A comprehensive record of flood events is available from the Dartmouth Flood Observatory (DFO) founded in 1993 at Dartmouth College, NH, United States. In 2010, the observatory moved to the Community Surface Dynamics Modeling System (CSDMS) (http://csdms.colorado.edu/, last access: 1 March 2016) as a division of the Institute of Arctic and Alpine Research (INSTAAR) at the University of Colorado, Boulder, United States . Information in this archive is based on instrumental measurements and remote-sensing sensors. These events are validated based on officially reported flood details by governmental and news agencies . The DFO mostly takes advantage of orbital remote-sensing sensors to identify, measure, and monitor global flood events by gathering globally consistent information on surface water changes, in particular since 1999. Floods are detected using MODIS (Moderate-Resolution Imaging Spectroradiometer) sensors (approximately 250 m footprint pixel), and river discharges are measures using satellite microwave data such as AMSR-E (Advanced Microwave Scanning Radiometer for Earth Observation System (EOS) from Global Change Observation Mission – Water, GCOM-W). The discharge values and runoff coefficients are then calculated from the Water Balance Model (WBM) embedded with the specific soil type, surface gradient, soil permeability, and land use–land cover (LULC) characteristics. These remote-sensing and model outputs are employed conjunctively to map the potential flood inundation extents frequently. Then, a number is assigned to the flood if (a) it is unusually “large” compared to the typical annual high water and previously mapped water–land extents, and/or (b) if there are significant damages caused to the structures, extensive land inundation, and fatalities .

It is important to note that the quality of data has improved in recent times. The improvements in the level of media reporting and information quality have improved the reliability of the data. At the same time, the likely improvements in the accuracy of in situ measurements, advances in satellite and ground-based sensors, data storage, and transfer facilities also contributed to the data quality. Moreover, have discussed that the frequent temporal sampling of satellite-based observations and ground sources (media reporting) determines the accuracy level amongst the (non-)flood event candidates. The dataset covers flood events at the global scale from 1 January 1985 to present. Any recent flood event is added immediately to the data archive. In this study, we considered 31 years of global flood events from 1 January 1985 to 31 December 2015. This comprehensive dataset includes information on the location of the flood events (longitude, latitude, and the name of the country), flood beginning and end dates, their duration (which is the number of days between the flood beginning and end dates), and damages due the flood (which is an estimation of flood-induced damage according to all the relevant sources). It is reported by the DFO that occasionally when there is no flood beginning date mentioned in the news report, they assume middle of the month as the start date (http://floodobservatory.colorado.edu/Archives/ArchiveNotes.html, last access: 1 March 2016). We verified the fraction of such events among the total events and found that less than 5 % (194 out of 4311 events globally over the 31 years) have such an assumption. We also explored the distribution of the month of occurrence for the flood beginning and end dates across the globe. While investigating these records, it become apparent that the days 1, 5, 10, 15, 20, and 25 also have an increased number of flood counts (around 4 %), which suggests a reporting bias or a phenomenon of rounding off to the nearest fifth day. While there was information on the phenomenon of rounding off to the middle of the month in the DFO data description, we did not find any relevant information on the pattern every fifth day. However, we did not find any systematic spatial pattern for these apparent reporting biases (see Appendix B for more details). The DFO is the only global dataset of observed flood events. Many of the prior studies either focused on rainfall-based datasets or model-based river flow data. In this regard, the present study adds a new dimension to the flood literature, especially the understanding of the long-duration floods at the global scale.

The flood events are spatially aggregated to five climate zones – tropics (23.5∘ S to 23.5∘ N), Northern Hemisphere subtropics (23.5–35∘ N) and midlatitudes (35–55∘ N), and Southern Hemisphere subtropics (23.5–35∘ S) and midlatitudes (35–55∘ S) (Environmental Literacy Council, ). We chose these spatial aggregations along the latitudinal belts to be consistent with the global circulation dynamics, zonally symmetric thermal forcing , temperature variabilities, and precipitation patterns . In addition, such specifications will result in achieving higher coherency in satellite-based data acquisition in particular for the passive sensors because of varying solar reflectivity and ascending–descending satellite orbits along different latitudes . Figure 1 represents the schematic of the five climate zones. We also show the geographical locations of four countries (United States, China, India, and Thailand) that have already experienced high rates of long-duration floods among all the countries from 1985 to 2015.

