Characteristics of Precipitation and Floods during Typhoons in Guangdong Province

Abstract

The spatial and temporal characteristics of precipitation and floods during typhoons in Guangdong province were examined by using TRMM TMPA 3B42 precipitation data and the Dominant River Routing Integrated with VIC Environment (DRIVE) model outputs for the period 1998–2019. The evaluations based on gauge-measured and model-simulated streamflow show the reliability of the DRIVE model. The typhoon tracks are divided into five categories for those that landed on or influenced Guangdong province. Generally, the spatial distribution of precipitation and floods differ for different typhoon tracks. Precipitation has a similar spatial distribution to flood duration (FD) but is substantially different from flood intensity (FI). The average precipitation over Guangdong province usually reaches its peak at the landing time of typhoons, while the average FD and FI reach their peaks several hours later than precipitation peak. The lagged correlations between precipitation and FD/FI are hence always higher than their simultaneous correlations.

1. Introduction

Typhoons are one of the most devastating natural disasters, causing extreme precipitation, high winds, storm surges, mudslides, and floods [1,2,3]. Several typhoons strike China each year [4,5,6]. A lot of rainstorms are caused by typhoons in China [4,7]. Furthermore, rainstorms and extreme precipitation can cause floods, which lead to property damage and casualties. However, there is a lack of understanding regarding how typhoons result in floods. Therefore, it is a pressing task to improve our understanding of typhoon floods, especially in typhoon-prone zones. Guangdong province is one of the most developed provinces in China. It is bordered by the South Sea to the south and has elevated terrains to its north. Typhoon floods appear more frequently in Guangdong than in other provinces, and the number of typhoon disasters is increasing yearly [8,9]. Therefore, it is of great importance to analyze the characteristics and causes of typhoon-related rainstorms and study these events-induced secondary flood disasters in Guangdong province.

Hydrologic models are considered to be efficient tools for flood monitoring and flood forecasting [10,11,12,13]. The development of a flood warning system based on a hydrologic model has the advantage of providing helpful information for flood prediction and evaluation [14]. Wu et al. developed the Dominant River Routing Integrated with VIC Environment (DRIVE) model by coupling the DRTR model with the VIC land surface model [13,14,15,16,17,18]. The DRIVE model has the advantages of high spatiotemporal resolution and an adaptive space–time scale, as it includes flood inundation modules and is suitable for flood simulation at different spatial and temporal scales. The DRIVE model has been validated by using its outputs and observed streamflow at 1121 gauges, 2086 remote sensing-based global flood inventory, and other real-time event evaluations [13].

Accurate precipitation data with fine resolution are critical for hydrologic models. Rain gauges can provide accurate measurements but are sparse, especially in remote areas [19]. Weather radar records are always affected by atmospheric conditions and high terrain in mountainous areas [20,21,22,23,24]. The development of remote sensing satellite precipitation data provides convenience for flood simulation and forecasting [13,25,26]. The Tropical Rainfall Measuring Mission (TRMM), Multi-satellite Precipitation Analysis (TMPA) product [27], have been applied in many previous studies successfully [25,26]. Driving a validated hydrological model with high resolution TMPA precipitation product, this study provides a tool for predicting runoff response and thus potentially mitigating future typhoon flood destruction. Thus, the objectives of this study are: (1) Quantifying the spatial and temporal characteristics of rainfall and floods during the typhoon period in Guangdong province; (2) exploring the simultaneous and lagged correlation relations between precipitation and floods during the typhoon period in Guangdong province.

Data and Methods are provided in Section 2. In Section 3, we evaluate the DRIVE model performance in the Guangdong area. Section 4 describes the spatial and temporal characteristics of typhoon precipitation and typhoon floods, and also the relationships between precipitation and floods. A discussion is presented in Section 5, and the findings are summarized in Section 6.

