Multi reanalysis data driven SWAT model building and its application in hydrology response to climate change in cau river basin of northern vietnam

分 类 号 密 级 U D C 编 号 10486 博 士 学 位 论 文 基于多源再分析数据的 SWAT 模型构建及 其在越南北部 Cau 河流域水文对气候变化 的响应研究 研 究 生 姓 名 ： DAO DUY MINH 指 导 教 师 姓 名 、 职称 ： 陈. Investigating the possibility of CFSR and CMADS data in hydrometeorological studies in the Cau river basin, Northern Vietnam

分 类 号 - 密 级

博 士 学 位 论 文

基于多源再分析数据的 SWAT 模型构建及 其在越南北部 Cau 河流域水文对气候变化

二〇二二年五月

Multi reanalysis data-driven SWAT model building and its application in hydrology response to climate change in Cau river basin of northern Vietnam

Multi reanalysis data-driven SWAT model building and its application in hydrology response to climate change in

Cau river basin of northern Vietnam

By

DAO DUY MINH

Ph.D Dissertation

Submitted to

State Key Laboratory of Information Engineering in Surveying,

Mapping and Remote Sensing (LIESMARS)

CARTOGRAPHY and GEOGRAPHIC INFORMATION SYSTEM

Prof Xiaoling Chen A Prof Jianzhong Lu

May, 2022

(1) 探讨 CFSR 和 CMADS 数据用于越南北部 Cau 河流域水文气象研究的可能性

由于第 3 章的研究重点是评估水文研究中重新分析数据的可能性，因此本节保留了GMS 控制模型中修正的参数，在 SWAT 模型中使用月尺度的 CMDAS 天气数据集(以下

简称 SWAT 模型，使用 CMADS 气象数据和校准的 GMS、SUC-CG 2011)和验证(2012-2013)期间，利用 GCM、CMADS 和 SUC-CG 在 Gia Bay 水物站的模拟结果对观测流量进行验证。根据每月时步的推荐性能评级记录结果为“良好”，将评价指标采用 R2>0.8 和 NSE>0.7。这些指标的结果略优于使用常规的 CMADS 进行

参数)。在校准(2009-校准, PBIAS 值达到-5.47 和-9.3% (相比之下，CMADS 分别为-16.19 和-19.35%)。分析表明，与传统策略相比，该方法显著提高了模型的性能。如果对参数进行校准以提高模型的性能，这种方法为水文研究重新分析数据的潜力提供了一种新的解决方案。

(2) 气候预测的降尺度

第四章的研究重点是气候变化项目中全球尺度的气象数据降尺度到高分辨率局部尺度数据。偏差校正空间分解方法 BCSD (Wood et al 2004)被认为是最可靠、最有效的方法之一，在世界许多地区的各种气候相关影响评估研究中被广泛采用。在本研究中，我们首先对 BCSD 方法进行了详细的描述，以方便未来的用户，这在以前的研究中没有很好的文献记录。将 BCSD 降尺度过程分为偏置校正(BC)和空间分解(SD)两个主要阶段，将 GCM 数据从 1°× 1°的中间分辨率转化为 0.1°× 0.1°的目标分辨率。据我们所知，这是在越南盆地研究中发现的最佳空间分辨率。

(3) 气候变化对 Cau 河流域水文过程的影响

CMADS 数据在访问、数据使用和在中国的研究中令人鼓舞的表现方面已被广泛使用，并具有许多便利性。 我们的研究是在越南水文研究中引入该数据集的首次尝试。 此外，首次使用 GCM-CMIP6 数据对 CRB 上的温度、降水和径流变化进行的预测是最新的，目前尚未广泛使用

论文原创性声明

本人郑重声明： 所呈交的学位论文， 是本人在导师指导下，独 立进行研究工作所取得的研究成果。除文中已经标明引用的内容外, 本论文不包含任何其他个人或集体已发表或撰写的研究成果。对本 文的研究做出贡献的个人和集体，均已在文中以明确方式标明。本声明的法律结果由本人承担。