Next, for each latitudinal belt, the total number of floods per year (calendar year from 1 January to 31 December), the duration of these floods, and their location (name of country) are processed. This procedure is formulated as follows: FCt,r=totalnumberoffloodevent(s)inlatitudinalbeltrandyeart[count(s)], FDt,r=duration(s)offloodevent(s)inlatitudinalbeltrandyeart[day(s)], FLt,r=location(s)offloodevent(s)inlatitudinalbeltrandyeart[nameofcountry(ies)], where FC indicates the flood counts (frequency), and FD and FL denote the vectors of flood duration and flood location for each of these flood events, respectively. The superscripts r and t denote the latitudinal belt (r = {global, tropics, midlatitudes (N and S), subtropics (N and S)}) and year (t = {1985, 1986, …, 2015}).

In addition, the number of floods in each latitudinal belt are also categorized in terms of their duration. We denote the event as a short-duration flood FCSt,r if the duration is between 1 and 7 days, moderate-duration flood FCMt,r if the duration is between 8 and 20 days, and as a long-duration flood FCLt,r if the duration is greater than or equal to 21 days. These categories are also consistent with the DFO's flood classification . The subscripts “S”, “M”, and “L” stand for short-, moderate-, and long-duration flood events, respectively.

We used large-scale ocean–atmospheric teleconnections to investigate the extent to which the trends in the floods can be related to natural variability in the climate–atmospheric system. Since the climate system has as quasi-periodic nature that often manifests as wet and dry regimes, it is important to understand whether the trends, if observed, can be attributed to these natural oscillations. Hence, we used the El Niño–Southern Oscillation (ENSO), Pacific Decadal Oscillation (PDO), North Atlantic Oscillation (NAO), and Atlantic Multidecadal Oscillation (AMO) as proxies for interannual, decadal, and multidecadal climate variability.

We obtained 31 years (1985–2015) of ENSO data (aggregated based on the monthly anomalies of Niño 3.4) from the HadISST1 dataset . Monthly AMO and PDO anomalies are obtained from the NOAA/Earth System Research Laboratory at http://www.esrl.noaa.gov/psd/data/climateindices/list (last access: 1 March 2016) , and then averaged to yearly time series from 1985 to 2015. Similarly, the monthly NAO indices are obtained from the NOAA/National Weather Service, Climate Prediction Center at http://www.cpc.ncep.noaa.gov/products/monitoring_and_data/ (last access: 1 March 2016) and averaged to yearly time series.

In addition to the frequency of the floods (FCt,r), we calculate a set of “resistant measures” to evaluate the existence of any significant monotonic time trend in the probability distribution of flood duration. Four moment indicators are selected because of their scale-invariant characteristics suitable for such asymmetric distributions. These metrics include the median, median absolute deviation (MAD), resistant skewness, and the 90th percentile of the distribution of flood durations in each year. Each of these metrics is computed as a time series of 31 years (1985–2015) for each of the six spatial scales (i.e., global, tropics, midlatitudes – N, midlatitudes – S, subtropics – N, subtropics – S). It is straightforward to calculate the median and 90th percentile from the distribution of flood duration each year. We explain the formulation and the properties of the other two metrics here.

For a specific country, we calculate the relative flood frequency of short, moderate, and long durations with respect to the total flood events occurring in that country. This can help us identify what flood duration class has occurred more frequently from 1985 to 2015 in that country. Correspondingly, the reported flood damage for that event has also been noted along with its relative damage in reference to the total flood damages in that country from 1985 to 2015.

In order to investigate the association between flood duration and damage at the country scale, we present a linear model for flood damage (Fdamage) as a function of flood duration (FD) in the log space as follows: Fdamage=αFDβ⟹log⁡Fdamage=log⁡(α)+βlog⁡FD, where α and β are the intercept and scaling exponents, respectively, of flood damage for a specific country. The parameter β in this formulation captures the change in flood damage due to changes in flood duration.

Most of the global precipitation studies indicate that there is a recent increase in both the annual precipitation and extreme rainfall intensities . Consequently, our goal here is to investigate whether we see a significant trend in the frequency and duration of floods during the last 3 decades. Based on this, the main hypotheses (H1, H2, H3, and H4) and the evaluation procedure are presented in Table 1.

