 to 2013 through all the processes in this section. Please describe more details about all the procedures and development processes in this section using figures (e.g. flow chart, etc.).
2. line 174: Please check the “LCLCC”, it should be “LULCC”.
3. line 186: What is the Pakse gauge?
4. lines 196 to 209 and section 3.1: Please do not confuse calibration and validation. The sentence “SWAT output discharge was calibrated based on MRC stream gauges within Cambodia to validate the model’s performance” should be corrected as model calibration process, I think. Typically, the model calibration and validation are implemented with different time series (i.e., years). So, maybe all the “validation” in these lines and sections can be corrected as “calibration”.
5. lines 236 to 237: Please insert the citation of reference (below would be good reference for it) for the use of spatial distribution of rainfall in the hydrological simulation.
Cho, Y. and B. A. Engel, 2017. NEXRAD Quantitative Precipitation Estimations for Hydrologic Simulation Using a Hybrid Hydrologic Model. Journal of Hydrometeorology, 18 (1), 25-47.
6. Figures 2 and 3: Some statistical values (e.g. R2, ENS, etc.) of model calibration results (performance) should be presented in these figures.
7. lines 291 to 295: Please provide more details of multiple regression analysis results with added variables using equations and figures.
8. line 312: Table 1 should be corrected as Table 3.
9. lines 316, 463, and 475: Please check the typo errors in these lines.
Author Response
Dear Forests Editors,
Thank you for the diligent review of the paper “Combining global remote sensing products with hydrological modeling to measure the impact of tropical forest loss on water-based ecosystem services.” All the of the requested edits and questions have been addressed in the manuscript (all changes are in MS Word track changes) and directly in our responses below. In our responses below we have identified the specific lines in the paper that have been edited. We believe these edits should address all the reviewer’s comments. Please see the responses below for reviewer 1. Please note that the line numbers indicated below refer to the newly edited manuscript.
Sincerely,
Michael Netzer – corresponding author
1. section 2.1: Not clear to get a LULCC maps for every year from 2001 to 2013 through all the processes in this section. Please describe more details about all the procedures and development processes in this section using figures (e.g. flow chart, etc.).
Response: To help clarify the development of the LULCC maps we have added explanatory text describing the process (lines 160-181), and added a diagram of the process that shows step by step how the maps were made (Figure 1 – line 182).
2. line 174: Please check the “LCLCC”, it should be “LULCC”.
Response: This was corrected. Thanks for catching that.
3. line 186: What is the Pakse gauge?
Response: We added explanatory text (lines 196-197). “The Pakse gauge (a hydrological gauging station maintained by the Mekong River Commission in the town of Pakse) was…”
4. lines 196 to 209 and section 3.1: Please do not confuse calibration and validation. The sentence “SWAT output discharge was calibrated based on MRC stream gauges within Cambodia to validate the model’s performance” should be corrected as model calibration process, I think. Typically, the model calibration and validation are implemented with different time series (i.e., years). So, maybe all the “validation” in these lines and sections can be corrected as “calibration”.
Response: To address the confusion between calibration and validation we have removed all reference to validation, and correctly identified it as calibration of the model. These corrections are primarily in Section 2.2. and 3.1.
5. lines 236 to 237: Please insert the citation of reference (below would be good reference for it) for the use of spatial distribution of rainfall in the hydrological simulation.
Cho, Y. and B. A. Engel, 2017. NEXRAD Quantitative Precipitation Estimations for Hydrologic Simulation Using a Hybrid Hydrologic Model. Journal of Hydrometeorology, 18 (1), 25-47.
Response: this was added.
6. Figures 2 and 3: Some statistical values (e.g. R2, ENS, etc.) of model calibration results (performance) should be presented in these figures.
Response: Statistical values for Nash-Sutcliffe efficiency (NSE) and R2 were added to Figure 2 (now Figure 3) and Figure 3 (now Figure 4). The addition of these statistics required explanation in the methods section lines 212-228.
7. lines 291 to 295: Please provide more details of multiple regression analysis results with added variables using equations and figures.
Response: We have added Table 3 to address this comment, lines 317-320.
8. line 312: Table 1 should be corrected as Table 3.
Response: Corrected. Thanks.
