Stochastic Rainfall Model Research Articles

STORM v.2: A simple, stochastic rainfall model for exploring the impacts of climate and climate change at and near the land surface in gauged watersheds

STORM v.2: A simple, stochastic rainfall model for exploring the impacts of climate and climate change at and near the land surface in gauged watersheds

A stochastic rainfall model for reliability analysis of rainfall-induced landslides

ABSTRACT Rainfall is an important trigger for a large number of landslides. Physics-based quantitative risk assessment of rainfall-induced landslides requires modelling of rainfall uncertainty in addition to soil parameters. Explicit modelling of rainfall uncertainty should consider the uncertainties in rainfall intensity, duration, as well as the occurrence of rainfall events. Towards this aspect, this study proposes a stochastic rainfall model, where the joint distribution of rainfall duration and intensity is constructed using copula theory and the occurrence of rainfall events is described using Poisson process. Based on the proposed stochastic rainfall model, a computationally efficient reliability method with a machine learning-based surrogate model is developed for assessing the failure probability of a slope subjected to rainfall infiltration within a given time period. An illustrative example with real rainfall data from Singapore is utilised to demonstrate the proposed approach. The results suggest that the slope failure probability is less sensitive to the uncertainty in rainfall occurrence, but highly sensitive to the adopted inter-event time definition for characterisation of the rainfall event. Overall, this study provides a useful stochastic rainfall model and some practical guidelines for quantitative risk assessment of rainfall-induced landslides.

Constructing multivariate distribution of rainfall characteristics: A Bayesian vine algorithm

Constructing a multivariate distribution model for rainfall characteristics is an important task for risk-based hydrologic design. However, most stochastic rainfall models are limited to bivariate frequentist distribution and cannot be easily generalized to multivariate (≥3) Bayesian settings. Towards this problem, this study presents a Bayesian vine algorithm. The presented algorithm translates the high-dimensional Bayesian update task into multiple lower-dimensional update tasks using the vine approach. The algorithm also computes the evidence, allowing Bayesian model selection of marginals and copulas in the joint distribution. Two numerical examples, a synthetic rainfall data from Canada, and a real rainfall data from China are utilized to demonstrate the algorithm’s capability. The results suggest that the algorithm can satisfactorily identify and characterize the underlying marginal and dependence for the rainfall characteristics. Since the high-dimensional Bayesian update task is translated into lower-dimensional Bayesian updates, the algorithm is scalable, stable, and computationally efficient in higher dimensions. A 20-year continuous hourly rainfall record can be characterized for its depth, duration, and intensity in around 60 s without any convergence or tuning issues. Overall, this study provides a powerful Bayesian tool for constructing stochastic rainfall models, which in turn can be used in multiple civil engineering disciplines concerned with rainfall interaction. This study does not consider temporal/serial dependence but only the cross-dependence among the rainfall characteristics.

Neglecting hydrological errors can severely impact predictions of water resource system performance

Risk-based decision making for water resource systems often relies on streamflow predictions from hydrological models. These predictions are integral for estimating the frequency of high consequence extreme events, such as floods and droughts. However, streamflow predictions are known to have errors due to various factors such as incomplete hydrological understanding, parameter misspecification, and uncertain data. Despite these errors being well known, they are frequently neglected when undertaking risk-based decision-making. This paper demonstrates that neglecting hydrological errors can impact on drought risk estimation for high stakes decisions with potentially severe consequences for water resource system performance. A generic framework is introduced to evaluate the impact of hydrological errors for a wide range of water resource system properties. This framework is applied in two Australian case study catchments, where we use a stochastic rainfall model, the GR4J hydrological model, a residual error model, and a simplified reservoir storage model to estimate water resource performance metrics (risk and yield). The results underscore the impact of neglecting hydrological errors on decision-making. In one case study catchment, the yield was over-estimated by ∼15%–55%, resulting in the (actual) risk of running out of water being ∼2–30 times larger than reservoir design. The magnitude of these errors in water resource performance metrics is striking, especially considering that the streamflow predictions appear reasonable based on typical performance metrics (e.g., NSE of ∼0.7). The errors in performance metrics stem from the complex propagation of hydrological errors through the water resource system modelling chain. By accounting for critically important hydrological errors we can mitigate highly erroneous risk estimates and improve decision-making related to water resource management.