We begin our investigation with H1, the hypothesis that there is no monotonic trend in the annual frequency of the flood events. We test this hypothesis using the Mann–Kendall (MK) trend test . The MK test uses the ranks of the data and assumes no underlying probability distribution . The test statistic is based on a pairwise comparison among the values and is independent of the distribution of the original series. The magnitude of the slope of the trend is estimated using the method of Sen, the median of the pairwise slopes among the elements of the series . Ties in the data are adjusted using an assumption that the number of ties is equal to an even number of positive and negative differences . Statistical significance is evaluated at the 5 % significance level, the probability of incorrectly rejecting the null hypothesis.

In hypothesis H2, we explore whether there is a change in the probability distribution of the flood duration over time. We test this hypothesis by applying the MK trend test on the three resistance moments (median, MAD, and skewness) and the 90th percentile (extreme flood duration) of the annual distribution of the flood duration. H3 is intended to investigate the changes in the patterns of flood frequencies for each category: short-, moderate-, and long-duration floods. Lastly, in H4, we investigate the potential large-scale atmospheric teleconnections to which the observed trend(s) in H1 and H2 can be related by using a GLM framework.

Our hypothesis (H4) is that the detected time trend is due to cyclical climate influences (i.e., oscillatory behavior) associated with the large-scale ocean–atmospheric interactions. Hence, for all the cases in which the null hypothesis of no trend is rejected, we attempted to understand whether the trend relates to large-scale climate oscillations. For this purpose, we employed a GLM framework on the time series of the above-developed metrics with ENSO, AMO, PDO, and NAO as covariates. GLMs are the mathematical extension of classical linear regression models to include a broad class of model assumptions such as linear, Poisson, exponential, log-linear, and so on with specified link functions . For all the spatial scales at which we see a statistically significant trend, a GLM is fit to the time series (1985–2015) of FC, FDMedian, and FD90 with climate covariates. FC=a+b1ENSO+b2AMO+b3PDO+b4NAO,FDMedian=a+b1ENSO+b2AMO+b3PDO+b4NAO,FD90=a+b1ENSO+b2AMO+b3PDO+b4NAO, where a, b1, b2, b3, and b4 are the GLM's coefficients (parameters). We then select the best model using the forward and backward stepwise regression and obtain the residuals of the best model in each case. The residuals represent the values for FC, FDMedian, and FD90 after adjusting for the influence of exogenous variables. In other words, they reveal the variability beyond what could be attributed to exogenous climate factors. The analysis of the time trends in the residuals will help discern any unexplained trend after accounting for background variability due to the climatic modulation (e.g., ). The models are fit using the “stepwiseglm” toolbox in MATLAB 2017a that uses the forward and backward regression algorithm. We used the deviance information criterion for the best model selection among a finite set of models. Results from the models are presented in Sect. 4 in which we discuss the associations.

The MK test (Eqs. A1–A3) is applied to each time series of FC (i.e., global, tropics, midlatitudes – N, midlatitudes – S, subtropics – N, and subtropics – S) for the detection of monotonic trends. Figure 2 presents the time series of FC for the global scale and the five latitudinal belts. A solid LOESS (local regression) curve is shown if the trend is significant. Alternately, a dashed LOESS curve is shown for the time series that do not exhibit a statistically significant trend. The detailed statistics derived from the trend analysis are given in Table 2.

A total of 4311 flood events occurred during last 3 decades worldwide. The results of the MK test on the annual frequency of global floods indicate that there is a statistically significant monotonic trend with τ (Kendall correlation coefficient between FC and time) and β (robust Sen slope) values of 0.26 and 2.12, respectively. A total of 2020 events (out of the 4311 floods) occurred in the tropics. The hypothesis that there is no trend in the frequency of floods in the tropics is rejected. This is also the case for both the subtropics (S) and midlatitudes (S). However, while we see an uptrend in the number of floods in the midlatitudes (S) post-2000, we urge caution in interpreting this trend as zeros dominate the time series. Finally, for both the subtropics (N) and midlatitudes (N), the hypothesis that there is no trend in the annual frequency of floods cannot be rejected at the 5 % significance level.

H1. There is a statistically significant increase in the frequency of floods at the global scale, and over the tropics, subtropics (S), and midlatitudes (S). The temporal pattern of the data for global floods resembles that of the tropics.

The MK trend tests are performed on the time series of the median, MAD, resistant skewness, and the 90th percentile of the flood duration. The following four subsections elaborate the results for each metric.

