Interactive comment on “ Tropical drought risk : estimates combining gridded vulnerability and hazard data

In general, the methodology is scientifically sounded, but I have some comments: The assessment of drought hazard seems to be straight forward. However, it would be useful if authors provide the time periods of analyses for each studied catchments, then the reader can have an idea of whether the periods are long enough to reflect the variability of the precipitation. It is not clear if they used the same period from all the catchments. Related to this, in line 144, authors mention they used a period from 1980-2018 to analyze CHIRPS data, but in line 156, they say they considered a period from 1980-2017.

sectors were affected? Since such characteristics are climate, region-and sector specific, there is a demand 65 to design locally suitable drought risk assessment approaches and related data sets .
The scale of analysis matters. Widely applied monthly scale standardized indices (eg. SPI 12) are useful for large scale drought risk assessment . Tropical climates are often dominated by a strong seasonality and a topography-influenced spatial rainfall variability. Few days without 70 rainfall might lead to a severe precipitation deficit that can affect cattle grazing and rain-fed agricultural production. Indices based on monthly hydro-meteorological values might not detect short-term deficits in quickly responding catchments. For tropical regions, it has therefore been proven useful to assess meteorological and hydrological drought hazard at a daily timescale (Nauditt et al., 2017;Firoz et al., 2018).
Also the spatial distribution and coverage of hydro-climatic observations used to detect drought anomalies 75 are of key importance for hazard assessment. During drought, topography, geology, soil and land-cover catchment characteristics as well as human water interventions influence hydrological processes, catchment storage and release and therefore play a major role in the evolution of low flows (Bruijnzeel, 2004;Calder et al., 2007;Birkel et al., 2012;Stoelzle et al., 2014;van Loon and Laaha, 2015;Van Loon et al., 2016).
Altogether these influences cause a strong variability of climatic and hydrological drought hazard in tropical 80 space (Nauditt et al., 2019b).
Daily time step data, needed to effectively evaluate drought hazard in tropical catchments, are rarely available. Sheffield et al. (2018) highlight the potential of satellite remote sensing and reanalysis data products to improve water resources management in regions with sparse in-situ monitoring networks. Open 85 access high resolution remote sensing data products are continuously increasing in quantity (AghaKouchak et al., 2015;Mariano et al., 2018). In this context, a variety of gridded datasets are available, including daily precipitation (Funk et al., 2015;Baez-Villanueva et al., 2018&2020), surface water (Beck et al., 2016), groundwater (Thomas et al., 2014), reservoirs (AghaKouchak et al., 2018), soil moisture (Samaniego et al., 2018;Tijdeman and Menzel, 2020), and vegetation (Pinzón and Tucker, 2014;Nguyen et al., 2019). 90 Especially vegetation condition indices like fAPAR, NDVI, EVI and VCI play an increasing role for drought monitoring and research in data scarce regions. They can provide spatially distributed information on soil and vegetation moisture anomalies on the ground (Heydari et al., 2018;Recuero et al., 2019) that is not dependent on sparsely monitored hydro-climatic data.
Exposure and vulnerability information are also sparse, especially in rural tropical regions. Vulnerability evaluation should be ideally based on historical drought impact data Bachmair et al., 2016;Blauhut et al., 2016), but these are usually not systematically monitored and recorded; rare examples being the US Drought Impact Reporter (droughtreporter.unl.edu) , the European Drought Impact Database  or observer-based systems such as the Czech INTERSUCHO (www.intersucho.cz). 100 Alternatively, vulnerability data is often replaced by exposure related information (Carrão et al., 2016;Naumann et al., 2018;Naumann et al., 2019), that is available as gridded socioeconomic data sets showing the spatial distribution of population-, livestock-and crop densities as well as socioeconomic, demographic and infrastructural characteristics. Such remote sensing and gridded data-based drought risk assessment approaches have often been carried out at global or regional scale (Carrão et al., 105 2016;Hagenlocher et al., 2019), but have rarely been applied to local and catchment scale drought risk. This study evaluates the performance of gridded datasets related to hydro-climatic and socio-economic information to derive relevant drought risk information for catchments of different sizes (between 5 450 and 49 382 km²) and differing tropical climates.

110
In line with the above, the overall aim of this study is to identify and characterize drought risk hotspots in rural and data scarce tropical regions as a basis for drought management.