Understanding hydrologic controls of sloping soil response to precipitation through machine learning analysis applied to synthetic data

Abstract. Soil and underground conditions prior to the initiation of rainfall events control the hydrological processes that occur in slopes, affecting the water exchange through their boundaries. The present study aims at identifying suitable variables to be monitored to predict the response of sloping soil to precipitation. The case of a pyroclastic coarse-grained soil mantle overlaying a karstic bedrock in the southern Apennines (Italy) is described. Field monitoring of stream level recordings, meteorological variables, and soil water content and suction has been carried out for a few years. To enrich the field dataset, a synthetic series of 1000 years has been generated with a physically based model coupled to a stochastic rainfall model. Machine learning techniques have been used to unwrap the non-linear cause–effect relationships linking the variables. The k-means clustering technique has been used for the identification of seasonally recurrent slope conditions in terms of soil moisture and groundwater level, and the random forest technique has been used to assess how the conditions at the onset of rainfall controlled the attitude of the soil mantle to retain much of the infiltrating rainwater. The results show that the response in terms of the fraction of rainwater remaining stored in the soil mantle at the end of rainfall events is controlled by soil moisture and groundwater level prior to the rainfall initiation, giving evidence of the activation of effective drainage processes.

Evaluating stochastic rainfall models for hydrological modelling

Stochastic rainfall models are important tools for evaluating hydrological risks such as flooding and drought because of their ability to randomly generate alternative plausible climatic timeseries. The stochastic generation of climatic timeseries is not an end in itself, since they are typically applied to a catchment to determine the performance of water-related infrastructure systems, such as reservoirs or flood-control measures. This methodology typically involves a train of models to determine the end-of-system impact, yet the evaluation of stochastically generated rainfall timeseries is usually a stand-alone procedure focused on metrics directly related to the stochastic generator. This paper demonstrates discrepancies in this approach by evaluating two, daily-timestep, stochastic rainfall models in terms of rainfall metrics and their subsequently generated flow metrics after rainfall-runoff transformation. The two models are a Markov-based model and a latent-variable model, where each model is calibrated and evaluated showing ‘overall good’ performance. Stochastically generated timeseries, alongside observed rainfall timeseries are inputted to a calibrated catchment model (GR4J) to derive daily flow timeseries. Whereas the rainfall metrics typically showed ‘good’ performance, streamflow-based metrics are not necessarily ‘good’. The procedure is repeated for 277 stations from Australia and 106 stations from the United States of America. Depending on the strictness of the flow-based comparison and region analysed, using the Markov-based model 12–26% of sites were classified as ‘poor’ performing, and 1%-9% of sites were classified as ‘poor’ using the latent-variable model. The results demonstrate that catchment-based performance of flow metrics is more holistic since it magnifies features of the rainfall not otherwise visible to rainfall-based evaluation.

A climate stress testing method for changes in spatially variable rainfall

With emerging evidence that anthropogenic climate change is affecting the spatial pattern of rainfall, it is important to understand how such changes affect a range of climate-sensitive systems. A method for generating multi-site rainfall sequences is presented that enables the exploration of changes in spatial rainfall patterns separately from various other potential changes (e.g. to the averages, extremes, seasonality), thereby enabling the separation of key drivers of change as part of climate stress tests. The proposed method is based on the ‘inverse approach’ to stochastic weather generation, and is evaluated through a case study with rainfall at 12 sites in the Barossa Valley, South Australia, using a multi-site latent variable stochastic rainfall model. The new method accurately captures (1) observed marginal attributes at each site (i.e. annual totals, seasonality, extreme rainfall, number of dry days, and intermittency), (2) spatial correlation between sites, and (3) observed attributes for areal average rainfall. We then apply the new method to stress test the response of streamflow in the region to changes in spatial rainfall patterns, quantifying how reductions in spatial correlation (while holding all other attributes constant) will lead to reduced mean annual flow and extreme flows, and increased number of low-flow days. The results show that not only can a broad range of rainfall properties (including both marginal and spatial attributes) be simulated using the inverse approach, but also that changes to the spatial properties of rainfall can have large implications on water resources systems, with the potential to amplify or dampen other changes to rainfall attributes.