Given that we find statistically significant trends in the tails of the distribution (90th percentile of the duration of floods), we were interested in exploring whether there would be a trend in the frequency of the long-duration floods as well. To investigate this, we performed the MK test on the frequency of long-duration floods (FCL) for the tropics, subtropics, and midlatitudes. We also performed these tests on short-duration flood frequency (FCS) and moderate-duration flood frequency (FCM). We present these results in Table 7.

As it can be seen from Table 7, there is no monotonic trend in the frequency of short-duration floods occurring across all the spatial scales, indicating that the number of short-duration floods has not changed significantly over the last 3 decades worldwide. However, this phenomenon is not true for moderate- and long-duration floods. In fact, the frequency of both moderate- and long-duration floods has increased in the tropics. There is also an increasing trend in moderate-duration floods in the subtropics (S) and long-duration floods in the midlatitudes (N). These findings are consistent with the results from H2, where we see a trend in the skewness and the tails of floods in these belts. An increase in the frequency of moderate- and long-duration floods will result in a shift of the quantile of flood duration distribution, thereby changing the skewness and the tails.

For the long-duration flood events in the tropics, the total number of events has increased from 60 before 2000 to 249 after 2000. Similarly, the total number of events in the midlatitudes has increased from 27 to 70 post-2000. In other words, there are 4 times more long-duration floods that occurred during the most recent 15 years than before the year 2000. The increase across the midlatitudes (N) is around 2.5 times pre- and post-2000.

H3. In summary, frequency of moderate- and long-duration flood classes has changed recently, but remains unchanged for the short-duration floods in all the latitudinal belts. The annual frequencies of moderate- and long-duration flood events have increased across the tropics and midlatitudes (N) (on the scale of 4 and 2.5 events per year, respectively) over last 3 decades.

There were 4311 flood events that occurred from 1985 to 2015 around the world. According to Tables 2 and 7, globally, the total number of short-, moderate-, and long-duration flood events was 2508 (≈ 59 %), 1151 (≈ 27 %), and 560 (≈ 13 %), respectively. In addition to the aggregate analyses at the latitudinal level, we also explored the country-scale vulnerability to short-, moderate-, and long-duration floods. We interpret vulnerability as the expected value of the damage due to floods, i.e., the severity of the consequence of the floods .

For this purpose, we first excluded countries which had less than 31 flood events to ensure that we investigate only those counties that have experienced at least one flood per year on average. This screening resulted in 28 countries with a minimum of 31 flood events during the last 3 decades. These 28 flood-prone countries are sorted as follows: the United States (388 events), China (344 events), India (226 events), Indonesia (190 events), Philippines (181 events), Australia (121 events), Vietnam (107 events), Brazil (96 events), Bangladesh (88 events), Mexico (80 events), Iran (77 events), Afghanistan (74 events), Russia (69 events), Thailand (66 events), Pakistan (66 events), Nigeria (57 events), Malaysia (54 events), Kenya (49 events), Canada (48 events), Colombia (44 events), Peru (43 events), Turkey (41 events), Nepal (40 events), France (40 events), Romania (38 events), Ethiopia (35 events), Somalia (34 events), and New Zealand (31 events).

Then, the fraction of flood frequencies for each country and duration class – short, moderate, and long – is calculated. Figure 7a presents these fractions for the 28 countries using the ternary plot. For 23 of these countries, we have the data on the damages due to the floods. We computed the expected value of the damages for each country and plotted the fractional damage due to short-, moderate-, and long-duration floods as the second ternary plot in Fig. 7b. The color bars indicate the total number of events (Fig. 7a) and the total flood damage (Fig. 7b). In each plot, the location of the country shows the relative fraction of short-, moderate-, and long-duration flood frequency and damage. For example, in Fig. 7a, the United States is identified as a red circle in the top corner with > 60 % of floods being short duration, between 20 and 30 % of the floods being moderate, and only < 10 % of them being long-duration floods. However, in terms of the vulnerability to floods (Fig. 7b), the United States is located in the bottom right corner of the triangle, indicating that most of the vulnerability is due to low-probability long-duration floods. Similar observations can be made for Vietnam, Mexico, Indonesia, Australia, and Malaysia, to name a few. These countries have a very low probability of long-duration floods, but the consequence of these floods is the most important in terms of the vulnerability. It is also noteworthy to emphasize that for most of the countries, the overall damage is dominated by the damage due to moderate- and long-duration floods. This can be seen from the fact that many of the countries are found in the bottom left and right corners of the ternary plot.