9. lines 316, 463, and 475: Please check the typo errors in these lines.
Response: All corrected. These corrections are on line 337 (corrected Chklok to Chklong), line 484 (delete extra comma), line 496 (delete extra period)
Reviewer 2 Report
I read the paper by Netzer et al., with interest. It is well written and structured. I have some comments, though, about the global tree cover and tree-loss data from Hansen et al. (2013). This study is based on this data but :
- I will be interested to see the original data of Hansen over the study area
- How forest loss in figure 4 and in table 2 was computed?
- Why the spatial patterns of the forest loss in figure 4 do not correspond to the forest loss in Fig 6?
- What is the reason on only selecting surface runoff and sediment yield and no other hydrological variables?
- Gathering data with different spatial resolutions may have an impact on the analyses. Can the authors comment on this?
Without understanding the original Hansen data, it will be difficult to understand the work of Netzer et al.,. I therefore suggest the authors to describe in details the Hansen data.
Author Response
Dear Forests Editors,
Thank you for the diligent review of the paper “Combining global remote sensing products with hydrological modeling to measure the impact of tropical forest loss on water-based ecosystem services.” All the of the requested edits and questions have been addressed in the manuscript (all changes are in MS Word track changes) and directly in our responses below. In our responses below we have identified the specific lines in the paper that have been edited. We believe these edits should address all the reviewer’s comments. Please see the responses to reviewer 2 below. Please note that that the edited lines indicated below correspond with the newly edited manuscript.
Sincerely,
Michael Netzer – corresponding author
Reviewer 2
I read the paper by Netzer et al., with interest. It is well written and structured. I have some comments, though, about the global tree cover and tree-loss data from Hansen et al. (2013). This study is based on this data but:
1. I will be interested to see the original data of Hansen over the study area?
Response: Figure 5 already shows this in red labeled “Forest loss”. We have added a more explicit reference in the caption.
2. How forest loss in figure 4 and in table 2 was computed?
Response: Forest loss is taken directly from the Hansen dataset without any further processing (other than basic GIS manipulation like projection). Figure 4 (now Fig 5) and Figure 6 (now Fig 7) show the forest loss data directly from Hansen, while Table 2 calculated area of loss within the boundaries of each watershed which were delineated using the SRTM 90m DEM.
3. Why the spatial patterns of the forest loss in figure 4 do not correspond to the forest loss in Fig 6?
Response: Figures 4 (now 5) and 6 (now 7) show the exact same forest loss layers and therefore the same spatial patterns. They may appear to be different due to the different color patterns, but upon close analysis they are the same.
4. What is the reason on only selecting surface runoff and sediment yield and no other hydrological variables?
Response: Surface runoff and sediment yield were selected for two reasons: 1) because they provide useful indicators for two of the most common and important ecosystem services (ES): flood mitigation and erosion prevention. Given the main focus of the paper is on quantifying how forest loss impacts ecosystem service provision, we tried to focus on SWAT outputs that best represented the most important ES and, 2) we were able to perform some model calibration on river discharge and sediment flow, which are correlated strongly with surface runoff and sediment yield. We therefore were more confident discussing conclusions based on these variables than other variables such as nutrient runoff, for which we had no calibration or validation data.
5. Gathering data with different spatial resolutions may have an impact on the analyses. Can the authors comment on this?
It is true, there can be a loss of accuracy when using data with different spatial resolutions. However, in this case any loss would be extremely minimal, and there would be no way to alleviate it. The reason for this is that our key dataset in assessing change was the Hansen (et al 2013) forest loss layer (sampled at a resolution of 100x100m). The forest loss layer was essentially overlaid with the GlobeCover (GLC 2015) dataset that is 300x300m resolution. As forest was turned into non-forest (using the forest loss layer) it simply referenced the GlobeCover land use class below it (at that location) so it could be reclassified. Therefore, both the forest loss and the Globe Cover maintain the integrity of their respective resolutions. Also, there are no other global datasets available that classify non-forest land cover at resolution of 100x100m, therefore the difference in resolution could not be avoided.
Without understanding the original Hansen data, it will be difficult to understand the work of Netzer et al. I therefore suggest the authors to describe in details the Hansen data.
Response: We have added extra text in a number of spots, including lines 123-128.