Nonstationary Annual Maximum Flood Frequency Analysis Using a Conceptual Hydrologic Model with Time-Varying Parameters

Recent evidence of the impact of watershed underlying conditions on hydrological processes have made the assumption of stationarity widely questioned. In this study, the temporal variations of frequency distributions of the annual maximum flood were investigated by continuous hydrological simulation considering nonstationarity for Weihe River Basin (WRB) in northwestern China. To this end, two nonstationary versions of the GR4J model were introduced, where the production storage capacity parameter was regarded as a function of time and watershed conditions (e.g., reservoir storage and soil-water conservation land area), respectively. Then the models were used to generate long-term runoff series to derive flood frequency distributions, with synthetic rainfall series generated by a stochastic rainfall model as input. The results show a better performance of the nonstationary GR4J model in runoff simulation than the stationary version, especially for the annual maximum flow series, with the corresponding NSE metric increasing from 0.721 to 0.808. The application of the nonstationary flood frequency analysis indicates the presence of significant nonstationarity in the flood quantiles and magnitudes, where the flood quantiles for an annual exceedance probability of 0.01 range from 4187 m3/s to 8335 m3/s for the past decades. This study can serve as a reference for flood risk management in WRB and possibly for other basins undergoing drastic changes caused by intense human activities.

Comparison of rainfall generators with regionalisation for the estimation of rainfall erosivity at ungauged sites

Abstract. Rainfall erosivity values are required for soil erosion prediction. To calculate the mean annual rainfall erosivity (R), long-term high-resolution observed rainfall data are required, which are often not available. To overcome the issue of limited data availability in space and time, four methods were employed and evaluated: direct regionalisation of R, regionalisation of 5 min rainfall, disaggregation of daily rainfall into 5 min time steps, and a regionalised stochastic rainfall model. The impact of station density is considered for each of the methods. The study is carried out using 159 recording and 150 non-recording (daily) rainfall stations in and around the federal state of Lower Saxony, Germany. In addition, the minimum record length necessary to adequately estimate R was investigated. Results show that the direct regionalisation of mean annual erosivity is best in terms of both relative bias and relative root mean square error (RMSE), followed by the regionalisation of the 5 min rainfall data, which yields better results than the rainfall generation models, namely an alternating renewal model (ARM) and a multiplicative cascade model. However, a key advantage of using regionalised rainfall models is the ability to generate time series that can be used for the estimation of the erosive event characteristics. This is not possible if regionalising only R. Using the stochastic ARM, it was assessed that more than 60 years of data are needed in most cases to reach a stable estimate of annual rainfall erosivity. Moreover, the temporal resolution of measuring devices was found to have a significant effect on R, with coarser temporal resolution leading to a higher relative bias.

Mapping storm spatial profiles for flood impact assessments

Synthetic design storms are often used to plan new drainage systems or assess flood impacts on infrastructure. To simulate extreme rainfall events under climate change, design storms can be modified to match a different return frequency of extreme rainfall events as well as a modified temporal distribution of rainfall intensities. However, the same magnitude of change to the rainfall intensities is often applied in space. Several hydrological applications are limited by this. Climate change impacts on urban pluvial floods, for example, require the use of 2D design storms (rainfall fields) at sub-kilometer and sub-hourly scales. Recent kilometer scale climate models, also known as convection-permitting climate models (CPM), provide rainfall outputs at a high spatial resolution, although rainfall simulations are still restricted to a limited number of climate scenarios and time periods. We nevertheless explored the potential use of rainfall data obtained from these models for hydrological flood impact studies by introducing a method of spatial quantile mapping (SQM). To demonstrate the new methodology, we extracted high-resolution rainfall simulations from a CPM for four domains representing different urban areas in Switzerland. Extreme storms that are plausible under the present climate conditions were simulated with a 2D stochastic rainfall model. Based on the CPM-informed stochastically generated rainfall fields, we modified the design storms to fit the future climate scenario using three different methods: the SQM, a uniform quantile mapping, and a uniform adjustment based on a rainfall–temperature relationship. Throughout all storms, the temporal distribution of rainfall was the same. Using a flood model, we assessed the impact of different rainfall adjustment methods on urban flooding. Significant differences were found in the flood water depths and areas between the three methods. In general, the SQM method results in a higher flood impact than the storms that were modified otherwise. The results suggest that spatial storm profiles may need to be re-adjusted when assessing flood impacts.

