the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Flood estimation for ungauged catchments in the Philippines
Abstract. Flood magnitude and frequency estimation are essential for the design of structural and nature-based flood risk management interventions and water resources planning. However, the global geography of hydrological observations is uneven; in many regions, such as the Philippines, data are spatially and/or temporary sparse, limiting the choice of statistical methods for flood estimation. We evaluate the potential of pooling short historical data series for ungauged catchment flood estimation. Daily mean river discharge data were collected from 842 sites, with data spanning from 1908 to 2018. Of these, 513 candidate sites met criteria to estimate a reliable annual maximum flood. Using the index flood approach, a range of controls were assessed at national and regional scales using land cover and rainfall datasets, and GIS-derived catchment characteristics. Multivariate analysis for predictive equations for 2 to 100 year recurrence interval floods based on catchment area only have R2 ≤ 0.59. Additionally, adding a rainfall variable, the median annual maximum 1-day rainfall, increases R2 to between 0.56 for Q100 and 0.66 for Q2. Very few other variables were significant when added to multiple regression equations. Although the Philippines exhibits regional climate variability, there is limited spatial structure in predictive equation residuals and region-specific predictive equations do not perform significantly better than national equations. Relatively low R2 values are typical of studies from tropical regions. The predictive equations are suitable for use as design equations for the Philippines but uncertainties must be assessed. Our approach demonstrates how combining individually short historical records, after careful screening and exclusion of erroneous data, generates large data sets that can produce consistent results. Extension of continuous flood records is required to reduce uncertainties but national-scale consistency suggests that extrapolation from a small number of carefully selected catchments could provide nationally reliable predictive equations with reduced uncertainties.
- Preprint
(1646 KB) - Metadata XML
-
Supplement
(1677 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on hess-2024-188', Anonymous Referee #1, 07 Aug 2024
This study evaluates 11 physical variables for index flood estimation across catchments in the Philippines, aiming to enhance flood estimation for ungauged catchments. The authors show significant effort in data collection and selection, and they present extensive analyses in this manuscript. Notice that the authors claim their study is applicable to ungauged catchments. However, if my understanding is correct, the analyses presented do not show anything regarding this applicability. While they propose using more local information to improve flood estimation—a common approach in many studies—they suggest this could benefit ungauged sites. Although this suggestion might be correct, it is overstated in the title since there are no relevant analyses or validation to support this claim. In addition, the manuscript has several critical issues regarding the quality: 1) Unclear Critical Information - There is confusion regarding the number and details of study sites, as well as incomplete or unclear descriptions of methodologies. 2) Lack of Novelty and Significant Findings - The framework lacks innovation and the findings are not particularly groundbreaking (as noted by the authors in line 527). 3) Quality and Clarity - The structure of the manuscript, along with its figures and tables (including captions), lacks quality and clarity. There are numerous mistakes throughout the document. I found it challenging to understand the authors' main points, both from the text and the figures. While the authors' efforts in conducting numerous analyses are commendable, they are strongly encouraged to improve the manuscript by enhancing its accuracy, clarity, and focus. Some specific comments (but not all) for improvement are listed below for reference:
- How many catchments are analyzed exactly? Is it 513 or 466? The abstract states 513, but other parts of the manuscript (e.g., Figure 1, Table 2, line 172) suggest it is 466. Lines 164-172 are particularly confusing: how does 513 minus 205 result in 466 sites?
- The catchment area sizes are analyzed in this study, but this information is missing in the data section.
- Why do the catchment area groupings differ between Table 2 and Figure 3? The former uses four groups (100-200, 200-400, 400-800, <800), whereas the latter uses five groups (<25, 25-50, 50-250, 250-2500, >2500), with so different ranges.
- Is the area grouping range rational? Comparing catchments across such varied groupings (<25, 25-50, 50-250, 250-2500, >2500) seems to be strange for representing hydrological responses.
- The title of Section 5.1 is not coherent with its content. Moreover, it is difficult to discern the patterns the authors aim to show (lines 232-234) because Figure 3 is unclear. Improving the color and marker settings or changing the plot type (e.g., stacked bar plots) might be helpful.
- Line 233: Figure 3 refers to area, not climate.
- Line 233: All the others should point to Figure S5, not Figure S6.
- There are references to regions 1-13 in many analyses, but no introduction or proper definition of these regions is provided. The first mention appears on line 78, without any definition.
- Line 213 claims that the authors apply the FEH approach described earlier, but there is no earlier description. The term is introduced on line 50, but without sufficient detail.
- Improper structure: Lines 220-227 fit better in the data section rather than the methods section.
- Line 245: Why are only 71 sites analyzed here instead of the full 466 sites?