2. Data and Methods

This study used the DRIVE model which is coupled by the DRT-based runoff routing (DRTR) model [13,14,15,16,17,18] with the Variable Infiltration Capacity (VIC) land surface model [28,29]. The DRT-based runoff routing (DRTR) model is based on the hierarchical DRT method [14,15,16,17]. It includes a package of hydrographic upscaling algorithms and resulting global datasets (flow direction, flow distance, drainage area, river network, slope, etc.) for large-scale hydrologic modeling [13].

The VIC model includes snow and soil frost dynamics module [30,31], it has advantages in snowmelt-dominated basins, such as mountainous regions [13,32,33,34]. The VIC model considers the subgrid heterogeneity of infiltration capacity through statistical variable infiltration curves [35], and also considers subgrid parameterization and processes on fractional subgrid areas for different land cover types and elevation bands. The VIC model has been modified from its original individual grid cell-based mode to a mode that is able to simulate spatially distributed runoff at each time step without changing its model physics. More details of the DRIVE model can be found in the literature [13,36,37].

DRIVE retrospective simulations were performed for the period of 1998–2019. The meteorological inputs of the DRIVE model include precipitation, temperature (min and max), wind speed. The three-hourly TRMM/TMPA 3B42 data [27] are used as precipitation input in this paper, and the three-hourly wind speed and temperature applied here are from The Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) [38]. All input data can be obtained from https://disc.gsfc.nasa.gov/, accessed on 1 September 2020. The outputs of the DRIVE model include discharges and routed runoff with time resolution of three-hourly and spatial resolution of 0.125° × 0.125°, which can be obtained from the NASA’s real-time Global Flood Monitoring System (GFMS) via http://flood.umd.edu, accessed on 1 September 2020. A grid cell is classified as flooded when the routed runoff is greater than the flood threshold. Detailed information of defining and determining flood threshold is provided in prior reports [39,40]. Differences (∆) between routed runoff and a threshold are calculated at grids. If ∆ > 0, the sum of the total hours is calculated as the flood duration (FD); then, the sum of the total values of ∆ when ∆ > 0 is flood intensity (FI) at each grid.

The typhoon best track data from 1998 to 2019 are obtained from Tropical Cyclone Data Center, China Meteorological Administration (http://tcdata.typhoon.org.cn, accessed on 1 September 2020) [41,42].

Topographic conditions in Guangdong province are complex (Figure 1). There are many small mountains in the north, east, and west of Guangdong; the Pearl River Delta and the coastal areas of Guangdong are low-lying areas. According to the terrain condition, we classified the typhoon landing paths into three categories: the first class consists of typhoons landing in the Pearl River Delta, the second class landing in the east of Guangdong, and the third class landing in the west of Guangdong. Besides the landfall typhoons, there are also typhoons that did not land in Guangdong but influenced Guangdong. We divided these typhoons into two categories: the first class consists of those moving northward and the second class moving westward. Thus, five categories of typhoons are discussed in this study. We only consider the named typhoons that landed on or influenced Guangdong during 1998–2019.

3. Model Validation

To validate the model performance, the DRIVE outputs were first compared with the available hydrological gauge observations from eight local stations for the years 2017 to 2019 (https://zjhy.mot.gov.cn/index.html, accessed on 1 September 2020). These eight stations are Boluo, Chaoan, Feilaixia, Gaoyao, Makou, Sanshui, Shijiao, and Tianhe (Figure 1). The hydrological gauges are mainly located in the river channels with an exception of Chaoan. There are mountains in the northern, eastern, and western parts of Guangdong, and the rivers are mainly located in the valley and low-lying areas (Figure 1).

Indicators including the correlation coefficient (R), the Nash–Sutcliffe efficiency (NSE) [43], and Percent Bias (PBIAS) [44], were used to evaluate the agreement between the simulated outputs and observations. It is worth noting that positive PBIAS values indicate model underestimation bias, while negative values indicate model overestimation bias.