Stochastic daily rainfall generation on tropical islands with complex topography

Abstract. Stochastic rainfall generators are probabilistic models of rainfall space–time behavior. During parameterization and calibration, they allow the identification and quantification of the main modes of rainfall variability. Hence, stochastic rainfall models can be regarded as probabilistic conceptual models of rainfall dynamics. As with most conceptual models in earth sciences, the performance of stochastic rainfall models strongly relies on their adequacy in representing the rain process at hand. On tropical islands with high elevation topography, orographic rain enhancement challenges most existing stochastic models because it creates localized precipitations with strong spatial gradients, which break down the stationarity of rain statistics. To allow for stochastic rainfall modeling on tropical islands, despite non-stationarity of rain statistics, we propose a new stochastic daily multi-site rainfall generator specifically for areas with significant orographic effects. Our model relies on a preliminary classification of daily rain patterns into rain types based on rainfall space and intensity statistics, and sheds new light on rainfall variability at the island scale. Within each rain type, the distribution of rainfall through the island is modeled by combining a non-parametric resampling of past analogs of a latent field describing the spatial distribution of rainfall, and a parametric gamma transform function describing rain intensity. When applied to the stochastic simulation of rainfall on the islands of O`ahu (Hawai`i, United States of America) and Tahiti (French Polynesia) in the tropical Pacific, the proposed model demonstrates good skills in jointly simulating site-specific and island-scale rain statistics. Hence, it provides a new tool for stochastic impact studies in tropical islands, in particular for watershed water resource management.

Development and Application of a Rainfall Temporal Disaggregation Method to Project Design Rainfalls

A climate model is essential for hydrological designs considering climate change, but there are still limitations in employing raw temporal and spatial resolutions for small urban areas. To solve the temporal scale gap, a temporal disaggregation method of rainfall data was developed based on the Neyman–Scott Rectangular Pulse Model, a stochastic rainfall model, and future design rainfall was projected. The developed method showed better performance than the benchmark models. It produced promising results in estimating the rainfall quantiles for recurrence intervals of less than 20 years. Overall, the analysis results imply that extreme rainfall events may increase. Structural/nonstructural measures are urgently needed for irrigation and the embankment of new water resources.

The Advanced Meteorology Explorer: a novel stochastic, gridded daily rainfall generator

Synthetic rainfall simulations from stochastic models are commonly used for water resource management, as they are able to provide a wider range of meteorological conditions than those seen in the observed record. Here, we present a novel stochastic rainfall modelling framework, the Advanced Meteorology Explorer (AME), which combines and extends existing methods to enhance model flexibility, and meet a number of key water industry needs. This framework allows for the simulation of physically consistent synthetic daily rainfall data, coherently in space and time, on a high-resolution grid over a region of interest. The AME uses an advanced hidden Markov model structure within a Bayesian hierarchical framework to represent daily rainfall at a set of locations in a region, conditional on important climate drivers. The climate drivers included in the rainfall model at each location are able to vary using penalised regression, ensuring a transferable model that can be applied to different locations without adaptation. The dependence between locations is modelled following a flexible copula approach, able to capture varying dependence structures within the data, allowing for spatially coherent simulations at the modelled locations. Simulations are then interpolated to a high-resolution grid using a terrain adjusted inverse-distance weighted interpolation method.The AME framework is applied to 105 years (1914–2018) of daily rainfall data at 39 sites in the Greater Anglian region of the UK, and used to generate 1000 alternative realisations of the same period on a 5 km grid over the region. Validation of these simulations shows how the AME framework is able to accurately capture rainfall occurrence and intensity, as well as long-duration meteorological drought behaviour, important for quantifying water resource risk in this dry region. This framework has the potential to be applied to other regions, incorporate additional weather variables and the effect of climate change, and the resulting simulations can be used for environmental risk assessment in any industry impacted by rainfall.

Can we estimate flood frequency with point-process spatial-temporal rainfall models?

Stochastic rainfall models are commonly used in practice for long-term flood risk management. One of the most widely used model types is based on point processes. Despite the widespread use of such models, whether their known simplifications in describing the space–time structure of rainfall will affect the accuracy of flood estimation has not been quantified. In this study, we quantify the biases introduced by the rainfall model limitations to flood estimates in two medium-sized river catchments (717 km2 and 844 km2) in the South East of the UK. To achieve this, we used nine years of hourly radar rainfall data, a dense network of hourly rain gauges, a spatial–temporal rainfall stochastic model based on point processes, and a fully distributed hydrological model. We modelled the corresponding catchment water dynamics using observed and simulated hourly rainfall and then assessed whether the errors introduced by the stochastic model will propagate in the river flow dynamics. Our results show that the stochastic rainfall model properly captures the point-scale rainfall statistics, including point extremes and the cross-site spatial correlations. However, the model results in a bias on extremes of areal statistics, including an overestimation of the areal reduction factor, extreme areal mean precipitation, and the areal fraction of rain (wet area ratio). Using this as input for continuous hydrological simulations, we find that the flow duration curves are well preserved, particularly in the high flow seasons (relative bias is less than 7%). The model also reproduces well the flood frequency curves at a daily scale with an averaged relative bias of 0.36–16.9% at 10-year return levels, confirming its ability to infer the long-term flood risk for medium-sized catchments. However, the summer-season hourly peak discharge is highly overestimated with a relative bias of over 163.5% at the same return level. The overestimation in summer hourly peak discharge is explained by the dominating convective systems in summer and the misrepresented spatial structure in the model.