- The figure captions should be more descriptive of the settings, but discussions of the results/patterns should not be included here.
- Tables could be improved by adding delineation lines, clarifying headers (e.g., the last item in the first column of Table 2), and removing unnecessary information.
In summary, I acknowledge that such a study is needed for the selected country, as the authors claim there are no other similar studies to such an extent. While the analyses are comprehensive, the manuscript lacks sufficient clarity in its structure and critical information, which hampers its transferability and overall readability.
Citation: https://doi.org/10.5194/hess-2024-188-RC1 - AC2: 'Reply on RC1', Pamela Louise M. Tolentino, 01 Oct 2024
-
RC2: 'Comment on hess-2024-188', Anonymous Referee #2, 27 Aug 2024
The paper by Hoey et al., entitled ‘Flood estimation for ungauged catchments in the Philippines’ aims at delivering design equations to estimate flood magnitudes and frequency in ungauged catchments across the Philippines.
The paper addresses an interesting topic, the design of equations to predict floods in data scarce region. The authors do enormous work to analyse the data. However, the methodology is not well defined/structured. Overall, the paper lacks clarity and needs to be restructured for better clarity.
General comments:
The title of the paper does not reflect the content of the paper appropriately. The authors propose and evaluate the index flood approach and multi-variate regression to estimate flood. The approach is evaluated only at catchments with data used to fit the regression equations. In a context of flood estimation in ungauged catchments, the paper is lacking the following steps: 1) a clear definition of the methodology used for regionalization to ungauged catchment, 2) a cross-validation to evaluate the performance of the designed equations in ungauged catchments. It is necessary to evaluate how well the proposed approach would perform in ungauged catchments. Therefore, I suggest the authors consider using some cross-validation approach (i.e., leave-one-out or k-cross validation) where a set of catchments are used to fit the design equations; then the remaining catchment(s) are used as pseudo-ungauged catchment(s) to evaluate the accuracy of the designed equations in ungauged catchments. In addition, the methodology used is not clearly described (see specific comments below). The paper could benefit from a flow diagram that describes the steps followed to estimate floods at ungauged catchments.
Specific comments:
- The authors fit and compare the accuracy of the three distributions showing that the choice of the distribution influences flow estimates and no distribution performs well at all sites. Then, the GLO distribution is selected to predict high flow magnitude in section 4.2 with no justification of this choice. In addition, Equation 5 , the factorial standard error for the GLO distribution, is only applicable when number of records is at least 20 years. The study region has only 71 sites (line 245), with more than 20 years data, among the 466 sites retained for the analysis (Line 234) . Why choose the GLO distribution if it cannot be applicable to all sites? Why choose only GLO instead of using the best fitted distribution at each site as shown in table 2?
- The last paragraph in section 4.1, the authors lists three ways the construct growth curves by combining the curves fitted from each site by region, climate type or catchment areas. Therefore, I am assuming that the resulting curves are regionalized curves where catchments from within each group(region or climate type, catchment area bin) would be represented by a single curve. In other words, catchments of each group would have a single QT value estimate from the regionalized growth curve. How the author use these results to perform a correlation analysis with catchment properties in Section 4.2 (Line 212: values of QT provided by the GLO analysis previously, were correlated with catchment properties”). ?
- The authors use precipitation dataset covering only 17 years (from 1998-2015) which is not sufficient to derive climatology of the catchments. In addition, this dataset are not available for a period that overlaps with flows records period. As suggestion, the author could use precipitation datasets from reanalysis products (e.g. ERA5) which are available from 1940 to present around the globe. It would cover the same period as flow records for most catchments and provide sufficient long timeseries to derive climatology for precipitation variables used in Table 3.
- Why the fitted curves from individual sites are not used to fit a relationship between flood magnitude and catchment area?
- It is unclear how the individual short historical records are combined to generate large dataset that produce consistent results.
- Results presented in Section 5.3.3 (Eq6a-6c) would be better presented in a table with equation and associated R2 and RMSE in the same way as in Table 5.
Overall, the way the methodology and the results are presented is incoherent and lacks clarity.
Citation: https://doi.org/10.5194/hess-2024-188-RC2 - AC1: 'Reply on RC2', Pamela Louise M. Tolentino, 01 Oct 2024
Data sets
Flood estimation for ungauged catchments in the Philippines: Annual Maximum Flow (AMAX) and catchment properties data T. B. Hoey et al. https://doi.org/10.5525/gla.researchdata.1666
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
316 | 111 | 44 | 471 | 32 | 18 | 14 |
- HTML: 316
- PDF: 111
- XML: 44
- Total: 471
- Supplement: 32
- BibTeX: 18
- EndNote: 14
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1