Specific objectives:
to identify the spatially distributed cumulative duration of hydrological and meteorological drought hazard 115 -to understand spatially varying and sector related drought vulnerability to visualize spatial distribution of drought risk in four tropical catchments that vary in size, topography, climate and water infrastructure development to attribute the relative spatial contribution of hazard and vulnerability related factors to drought risk 120 https://doi.org/10.5194/nhess-2020-360 Preprint. Discussion started: 14 November 2020 c Author(s) 2020. CC BY 4.0 License.

Study regions
We selected four rural catchments in tropical regions that were affected by severe drought disasters during the last decade. Figure 1 gives an overview on the characteristics of the four study regions, each differing in size, topography seasonality and level of human intervention. 125

Discharge data set: Hydrostreamer
Available discharge observations data in the study regions ( Figure 1) do not allow to display the spatial 135 variability in hydrological behaviour. We applied a recently developed downscaling tool, Hydrostreamer (Kallio, 2020) to the spatially coarse global discharge data product from the ISIMIP 2a (Gosling et al., 2017) experiment. Downscaling is carried out by areal interpolation, where the source runoff data are distributed to intersecting higher-resolution catchments, routed downstream, and optimized against observed streamflow (see detailed description in Kallio et al., 2019&2020. For the Muriaé catchment, SWAT2012 modelling 140 (Neitsch et al., 2011) provided 93 simulated discharges used for optimization (Nauditt et al., 2019b).
We requested available daily discharge observations for the optimization of the Hydrostreamer results. For the Muriaé, five daily discharge time series were obtained by the National Water Agency of Brazil (ANA, https://doi.org/10.5194/nhess-2020-360 Preprint. Discussion started: 14 November 2020 c Author(s) 2020. CC BY 4.0 License. 2019; Nauditt et al., 2019a;Nauditt et al., 2019b). For the Tempisque, data from two discharge stations were acquired by the Hydrological Department of the Electricity Institute of Costa Rica ICE (2019) 1980-2003 for the period 2003-2018 by the Institute for Aqueducts and Sanitation (AYA, 2019). For the Upper Magdalena, we obtained daily time series from 46 discharge stations from IDEAM (2019) and for the Srepok, daily data from three discharge stations (Kallio et al., 2019;Kallio, 2020;MRC, 2018) were acquired.

Precipitation
Station data that were used to validate the satellite based precipitation estimates for the Magdalena and the  (Nauditt et al., 2019b). 155 CHIRPS v2.0 showed good goodness-of-fit (GOF) performance in point to pixel evaluation and HBV rainfall runoff modelling (Baez-Villanueva et al., 2018;Nauditt et al., 2019b;Venegas-Cordero et al., 2020).
Additionally, CHIRPS v2.0 covers the longest time period (1981 to date) and has a higher spatial resolution compared to other available satellite based precipitation estimates products.

Vulnerability
We evaluated gridded data sets in terms of their suitability to represent vulnerability and selected the following data sets available for all our study regions:  (Robinson et al., 2014) Global Agricultural Lands 2000 cropland density (Ramankutty et al., 2008) GHS Population Grid 2015 population density (CIESIN, 1997(CIESIN, -2020 Major roads 2013 proximity to infrastructure (CIESIN, 1997(CIESIN, -2020 Global GDP PPP/HDI 2011 GDP  170 https://doi.org/10.5194/nhess-2020-360 Preprint. Discussion started: 14 November 2020 c Author(s) 2020. CC BY 4.0 License. Figure 2 illustrates the overall methodology applied in this study. We evaluate drought risk as a combination of hazard and vulnerability:

Drought Risk Assessment
Where represents drought risk, ℎ drought hazard, drought vulnerability and grid cell.
Hazard ( ℎ ) is defined by drought in meteorological, hydrological and vegetation condition indices.
Vulnerability ( ) is defined as the potential of a drought to cause damages in selected socioeconomic sectors using typical exposure information.
We used two groups of variables (hydro-climatic and socio-economic) and calculated index values for each 180 grid cell (i) for each variable. All layers were resampled to a spatial resolution of 30 m and equally weighted to obtain maps for each index, hazard, vulnerability and risk. Drought risk maps were then produced by equally weighting the hazard and vulnerability layers. More details about the methodological process are given in sections 2.3.1 and 2.3.2.