To further understand the relation between flood duration and flood damage, we fit log-linear models given in Eq. 6 for four selected countries: the United States, Thailand, India, and China. The results of the log-linear models for these four countries are shown in Fig. 8a. These countries are selected because they have the highest number of long-duration floods among all countries (Fig. 8b). Parameter β is the scaling exponent of the damages to the flood duration. Note that the scaling exponent is similar for the United States (0.89) and China (1.03) while India (0.23) and Thailand (0.56) have much smaller exponents. In total, 226 flood events occurred across India, of which around 43, 32, and 25 % of them were short-, moderate-, and long-duration events, respectively. In the United States, short-, moderate-, and long-duration flood events account for 66, 26, and 8 % of 388 flood events that occurred in the last 3 decades. However, the fraction of long-duration flood events is much higher for Thailand (30 % of total flood events). In China, around half of the flood events were related to the moderate- or long-duration flood classes (34 and 16 %, respectively). This opens up new questions about whether there are consistent relations like this across the globe and how different these scaling exponents would be. We do not pursue them as part of this investigation, however, in the spirit of examining flood duration and damages, in Fig. 8b and c, we present the data on flood duration and flood damage ranked for various other countries.

According to the DFO flood data from 1985 to 2015, the ranking results show that the frequency of short-duration floods for the United States, China, India, and the Philippines is respectively 255, 173, 133, and 122. For moderate-duration floods, the countries of China, the United States, India, and the Philippines have experienced 118, 101, 74, and 52 flood events, respectively. The long duration floods were seen mostly in India (55 events), China (53 events), the United States (32 events), and Thailand (20 events) from 1985 to the end of 2015. It should be noted that here we only presented the top 21 countries in each category.

As discussed in this section, the consequences of floods of different durations should be paid attention to, as this plays a big role in designing appropriate flood-proofing infrastructure and developing early warning systems and flood insurance payout structures. The relation between the duration of floods and the induced damages, and how they might vary across different countries, was also investigated here.

The flood archive data provided by the DFO have been collected using different methods of observation and validation since 1985 (see the summary of the methods in ). In addition, there are more flood warning systems and facilities, transmitting instruments, reporting networks, and communications today at different levels of social and governmental divisions that the DFO is using to provide more comprehensive flood information. They have improved their flood detection methods by including the MODIS products since 1999. MODIS products contain surface inundation information based on vertically and horizontally polarized backscatter acquired remotely from the radiance changes among water-, land-, and vegetation-covered surfaces . We acknowledge that there could be some uncertainties as a result of this since surface may also be interpreted as water in the presence of clouds, cloud shadows, and mountainous terrain . Further, inclusion of this improved technology will result in better monitoring of floods. This improvement is likely a potential driver of trend in the flood duration. In our analysis of H3, we find that there is no significant trend in the frequency of short-duration events across all latitudinal scales (Table 7), but a significant trend can be seen across the tropics for moderate- and long-duration flood events. The introduction of improved satellite products would have increased the chances of detecting more short-duration floods (small events) along with providing better resolution for longer floods. We think that it is not possible to see the systematic contributions of such products into only one specified type(s) of flood duration.

To validate the DFO's flood statistics, we have corroborated the DFO floods with the available in situ streamflow observations from the GRDC (The Global Runoff Data Centre, 56068 Koblenz, Germany, 2013, http://grdc.bafg.de, last access: 1 December 2017). Among the stations that had matching time periods and locations, we found that a high percent of stations (≈ 90 %) have very few errors (i.e., less than 7 days) when their flood durations were compared. The results demonstrate that the recorded flood information in the DFO appears to be reliable with respect to the GRDC river discharge measurements (see more details in Appendix B). However, we did identify a reporting bias in the start and end dates of the floods where the events with no reported flood beginning date are assumed to start in the middle of the month. Some biases were also identified for days 1, 5, 10, 15, 20, and 25, which had an increased number of flood counts. These biases will potentially lead to estimation uncertainties in the trend model. We believe that interpretation at the global level will remain robust due to the effect of averaging and large numbers; at the local level interpretation needs more attention to such reporting details.