Simulated and observed daily streamflow, together with the metrics for model performance, can be seen in Figure 2. Seven out of eight stations show high significant correlations with simulated data, with a correlation coefficient larger than 0.6. Six stations show positive NSE scores with the highest value of 0.53 at Shijiao station, while the other two stations have negative NSE. There are also six stations with small errors in PBIAS between −15% and 15% [44]. For all stations except for Sanshui and Tianhe, the model outputs can generally simulate the variation and magnitude of observed streamflow. However, for the Sanshui and Tianhe stations, there are magnitude biases with observed data; the model outputs underestimate the streamflow, but can generally capture the variations shown in observed streamflow. The possible reasons for the differences may be the influence of the dam which is not considered in the model or the errors in the observed streamflow or the systematic error in the model.

Overall, the DRIVE outputs can provide reasonable estimations for most gauges in Guangdong province; therefore, using outputs of the DRIVE model to examine the characteristics of typhoon floods in Guangdong province is reasonable.

4. Results

4.1. Spatial and Temporal Distribution of Precipitation and Floods during Typhoon Period

Prior to exploring the spatial and temporal distribution of precipitation and floods during the typhoon period, we first examined the tracks of 145 the typhoons that made landfall in or influenced Guangdong during 1998–2019. As shown in Figure 3, 21 typhoons landed on the Pearl River Delta region during 1998–2019, 14 typhoons landed on the eastern regions of Guangdong, and 29 typhoons landed on the west regions of Guangdong. As for the non-landing typhoons, 55 typhoons moved northward and 26 typhoons moved westward.

Figure 4 illustrates the spatial distributions of the average cumulative precipitation during 24 h before through 24 h after the typhoons made landfall in different regions. For the typhoons that landed on the Pearl River Delta region, the corresponding high values of precipitation appeared in the coastal areas of the Pearl River Delta region and the western regions of Guangdong province, and the areas with less precipitation were mainly located in the northern and northeastern regions of Guangdong. For the typhoons that landed on eastern Guangdong, the corresponding high values of precipitation appeared in the Pearl River Delta region and the eastern regions of Guangdong province. For the typhoons that landed on western Guangdong, the corresponding high values of precipitation appeared in the western regions of Guangdong province, and the areas with less precipitation were mainly located in the northern parts of east Guangdong. For the non-landing typhoons that moved northward, major precipitation zones appeared in a small part of eastern Guangdong, with the values of precipitation being apparently lower than corresponding high precipitation values that were caused by the landfall typhoons. For the typhoons that moved westward, the related high precipitation appeared in Jiangmen city and small parts of Zhanjiang city. Therefore, when typhoons make landfall in different areas of Guangdong, the corresponding areas of high precipitation differ as expected, and thus we will focus attention on the contrasting high precipitation areas that correspond to different tracks.

Spatial distributions of the average cumulative flood duration are depicted in Figure 5. For the typhoons that landed on the Pearl River Delta region, high flood durations show similar distributions to high precipitation, especially in the western areas of Guangdong; this indicates that flood duration during typhoon periods may have a high correlation with precipitation. Lower values of flood duration appear in the northern and northeastern regions of Guangdong. When typhoons made landfall in eastern Guangdong, high flood durations were mainly located in eastern regions of Guangdong. When the typhoons made landfall in western Guangdong, high flood durations were mainly located in western regions of Guangdong. It should be noted that there were also higher values in the eastern coastal areas, and this distribution can also be seen for the precipitation. For the non-landing typhoons that moved northward, the related high flood durations appeared in the eastern parts of Guangdong. For the non-landing typhoons that moved westward, the related high flood durations appeared in Zhanjiang and Jiangmen city, which also correlated with precipitation.