Drought Hazard
To obtain daily scale hydrological drought signals, we applied the widely used threshold method (e.g. Tallaksen, 2000) using a daily varying Q95 threshold (Fleig et al., 2006;WMO, 2008). We selected the period 1981-2018 that corresponds to the record length of CHIRPS v2.0 data. We defined more or equal than 12 days below a daily varying Q95 threshold as a long hydrological tropical drought event (ℎℎ ) and 5-11 190 days below that threshold as a short hydrological tropical drought event (ℎℎ ℎ ). We used pooling to remove single days when streamflow went above the threshold by less than 20 %. Resulting short and long hydrological drought indices (ℎℎ ) were derived as the cumulative drought duration of events for each grid cell. So (ℎℎ ℎ ) is the sum of all short-duration (5-11 days) events and (ℎℎ ) is the sum of long-duration (>=12 days) events. The cumulative duration of detected events was classified into five severity categories 195 (Sc). More than 75 short drought events during the period of 37 years were classified as the most severe https://doi.org/10.5194/nhess-2020-360 Preprint. Discussion started: 14 November 2020 c Author(s) 2020. CC BY 4.0 License.
short drought hazard and more than 50 events with more or equal than 12 days below Q95 were considered the most severe long drought hazard (Sc 5) ( Table 3).
The meteorological drought index ( ℎ ) evaluates the cumulative drought duration of precipitation drought events. To represent long and short meteorological drought events in tropical regions, we defined two classes 200 of drought intensity for precipitation deficits: >= 20 days with rainfall below 0.3 mm as a long meteorological drought: ( ℎ ) and 5-19 days as a short meteorological drought: ( ℎ ℎ ) with rainfall below 0.3 mm.
The number of detected events were classified into 5 severity categories (Table 3) The vegetation condition related drought hazard vc is represented by the vegetation condition in the driest month in record. We identified the driest month in record using the SPI. To understand the spatial variation 205 of vegetation condition, we used the Vegetation Condition Index (VCI) (Dutta et al., 2016;Kogan, 1995;Quiring and Ganesh, 2010) applied to NDVI imagery. The vegetation related drought index (vci) was established by inversely rating VCI values for each pixel. In contrast to the hydrological and meteorological indices, vci has a negative correlation with drought severity. Values between 50 % and 100 % indicate moisture rich vegetation conditions, values between 50 % and 35 % short drought conditions and values 210 below 35 % long drought conditions (Kogan, 1995). The detailed methodology is described in Nauditt et al., 2019b. VCI percentage values were classified into five severity categories (Table 3).
The overall drought hazard (dh) for each grid cell (i) is calculated by the equally weighted severity class (Sc) values (Table 3) of each hazard index as: Where dh is the drought hazard, i the location (grid cell) and Sc the severity class. ℎℎ ℎ represents the cumulative duration of short hydrological drought events based on number of events, ℎℎ the cumulative duration of long hydrological drought events, ℎ ℎ the cumulative duration of short meteorological drought 220 events, ℎ the cumulative duration of long hydrological drought events and the vegetation condition related hazard (Table 3).

Drought Vulnerability
We used open access gridded datasets for five socioeconomic exposure related variables to represent spatial drought vulnerability in the four study regions. All datasets were resampled using the nearest neighbor method to account for differences in grid cell resolution. Each data set was reclassified and given a rating based on positive or negative correlation to vulnerability. The overall drought vulnerability dv for each grid 230 cell i is calculated by the equally weighted severity class Sc values (Table 4) of each vulnerability index as: Where is overall drought vulnerability per grid cell, Sc the severity class, li the livestock density index, ci the crop density index, pi the population density index, ii the index for proximity to infrastructure and GDPi 235 the GDP index per grid cell. Table 4 gives an overview on the severity classification for each index.

Drought hazard (dhi), vulnerability (dvi) and risk (dri) in the four study regions 240
in Sc 5 and 31.1 % in Sc 4) as well as for proximity to infrastructure (ii) (43.5 % in Sc 5 and 25.7 % in Sc 4).
The remaining indices showed a nearly homogenous distribution across the severity classes.

Figure 6 Spatial distribution of drought vulnerability (dv) severity classes in the four study regions 310
https://doi.org/10.5194/nhess-2020-360 Preprint. Discussion started: 14 November 2020 c Author(s) 2020. CC BY 4.0 License. Figure 7 illustrates the spatial distribution of drought risk (dr) in the four study regions based on equally weighted hazard and vulnerability severity values.