A simple scheme to adjust Poisson cluster rectangular pulse rainfall models for improved performance at sub-hourly timescales

Although sub-hourly rainfall temporal characteristics play an important role in occurrences and magnitudes of urban flash floods, most stochastic rainfall models struggle to reproduce them when rainfall information is not available at sub-hourly timescales. This study suggests a simple approach to modifying Poisson cluster rectangular pulse rainfall models to resolve this issue. In this model, the original rectangles representing a rain cell are replaced by a bell shape that is mathematically represented as y = sin2(x). Such an alteration no longer makes it possible to obtain closed form expressions for the various rainfall statistics for model parameter calibration. Based on an assumption that this alteration would hardly alter the hourly and supra-hourly statistics, we calibrated the parameters of a rectangular pulse model with these statistics (from 1 to 32 h) and transplanted the same parameters to the sine-squared pulse model to generate 500 years of 5-minute rainfall. Importantly, this parameter transplant procedure does not require any sub-hourly information so that it amounts to a downscaling algorithm for the generation of sub-hourly rainfall. Our results show that both models reproduced well the rainfall mean, variance, autocovariance, proportion of dry periods, and extreme values at hourly to 72-hour timescales. However, the sine-squared pulse model performed significantly better at sub-hourly time scales. The same 500 years of synthetic rainfall data and 69 years of observed rainfall data were then used as input of a dual-drainage urban watershed model developed for a frequently flooded urban area of Seoul. While the traditional rectangular pulse model significantly underestimated the flooded area, the new sine-squared pulse model successfully reproduced the flooded area caused by the observed rainfall.

Radar and Rain Gauge Data Fusion Based on Disaggregation of Radar Imagery

AbstractRemotely sensed radar data and in situ rain gauge observations provide complementary information about rainfall. Specifically, radar data inform rainfall location and extent, while rain gauges provide accurate observations of the local rain intensity and its temporal variability. Drawing on the respective strengths of these two data sources, radar and rain gauge data fusion is becoming more and more common to derive both accurate and comprehensive rain estimates. However, combined precipitation estimates are often restricted to a resolution of 1 km in space and 1 h in time. In this paper, I propose to use radar data disaggregation to merge radar and rain gauge observations at a high space‐time resolution. To this end, a stochastic rainfall model is first trained on rain gauge observations and then combined with area‐to‐point geostatistical simulations to simultaneously: (i) disaggregate radar data, (ii) honor rain gauge observations, and (iii) reproduce local rain statistics. The proposed approach is applied to summer rain data collected in a small Alpine catchment, and data fusion provides rainfall reanalysis results at a resolution of 200 m in space and 2 min in time, including remote locations where in‐situ observation was not feasible.

A robust method to develop future rainfall IDF curves under climate change condition in two major basins of Iran

Current rainfall Intensity-Duration-Frequency (IDFs) curves generally were developed based on stationary extreme value assumption. However, climate change is expected to alter extreme rainfalls and invalidate the stationary assumption. So, it is crucial to develop future rainfall IDFs taking into account the impacts of climate change. Difficulty in preparing reliable rainfall time series with fine time step and capturing extremes for future climate scenarios has led to limited studies aimed at investigating the change of IDF curves due to climate change. In this paper, a robust stochastic rainfall model (NSRP) is employed to produce future rainfall IDFs for five stations across Karkheh and Karun basins in Iran. NSRP can generate long-term realistic rainfall series containing extremes. Moreover, it applies changes in different rainfall statistical characteristics, projected by GCMs, in the downscaling procedure. For each station, the model was calibrated using observed rainfalls series. Then, 3000-year daily rainfall series were generated, and historical IDFs were developed. Consequently, NSRP parameters were perturbed based on GCM projections. Then, 3000-year future rainfall series were generated, and future IDFs were developed. The climate change uncertainties were represented by employing two GCM (CanESM2 and HadGem2) and three emission scenarios (RCP2.6, RCP4.5, and RCP8.5). The results showed NSRP is able to reproduce observed extreme rainfalls in a wide range of time scales. Also, climate change will lead to a considerable increase in future extreme rainfall intensity in the study basins. As averages of all considered scenarios, rainfall intensities will increase between 22 and 206% in the future.