2022 年 04 月 25 日

of climate change on the water resources is extremely important for prevention and mitigation actions to be taken The Soil Water Assessment Tool (SWAT), a semi-distributed model, was developed to analyze the impacts of land use and climate changes on discharge, erosion, sedimentation, and water quality in gauged and ungauged watersheds (Arnold et al., 1998) SWAT has received international acceptance as a robust interdisciplinary catchment-scale modeling tool because user-friendly nature, broad application capability, and the fact that is well-evaluated, well-promoted, and well-supported

Recent studies by the United Nations Environment Programme (UNEP) indicate that Vietnam is one of the countries most affected by climate change with the air temperature will increase by approximately 1,3 to 4°C by end of the 21st century Under these circumstances, water sources in rivers including the Cau river basin (CRB), a large river in northern Vietnam may be adversely affected Surprisingly, this area has only been recognized for studies in the direction of assessing the current state of surface water quality Therefore, a thorough understanding of the current status and changing trends of hydrological processes under changing climate conditions in the CRB for developing sustainable water resources management in the state

Investigating the possibility of CFSR and CMADS data in hydrometeorological studies in the Cau river basin, Northern Vietnam

In Chapter Three, the potential application of two GCPs, the China Meteorological

Assimilation Driving Datasets for the SWAT model (CMADS) and Climate Forecast System Reanalysis (NECP-CFSR), are compared for the first time with data from ground-based meteorological stations over the CRB, northern Vietnam These products are used because they have higher spatial resolutions than other products and are openly available for the study areas, covering both temperature and precipitation, and can be used immediately in flow simulations This is a major advantage of CFSR and CMADS over satellite precipitation data that often lack associated temperature data and heterogeneous time scales

Major input data for SWAT include DEM, LULC, soil properties, and daily weather data (includes grid points and ground measurement stations located around or covering the catchment area) The period for collection and processing from 1 January 2008 to 31 December

2013 to ensure consistency in the evaluation and comparison of the performances of the input data The 2012 ArcSWAT version, an interface in ArcGIS used to perform simulations

controlled by CFSR_, CMADS_, and GMS_ The lack of gauge stations is a major issue in different parts of the world, including the CRB Besides, some uncertainties may arise during interpolation of measuring stations with grid-based monitoring data, so the evaluation is limited between grids containing corresponding measured observed values Hence, the climate aspect comparison was conducted using the point-to-grid approach, where the gauge stations were directly compared to their respective grids’ values

The mean CC value of Tmax and Tmin obtained from CFSR is > 0.92, while that of CMADS is > 0.96 In addition, the MAE ranged from 0.95 to 2.47, and the RMSE varied from 1.27 to 2.85 indicating that the GCPs are in good agreement with the temperature variation at the observation stations Although the negative PBIAS value at most stations reflects that both the CFSR and CMADS data tend to underestimate the Tmax and Tmin temperatures but CFSR and CMADS can be used as an alternative to GMS in the CRB hydrometeorological studies A difference is found in that the CMADS values underestimated the actual precipitation, with a PBIAS value of -16.64%, while CFSR overestimated with a PBIAS of 99.2% Therefore, the MAE value of CMADS was much lower than that of CFSR, 5.7 and 8.01 mm/day, respectively Furthermore, the analysis results of the seasonal statistical indicators obtained from the CFSR data show the largest mean errors, with MAE and RMSE values that are too large As expected,

at the pixel scale in the basin, the CFSR rainfall data was overestimated over most of the basin, with a prevalence value between 60% and 150% In contrast, the rainfall data of CMADS tends

to underestimate with an average PBIAS of -16%, but the data exhibit different states rainfall

is underestimated in the western mountains the while the data have slightly higher ratings in the southern plains

In areas with tropical climates such as the Cau river, rainfall is the major source and greatly affects the runoff simulation results The analysis showed that the rainfall data obtained from GMS and CMADS reached an agreement better than the agreement between CFSR and GMS In general, the SWAT model based on the GMS data is best suited during the calibration and validation periods at both daily and monthly scales The simulated flow reproduced by