Hazard (dh), vulnerability (dv) and drought risk (dr) in the four study regions -plausibility of identified drought risk hotspots
For the Muriaé, plausible results were obtained for the location of drought risk hotspots as well as for the spatial distribution of severe drought hazard and vulnerability. Risk hotspots were found in the downstream 325 area where most economic activities take place and precipitation rates are lower compared to those at higher elevations (CEIVAP, 2015;Nauditt et al., 2019b), along with a higher hydrological hazard also due to fractured geological and alluvial characteristics. Vulnerability is high all over the basin due to intensive livestock grazing and agricultural production as well as low GDP and large distances to road infrastructure.
These spatial characteristics for drought hazard, vulnerability and risk were confirmed by collaborating 330 stakeholders of the river basin committee CEIVAP (Comitê de Integração da Bacia Hidrográfica do Rio Paraíba do Sul) and the executive river basin agency AGEVAP (Agência da Bacia do Rio Paraíba do Sul).
Additionally, results coincide with field research, modelling and data analysis related to spatial variability of drought occurrence and impacts of involved and affiliated scientists (Nauditt et al., 2019a).

335
Also the locations of drought risk hotspots in the Tempisque were well defined by the analysis. Hotspots Institute NAWAPI).

Drought hazard assessment
In the Muriaé, severe hydrological hazard ℎℎ ℎ was found in the Northeastern agricultural upstream area, along the streams and at the river basin outlet with 20 % in Sc 5 and another 20% in Sc 4. During the longterm drought in 2014-2015, the river stretch near the catchment outlet station fell dry (Nauditt et al., 2019b;380 Ribbe et al., 2018), with impacts on aquatic and riparian ecosystems and water users. This event was aggravated due to the fractured geological and alluvial characteristics of the downstream river bed (Nauditt et al., 2019b). Cumulative duration of ℎℎ with 12 or more days however, was only present in 14.1 % of area in Sc 4 and 1.8 % in Sc 5 at the aforementioned basin outlet. The Northeastern agricultural upstream area in Minas Gerais State is prone to ℎℎ ℎ , most probably due to smaller catchment areas and fast 385 response to rainfall deficits.
Hydrostreamer provided excellent spatially distributed discharge simulations for the Muriaé catchment, as validated by station data and SWAT2012 modelling results for 93 stations; very valuable for drought and water resources management and planning. In the Southwestern downstream part severe meteorlogical hazard (Sc 5) was found for ℎ ℎ , ℎ for most grid cells. Low values were found for the upstream 390 region (Sc1 for most grid cells).
is following this spatial pattern with low hazard in the upstream and livestock and milk production is the main economic activity (Fischer et al., 2018).
Hydrological drought hazard was found to be most severe along the upper streams of the Tempisque from 400 which irrigation water is abstracted. This shows that our approach to assess hydrological hazard with a daily varying threshold is also suitable for anthropogenically intervened catchments, where human abstractions are leading to discharge anomaliesmost probably increasing as a response to a meteorological drought.
23.1 % of the basin area experienced more than 40 long drought events that lasted longer than 12 days (Q95 https://doi.org/10.5194/nhess-2020-360 Preprint. Discussion started: 14 November 2020 c Author(s) 2020. CC BY 4.0 License. of daily varying discharge and below) and 10.5 % of its area was affected by 60 moderate drought events in 405 the time period between 1918 and 2018. This shows the extreme low flows (down to 2.6 m³/s at Guardia station) the drought prone region is facing. A drought threshold of Q95 can therefore be considered as too low for a stream with a mean annual discharge of 27 m³/s in the case of the Guardia Station.
Strong spatial variation in meteorological drought hazard was found for ℎ ℎ and ℎ with most values in Severity Class 2 (52%). This might be due to data uncertainties in the hydrostreamer dataset and the underlying observed discharge data (Rodríguez et al., 2020). Hydrostreamer yielded in poorer performance compared to the other three study regions.
The Southwestern upstream region showed strongest meteorlogical hazard (Sc 5) for ℎ ℎ and ℎ followed by the Northeastern downstream part and similar spatial patterns for vci. The Southwestern 425 upstream region is exposed to a more marked tropical seasonality with two wet periods (April and May and October and November) and two long dry periods (June to October and November to April) (Rodríguez et al., 2020) while the lower part of the Magdalena receives more precipitation and is not exposed to such a marked seasonality.