Figure 6 depicts the spatial distributions of the average cumulative flood intensity; the blue lines indicate the river channels in Guangdong. When typhoons made landfall in the Pearl River Delta region, high flood intensity appeared in western Guangdong. For typhoons that landed on eastern Guangdong, the related high flood intensity appeared in the river areas of northern and eastern Guangdong, and in the low-lying coastal areas. When typhoons made landfall in western Guangdong, high flood intensity appeared in the areas with low terrain in western Guangdong and some small low terrain parts of northern and eastern Guangdong. For the non-landing typhoons that moved northward, the related high flood intensity appeared in the river areas of northern and eastern Guangdong. For the non-landing typhoons that moved westward, the related high flood intensity appeared in some small low terrain regions. In contrast to flood duration, flood intensity does not show a similar spatial distribution as precipitation, and appears to be more related to landform.

Time series plots can depict more clearly the variations in precipitation and floods during the typhoon period (Figure 7). Here the data were standardized for comparison. For the typhoons that landed on the Pearl River Delta region, the average precipitation of all areas of Guangdong showed a rising tendency during the 0 to 24 h period; the precipitation reached its peak at 24 h, and then decreased with time. The 24 h corresponds to the timeframe in which the typhoons make landfall; therefore, it is of great importance to focus on the high precipitation areas in Guangdong during the time that typhoons make landfall. Flood duration reached its peak at 33 h, about 9 h after the precipitation peak; this means that high precipitation will have a lagged impact on floods. Flood intensity reached its peak at 51 h, about 27 h later than the precipitation peak. For the typhoons that landed on eastern Guangdong, the precipitation reached its peak at 30 h, and flood duration reached its peak at 39 h, also 9 h after the precipitation peak. Flood intensity reached its peak at 63 h, about 33 h later than precipitation peak. For the typhoons that landed on western Guangdong, the precipitation reached its peak at 24 h, and flood duration reached its peak at 33 h, 9 h after the precipitation peak. Flood intensity reached its peak at 90 h, about 66 h later than precipitation peak. For the non-landing typhoons that moved northward, the related peak precipitation appeared at 33 h, and then there was a lower peak at 45 h. Flood duration reached its peak at 57 h, 24 h later than the precipitation peak. Flood intensity reached its peak at 72 h, about 39 h later than the precipitation peak. For the non-landing typhoons that moved westward, the related peak precipitation appeared at 15 h, and flood duration reached its peak at 33 h; flood intensity reached its peak at 39 h, 24 h later than the peak of precipitation. Time series of precipitation, flood duration, and flood intensity demonstrate that it takes time for floods to converge. The lag time of a flood peak compared to the peak precipitation differs with contrasting typhoon tracks. To a certain extent, these lag times can inform flood prediction during the typhoon period.

4.2. Correlations of Precipitation and Floods during the Typhoon Period

The time series plots confirm that the flood peaks always lag the precipitation peaks (Figure 7). The simultaneous and lagged correlations between averaged precipitation and floods are then further estimated (Figure 8). The values above the dotted line indicate the correlation coefficient is above the 90% confidence level. As shown in Figure 8, for the typhoons that made landfall on the Pearl River Delta region, the correlation between precipitation and flood duration is significant and reaches its peak at 9 h of lag time between precipitation and flood duration, and the significant correlation between precipitation and flood intensity reaches its peak at 24 h of the lag time between precipitation and flood intensity. For the typhoons that landed on eastern Guangdong, the correlation between precipitation and flood duration is significant and reaches its peak at 9 h time-lag, and correlation between precipitation and flood intensity reaches its peak at a 27 h time-lag. For the typhoons that landed on western Guangdong, the correlation between precipitation and flood duration is significant and reaches its peak at a 12 h time-lag, and the correlation between precipitation and flood intensity reaches its peak at a 15 h time-lag, but the correlation is not significant. For the non-landing typhoons that moved northward, the correlation between precipitation and flood duration is significant and reaches its peak at a 9 h time-lag. The precipitation and flood intensity are significantly correlated for this category, and the correlation reaches its peak at 27 h time-lag between precipitation and flood intensity. For the non-landing typhoons that moved westward, the correlation between precipitation and flood duration is significant and reaches its peak when precipitation leads by 9 h, and the correlation between precipitation and flood intensity is significant and reaches its peak at 24 h time-lag. These lag times provide evidence for the prediction of floods during typhoon periods in Guangdong.