Development and hydrometeorological evaluation of a new stochastic daily rainfall model: Coupling Markov chain with rainfall event model

Stochastic rainfall models have been widely used for hydrological modelling and climate change impact studies, the accuracy of which can substantially affect the reliability of water resources planning, hydraulic structure design and flood and drought risk assessment. The primary objective of this study is to develop a stochastic daily rainfall model through coupling a Markov chain model with a rainfall event model (SDRM-MCREM) to simultaneously preserve the statistical properties of rainfall time series and rainfall events. The newly developed model is applied to the Qu River basin, East China and its performance is evaluated at catchment scale. Results demonstrate that SDRM-MCREM shows a good performance in reproducing most of the rainfall time-series statistics (i.e. rainfall percentiles, average monthly and annual rainfall, inter-monthly rainfall variability and extreme rainfall) and rainfall event characteristics (i.e. distributions of wet and dry spells, occurrence frequency of different rainfall event classes, temporal rainfall patterns and their occurrence frequency in different rainfall event classes). The statistics of average runoff and extreme runoff are also well preserved by using the SDRM-MCREM simulations as input of hydrological modelling except that the interannual variability of rainfall and runoff is slightly underestimated. Moreover, SDRM-MCREM shows a great potential to be used for flood and drought risk assessment in reproducing the exceedance probabilities of high flows (e.g. annual maximum 1-day, 3-day and 5-day mean runoff) and low flows (e.g. annual minimum 7-day, 30-day and 90-day mean runoff).

A stochastic rainfall model that can reproduce important rainfall properties across the timescales from several minutes to a decade

A stochastic rainfall model that can reproduce various rainfall characteristics at timescales between 5 min and one decade is introduced. The model generates the fine-scale rainfall time series using a randomized Bartlett-Lewis rectangular pulse model. Then the rainstorms are shuffled such that the correlation structure between the consecutive storms are preserved. Finally, the time series is rearranged again at the monthly timescale based on the result of the separate coarse-scale monthly rainfall model. The method was tested using the 69 years of 5-minute rainfall data recorded at Bochum, Germany. The mean, variance, covariance, skewness, and rainfall intermittency were well reproduced at the timescales from 5 min to a decade without any systematic bias. The extreme values were also well reproduced at timescales from 5 min to 3 days. The past-7-day rainfall before an extreme rainfall event, which is highly associated with the extreme flow discharge was reproduced well too. The rainstorm shuffling approaches introduced here may be adopted as a standard procedure in combination with any Poisson cluster rainfall model. The methods are simple and parsimonious, yet significantly reduce the systematic underestimation of rainfall variance at coarse scales, and improve the reproduction of skewness, and extreme rainfall depths values at a range of time-scales, thereby addressing well-known shortcomings of Poisson cluster rainfall models.

ASSESSMENT OF THE USSURI RIVER FLOOD REGIME CHANGES REGARDING TO CLIMATE PROJECTIONS TO THE END OF THE XXI CENTURY

Detecting reaction of river systems to climate changes is one of the key problems in modern hydrology. Moreover, this is the case of the Far East of Russia, where moisture regime variability is well known. The presented article deals with the assessment results of changes in the flood characteristics of the Ussuri rivers basins under the expected climate changes, linked with global warming. The estimates are based on dynamic–stochastic modeling scheme, in which numerous synthetic precipitation series are input to the hydrological model for searching of all kinds of humidification conditions at the investigated catchments. There was used a regional hydrological Flood Cycle Model of small river basin that focuses on the description of runoff formation processes in the summer–autumn flood period. Climate changes input is based on the integration of the regional stochastic rainfall model and General Circulation Models (GCM) data to the end of XXI century (2071–2100). It is shown that an increase in precipitation sum leads to an unproportional response of the evaluated flood flow characteristics. The relative increase in the mean values of the maximum flow reaches 4–5 times and 2–3 times for the seasonal runoff depth depending on applied GCM and climate scenarios. The obtained results are consistent with the analysis of observational data using the “elasticity coefficient” and previous investigations. Therefore, flood risks could be expected at the investigated territory due to climate changes.