430
For the Srepok basin, hydrological drought hazard for both hhlong and hhshort was found to be most severe in the Vietnamese Southern upstream region in the Vietnamese highlands due to discharge alterations through hydropower operation, as well as in the Cambodian North-Central and Northwestern downstream region due to abstractions from agricultural activities.
Good results for Hydrostreamer downscaling results were obtained for the Srepok, being a study region of 435 Kallio et al. (2019Kallio et al. ( & 2020, providing a valuable discharge data set for water resources modelling, management and planning in the transboundary basin and the Mekong region.
Meteorological hazard was strong with 55.9 % in Scs 4 and 5 for ℎ , with all grid cells located in the Vietnamese Southeastern upstream part. ℎ ℎ was only detected in Scs 1-3, indicating that periods without rainfall of shorter duration during the wet period were less frequent. , in contrast, is homogenously 440 distributed all over the basin.

Vulnerability assessment
We applied open access gridded data sets to evaluate their suitability to provide drought vulnerability or exposure information for all of the four study regions. For all study regions, our indicator population density (pi) showed few grid cells with severity values higher then Sc 1 (classified as less than 50 persons per 1 km² 445 grid cell). This suggests that the classification we chose (Table 4), assuming that >50 persons would represent small settlements and agricultural communities in rural regions, might not be adequate. The number of persons per km² classified as vulnerable could be lower to also detect remote farmers. In contrast, low GDP in rural areas showed strongest severity (Sc 5 = < 1 million USD PPP per km² for the reference year 2011) for almost all grid cells in all study regions, outweighing the low pi values. We used this 450 classification assuming a low GDP for rural agricultural regions; however, our results suggest that higher classification values for Sc 1 and Sc 2 in order to display differences in GDP. In most risk studies, several exposure and vulnerability indicators are aggregated, regionally masked (Naumann et al., 2014, Carrão et al., 2016Hagenlocher et al., 2019) to show overall vulnerability. Reference values for indicator classification at least to our knowledge --are not available in literature. Generally speaking, a more detailed evaluation 455 of gridded socioeconomic data in terms of risk indicator classification, e.g. for tropical agricultural regions resulting in reference values would be a valuable contribution to future comparative risk studies.
To evaluate drought exposure and vulnerability of agricultural activities, we tested the data set "Global Agricultural Lands in the Year 2000". The resulting crop density evaluation, similarly as population density, is mostly distinguishing between grid cells with agriculture and no agricultural use, therefore a low 460 classification values were used (Table 4) resulting in a good representation of agricultural exposure in the four study regions, as confirmed by affiliated scientists and stakeholders. More detailed local information on crop types, eg. distinguishing between perennial crash crops and annuals, irrigated or non-irrigated agriculture, could further detail such site-specific exposure information. Similarly for Livestock density related vulnerability, we used a low number of animals grazing per grid cell to determine the low Scs (Table 4). 465 Proximity to Infrastructure served as a good proxy for the stage of development of a location. Although drought vulnerability and exposure largely depends on storage infrastructure or irrigation systems, there are no available data sets for the regions addressed in this study. FAO AQUASTAT, for example, provides such data for Africa but not for Latin America and South East Asia. Despite these shortcomings, our overall drought vulnerability index dvi showed good results for aggregated vulnerability. The Muriaé had most grid cells in 470 high severity classes (73.1 % in Sc 4) due to its prevailing sectors milk production and agriculture as well as its sparse road infrastructure. For the Tempisque, we found fewerbut well distinguished and located --grid cells in Sc 4 (81.8 % in Sc 3); because of the existence of large National Parks in the Northeastern part, less cattle grazing and its well-developed road network. The Upper Magdalena, (with 43.1 % in Sc4) showed high dvi values in the downstream part mainly due to crop density and infrastructure. Only for the Srepok, less dvi 475 was found. This can be attributed to little livestock and crop density and larger forested areas compared to the other study regions. dvi was mainly found for the cultivated areas along the streams as in Vietnam (upstream Southeast), irrigated rice areas in the Cambodian Northeast, and in the downstream Northwest.