While understanding such uncertainties is essential, especially while interpreting trends in limited data, it is also documented in the literature that there has been an increased exposure to floods in recent times. The number of people, residential and industrial properties, and assets exposed to the flood events has drastically increased . The type of vulnerability to flood risk is mostly connected to development of the country and its land use and environmental management . Recent studies by and in Africa and the United States, respectively, showed that there had been a considerable change in the flood frequency and magnitude in regions which have undergone intense urbanization.

While exposure of people to floods is the main concern in developing countries, exposure of assets and properties to floods is the vital concern for the developed countries . Recently, much residential and industrial infrastructure has moved to the flat and cheap lands of floodplains . The nature of geomorphological features of land has been modified to embrace these new developments. and have recently indicated that the increase in the reporting of floods can be linked to the rise in the land use development in the floodplains.

The frequency of heavy precipitation events has increased at the global scale . Using daily precipitation observations from the Global Historical Climatology Network (GHCN) dataset, showed that the distributions of precipitation indices in the 1979–2003 period are significantly different from the 1901–1950 period with a tendency towards wetter conditions. , in the fourth assessment report of the Intergovernmental Panel on Climate Change (IPCC), discussed that the annual precipitation intensity has increased over high latitudes during the period from 1901 to 2005, except the southwest of the United States, northwestern Mexico, and the Baja California Peninsula. This IPCC report also highlights the increasing contribution of extreme rainfall events to the total precipitation across Europe and the United States, which mostly occurred during the last 3 decades of the 20th century. tested 8326 land-based rainfall stations (with at least 30 years of record from 1900 to 2009) and found that the annual maximum daily precipitation has significantly increased for more than two-thirds of these stations at the global scale.

Theoretical studies also discussed the fact that mean global precipitation intensity increased by 1–3 % (conditional on available energy budgets) in proportion to the 1 ∘C increasing rate of surface air temperature. , , , , and among others have also argued that an increase in air temperature will increase the atmospheric water-holding capacity (Clausius–Clapyron relationship), leading to more intense and frequent precipitation events. Hence, fluctuating precipitation regimes would interrupt the current balances of components within the hydrological cycle and human activities . Consequently, a warmer and wetter atmosphere is likely to intensify the global water cycle that ultimately will result in more frequent and larger flood events.

The space–time distribution of these precipitation regimes is potentially related to the large-scale ocean–atmosphere circulations driven by the natural climatic variability . Natural climate variability often causes periods of increasing extremes (flood-rich cycle) or decreasing extreme events (flood-poor cycle) depending on the phase of the climate .

Hence, in an effort to investigate any significant relationship between the observed trend in the flood data (characterized in H1 and H2) and the variability in the climate and atmospheric circulation patterns, we considered large-scale atmospheric teleconnections and climate indices (with quasi-periodicity in nature that can lead to wet–dry regimes) to explain the trend, i.e., to place the short-term trends within a longer climate variability context as argued by and .

To our knowledge, this study is the first analysis of global flood events that exclusively focuses on the variability in the flood duration using the DFO dataset over the last 3 decades (i.e., 1985–2015). In this part, we are corroborating the presented results here with the most relevant previous studies. A high number of recent flood studies have focused on the regional scale, and/or have used the flood duration to calculate the flood magnitude (i.e., log⁡ × duration × severity × affected area). For instance, analyzed the DFO data from 1985 to 2016 and concluded that there had been a slight increase in the flood severity in both India and Australia. Similarly, it was reported by that there is an increasing tendency in the number of floods with large magnitude and severity in Europe. These are consistent with our findings.

Several flood-related studies analyzed the trends in the annual maximum streamflow and/or precipitation across multiple spatiotemporal scales. For example, an increasing trend in annual maximum precipitation intensities was found by in addition to the increasing trend in the extreme precipitation at the global scale, but the catchment characteristics and river geomorphology can substantially regulate the streamflow regimes despite the intensified rainfall trends . Recently, used the Global Runoff Data Center (GRDC) database to investigate the potential trends in the annual maximum streamflow and found decreasing trends for many stations in western North America but increasing trends in eastern North America, some parts of Europe and South America, and southern Africa. A complete comparative analysis is required in this regard, especially to identify the DFO locations with the river basins and then analyze the trends in those river basins. We believe that this involves developing a separate study in the future.