Figure 9 shows the correspondence between averaged cumulative precipitation and flood duration during the 24 h before though 24 h after the typhoons made landfall. Scatterplots can be used to show the spatial similarity of precipitation and flood duration. Significant correlations between precipitation and flood duration can be seen for the five classes of typhoon tracks, especially for the typhoons that moved northward, which reveal a correlation coefficient of 0.82. This result indicates that precipitation and flood duration tend to have similar spatial distributions during the typhoon period.

Figure 10 shows a low linear correlation between precipitation and flood intensity; none of the correlation coefficients are significant for the five classes of typhoon tracks. This indicates that the spatial distribution of flood intensity differs from precipitation, which is shown in Figure 4 and Figure 6. Flood intensity tends to be affected by other variables, such as catchment size, landscape, topography, or other hydrological conditions.

5. Discussion

Typhoon rainstorms and the following secondary flood and landslide disasters can seriously affect coastal areas [45]. A number of studies have focused on this topic [45,46,47]. Gutierrez-Lopez proposed the Huff curves to get typical hurricane precipitation hyetograms in Mexico [45]. Trošelj et al. have simulated the typhoon-induced extreme river discharges in Japan [46]. The results from this paper further provide a foundation for constructing maps of typhoon and heavy rainfall-induced floods and translational landslides [47].

In this paper, typhoons during the period 1998–2019 and related to Guangdong province were divided into five categories (Figure 3). For each class, the spatial distributions and temporal variations of flood intensity and duration were investigated by using the outputs from DRIVE model driven by the TMPA 3B42 rainfall product. Additionally, connections between precipitation and floods were examined.

The corresponding high precipitation, flood duration, and flood intensity differ for different typhoon tracks (Figure 4, Figure 5, Figure 6). The longer flood duration showed similar spatial distributions as the larger precipitation (Figure 4 and Figure 5). The higher flood intensity shows different distribution as the larger precipitation (Figure 4 and Figure 6), which is similar with past studies [39]. This could be because of the flood intensity tends to be affected by other variables, such as catchment size or landscape or other hydrological conditions [48,49].

For different categories of typhoons tracks, time series analysis shows the average precipitation over Guangdong province usually reached its peak at landing time, and the flood duration/flood intensity reached its peak later than the precipitation peak (Figure 7). As shown in previous papers, precipitation has a lagged impact on floods [40,50], because it takes time for precipitation to concentrate and form discharges. The concentration time depends on the spatial distribution of precipitation, the drainage area, topography, and vegetation cover [49].

6. Conclusions

The different tracks of typhoons that landed on or influence Guangdong result in different spatial distributions of precipitation, flood duration, and flood intensity. The flood duration generally has similar distribution with precipitation. Nevertheless, differences exist regarding spatial distributions between flood intensity and precipitation. The time series results also show that the average precipitation over Guangdong province usually reaches its peak at the landing time of typhoons, and the average flood duration and flood intensity reach their peaks several hours later than the peak of precipitation. Additionally, the lagged correlations between precipitation and flood duration/flood intensity are generally higher than the simultaneous correlations. These results can provide guidance for the effective warning and prediction of floods in Guangdong.

Finally, in future studies, the use of hydrological model including more details of dams and reservoirs will be helpful for improving the accuracy of the model outputs, and considering the impact of human activities such as dams and reservoirs in the study of flood hazards would be beneficial for adaptation and mitigation measures against the typhoon-induced flood hazards in this vulnerable region. Moreover, finer spatial-temporal resolution is suggested in future studies to better understand the relations between the precipitation and floods characteristics.