Conclusion
Droughts are causing severe damages to water abundant tropical countries worldwide. The implementation 480 of drought adaptation measures at the local scale need to be based on reliable information about spatially distributed drought risk --that is rarely available in data scarce tropical catchments. We propose a methodology for evaluating and mapping the distribution of drought risk for rural tropical regions, based on the combination of independent indicators of daily data based drought hazard and drought vulnerability.
We evaluated freely available gridded datasets regarding their suitability to assess drought hazard, 485 vulnerability and risk in four differing rural tropical study regions, the Muriaé river basin in South East Brazil, the Tempisque basin in Costa Rica, the upper catchment of the Magdalena river Basin, Colombia and the Srepok basin in Cambodia/Vietnam. We used daily scale meteorological and hydrological gridded data products and indices to evaluate tropical drought hazard, next to vegetation response to long term droughts, as well as vulnerability data related to the major sectors population, agriculture, livestock, infrastructure and 490 GDP that were available for Latin America and South East Asia. Results showed plausible spatial distribution of hazard, vulnerability and risk in the four study regions, as confirmed by local stakeholders, field surveys as well as through research of the authors. The following outcomes can be highlighted: -The hydrological drought index hhi, based on daily time series and a daily varying threshold Q95, was able to detect hydrological drought hazard in both, pristine and regulated streams, representing both 495 climate and human induced hydrological drought.
-The meteorological drought index mhi, based on daily precipitation data and periods of zero rainfall turned out to be suitable for tropical regions as shown by local impacts especially on livestock and rainfed agricultural production; -The subindices ℎ ℎ and ℎ and ℎℎ ℎ and ℎℎ give insights in the historical frequency of 500 long and short drought events, independent of general seasonal patterns.
-In combination with the above-described findings, the vegetation anomaly response (NDVI/VCI) to long term drought periods (SPI 12) reveals further vegetation, soil and groundwater related hazard, relevant for eg. forest fire related hazard.
-In light of the data scarcity in many tropical regions, the vulnerability related data sets and indicators 505 related to crop and livestock density as well proximity to infrastructure have an adequate spatial resolution to provide vulnerability information at the local scale.
-The individual hazard and vulnerability indicator results give insights on how the classifications can be further adapted to individual study regions, depending on climate, topography, seasonality and human influence. 510 -Drought hazard, vulnerability and risk maps and individual indicator maps provide decision support when selecting and designing drought adaptation measures to avoid future drought impacts.
These findings were discussed with representatives of local universities and public institutions working in the study regions, by looking at each indicator and combined hazard, vulnerability and risk. We recommend the replication of the approach in other tropical regions by using the developed R scripts. A further step will be 515 to make the Tropical Drought Risk Assessment R-Package available on CRAN. Potential modifications to adapt or further develop our approaches are described below. These depend on the location of the study region, basin size, data availability, hydro-climatic seasonality and major socioeconomic uses. Individual indices can be adapted by changing the severity classification (e.g. by increasing or reducing the duration of short and long droughts) or by introducing data sets that are not (yet) 520 available for the regions addressed in this study. Furthermore, locally defined ecological flow and water demand thresholds could feed into a re-definition of local hydrological drought hazard indices. In addition, drought risk scenarios can be developed, helpful to detect potential changes in future drought risk. These could be based on hydro-climatic projections, indicating longer drought periods, or a changing vulnerability based on socioeconomic projections. 525 However, in its current composition and design, our approach delivers a plausible representation of spatially distributed drought hazard, vulnerability and risk hotspots in data scarce and rural tropical regions. It offers a holistic, science based and novel solution to generate local drought risk knowledge that can feed into future https://doi.org/10.5194/nhess-2020-360 Preprint. Discussion started: 14 November 2020 c Author(s) 2020. CC BY 4.0 License. drought related research. The outcomes produced are a valuable source of information for regional planners and water managers that take decisions on infrastructural and drought adaptation measures. 530

6
Code availability R package is under preparation to be published on GitHub and possibly in Cran, in the meanwhile please contact the corresponding author.

Supplement link
Hazard and vulnerability indicator maps 535 8 Author contributions AN conceived and designed the study and overall methodology, performed the analyses and prepared the manuscript. KS contributed to the study design, methodology, manuscript writing and structure. ER, CR and RF supported field work, local validation of results and stakeholder communication in terms of research and information demand to support drought management and validation. KM simulated and evaluated high 540 resolution discharge data (hydrostreamer). HH, OB and JT supported methodological setup, R Coding, Package for GitHub and Cran and development of illustrations (R Graph Gallery), LR supported and accompanied the research by allocating time and financial resources for personnel, travels and field work.
All authors actively took part in the writing and editing process.

9
Competing interests 545 The authors declare that they have no conflict of interest.