Droughts caused by meteorological factors such as a long-term lack of precipitation can propagate into several types of drought through the hydrological cycle. Among them, a phenomenon in which drought has a significant impact on the ecosystem can be defined as an ecological drought. In this study, the Ecological Drought Condition Index-Vegetation (EDCI-veg) was newly proposed to quantitatively evaluate and monitor the effects of meteorological drought on vegetation. A copula-based bivariate joint probability distribution between vegetation information and drought information was constructed, and EDCI-veg was derived from the joint probability model. Through the proposed EDCI-veg, it was possible to quantitatively estimate how much the vegetation condition was affected by the drought, and to identify the timing of the occurrence of the vegetation drought and the severity of the vegetation drought. In addition, as a result of examining the applicability of the proposed EDCI-veg by comparing past meteorological drought events with the corresponding vegetation conditions, it was found that EDCI-veg can reasonably monitor vegetation drought. It has been shown that the newly proposed EDCI-veg in this study can provide useful information on the ecological drought condition that changes with time. On the other hand, the ecological drought analysis based on the type of land cover showed that the response of vegetation to meteorological drought was different depending on the land cover. In particular, it was revealed that the vegetation inhabiting the forest has a relatively high resistance to meteorological drought.

Drought is a phenomenon that is caused by several factors and can be divided into meteorological drought, agricultural drought, hydrological drought, and socioeconomic drought. In this study, the characteristics of propagating from meteorological drought to agricultural (or hydrological) drought in the Andong Dam basin and Hapcheon Dam basin located in the Nakdong River basin in Korea were investigated. Standardized precipitation index (SPI), standardized soil moisture index (SMI), and standardized runoff index (SRI) were used to characterize meteorological, agricultural, and hydrological droughts, respectively. SPI-m (1–12) on various timescales and SMI-1 (or SRI-1) were selected as drought propagation timeseries, such that a correlation analysis was performed to evaluate the correlation and propagation time between meteorological and agricultural (or hydrological) drought. Propagation probability was quantified using a copula-based model. The correlation between meteorological and agricultural (or hydrological) droughts was not significantly affected by seasons. A relatively strong correlation was found in summer. A relatively weak correlation was shown in autumn. In addition, it was found that there was a difference in correlation between the Andong Dam basin and the Hapcheon Dam basin. On the other hand, in both watersheds, the propagation time was as long as 2 to 4 months in spring and decreased to 1 month in summer.

Since vegetation is closely related to a variety of hydrological factors, the vegetation condition during a drought is greatly affected by moisture supply or moisture demand from the atmosphere. However, since feedback between vegetation and climate in the event of drought is very complex, it is necessary to construct a joint probability distribution that can describe and investigate the interrelationships between them. In other words, it is required to understand the interaction between vegetation and climate in terms of joint probability. In this study, the possibility of drought stress experienced by vegetation under various conditions occurring during drought was investigated by dividing drought into two aspects (atmospheric moisture supply and moisture demand). Meteorological drought indices that explain different aspects of drought and vegetation-related drought indexes that describe the state of vegetation were estimated using data remotely sensed by satellites in parts of Far East Asia centered on South Korea. Bivariate joint probability distribution modeling was performed from vegetation drought index and meteorological drought index using Copula. It was found that the relationship between the vegetation drought index and the meteorological drought index has regional characteristics and there is also a seasonal change. From the copula-based model, it was possible to quantify the conditional probability distribution for the drought stress of vegetation under meteorological drought scenarios that occur from different causes. Through this, by mapping the vulnerability of vegetation to meteorological drought in the study area, it was possible to spatially check how the vegetation responds differently depending on the season and meteorological causes. The probabilistic mapping of vegetation vulnerability to various aspects of meteorological drought may provide useful information for establishing mitigation strategies for ecological drought.

Several methods have been proposed to analyze the frequency of nonstationary anomalies. The applicability of the nonstationary frequency analysis has been mainly evaluated based on the agreement between the time series data and the applied probability distribution. However, since the uncertainty in the parameter estimate of the probability distribution is the main source of uncertainty in frequency analysis, the uncertainty in the correspondence between samples and probability distribution is inevitably large. In this study, an extreme rainfall frequency analysis is performed that fits the peak over threshold series to the covariate-based nonstationary generalized Pareto distribution. By quantitatively evaluating the uncertainty of daily rainfall quantile estimates at 13 sites of the Korea Meteorological Administration using the Bayesian approach, we tried to evaluate the applicability of the nonstationary frequency analysis with a focus on uncertainty. The results indicated that the inclusion of dew point temperature (DPT) or surface air temperature (SAT) generally improved the goodness of fit of the model for the observed samples. The uncertainty of the estimated rainfall quantiles was evaluated by the confidence interval of the ensemble generated by the Markov chain Monte Carlo. The results showed that the width of the confidence interval of quantiles could be greatly amplified due to extreme values of the covariate. In order to compensate for the weakness of the nonstationary model exposed by the uncertainty, a method of specifying a reference value of a covariate corresponding to a nonexceedance probability has been proposed. The results of the study revealed that the reference covariate plays an important role in the reliability of the nonstationary model. In addition, when the reference covariate was given, it was confirmed that the uncertainty reduction in quantile estimates for the increase in the sample size was more pronounced in the nonstationary model. Finally, it was discussed how information on a global temperature rise could be integrated with a DPT or SAT-based nonstationary frequency analysis. Thus, a method to quantify the uncertainty of the rate of change in future quantiles due to global warming, using rainfall quantile ensembles obtained in the uncertainty analysis process, has been formulated.

Many studies have applied the Long Short-Term Memory (LSTM), one of the Recurrent Neural Networks (RNNs), to rainfall-runoff modeling. These data-driven modeling approaches learn the patterns observed from input and output data. It is widely known that the LSTM networks are sensitive to the length and quality of observations used for learning. However, the discussion on a better composition of input data for rainfall-runoff modeling has not yet been sufficiently conducted. This study focuses on whether the composition of input data could help improve the performance of LSTM networks. Therefore, we first examined changes in streamflow prediction performance by various compositions of meteorological variables which are used for LSTM learning. Second, we evaluated whether learning by integrating data from all available basins can improve the streamflow prediction performance of a specific basin. As a result, using all available meteorological data strengthened the model performance. The LSTM generalized by the multi-basin integrated learning showed similar performance to the LSTMs separately learned for each basin but had more minor errors in predicting low flow. Furthermore, we confirmed that it is necessary to group by selecting basins with similar characteristics to increase the usefulness of the integrally learned LSTM.