SWAT_GMS at Gia Bay station is “Good”, with NSE> 0.79 and R2>0.68 The simulations performed using the SWAT_CMADS tend to underestimate the observed flow, with PBIAS values varying from -16.19 to -19.35%, but with R2> 0.76 and NSE> 0.78; thus, flow simulations performed by CMADS data were within "satisfactory" on the monthly scale according to the given criteria Finally, the SWAT_CFSR is not suitable for flow simulations over the CRB basin with, R2 and NSE values that are "Unsatisfactory" based on the given

criteria Some studies have also found that integrating temperature data from CFRS with the precipitation data of the other GCPs did not cause any difference compared to conventional simulations mainly because these data overestimate the actual precipitation values

Because the research focus of Chapter 3 is to evaluate the possibility of re-analytical data in hydrological studies, in this section the parameters corrected in the GMS control model are preserved, and use CMDAS weather dataset in the SWAT model on a monthly scale

(hereinafter referred to as SWAT model Using CMADS's meteorological data and Calibrated

parameters of GMS, SUC-CG) The observed flow was used for validation with simulation

results by GCM, CMADS, and SUC-CG at Gia Bay hydrological station during calibration (2009-2011) and validation (2012-2013) period The evaluation indicators with R2 value > 0.8,

while NSE > 0.7 records result as “good” by the recommended performance rating for the

monthly time step These indicators show slightly better results than using calibration by conventional CMADS with PBIAS values reaching -5.47 and -9.3% (compare to -16.19 and -19.35% for CMADS, respectively) The flow tends to peak in August, which is consistent with the rainiest times of the year for GMS, CMADS, and even SUC-CG However, simulation results from SUC-CG reproduce better at peak flows and degradation phases than CMADS The obtained flow curves are closer to the hydrological station than the CMADS in the flood season (May to October) showing SUC-CG can get better results than conventional CMADS simulations Despite the same tendency to underestimate the actual flow as SWAT_CMADS but SWAT_SUC-CG has a better PBIAS value The analyzes have shown that this method has significantly improved the performance of the model compared with the conventional strategy This approach provides an additional new solution to the potential of reanalyzed data for hydrological studies if the parameters are calibrated to improve the performance of the model

If the model input, especially the precipitation variable, is verified before application in hydrological studies (e.g CFSR) it gives the modeler confidence in the model outputs

Projections of Future Climate Change over the Cau river basin Using the BCSD Downscaling Method

Downscaling from global-scale meteorological data to high-resolution local-scale data

in climate change projects is the research focus of Chapter Four Global climate models

(GCMs) are robust tools for quantitatively assessing climate change impacts However, GCMs outputs are insufficient to provide accurate information for local to regional scale needs due to their inadequately coarse horizontal resolutions (typically at 100-300 km) The Bias Correction Spatial Disaggregation method, BCSD (Wood et al 2004) is widely used in climate-related impact assessment studies throughout the world and is regarded as one of the most trustworthy and successful methodologies In this study, we first present a detailed description of the BCSD method for the convenience of future users, which has not been well documented in previous research

Firstly, the observed station data were interpolated to a 0.1° x 0.1° gridded dataset

(hereinafter called OBS) by using the interpolation techniques for T2m (mean temperatures), daily Tmax, Tmin, and rainfall The newly-created gridded OBS dataset will be used further in this study to bias-correct GCM data and to estimate future climate patterns in the CRB Then, The BCSD downscaling process was divided into two major stages, namely Bias Correction (BC) and Spatial Disaggregation (SD), to spatially translate GCM data from the intermediate

resolution of 1° × 1° to the targeted high-resolution of 0.1° × 0.1° To our knowledge, this is the best spatial resolution found in a study of a basin in Vietnam

The correlation of 5 representative GCMs (including CNRM-ESM2-1, EC-Earth3, GFDL-ESM4, HadGEM-GC31-LL, MPI-ESM1-2-HR) is downscaled with meter-based data

in the CRB for the period 1985-2014 Based on the locations of all points on the scatter plot, it can be seen that BCSD produces similar monthly mean temperature and mean monthly precipitation outputs at all GCMs at the observation station in the basin The accuracy of the GCMs with CC is mostly greater than 0.99 for both T2m, Tmax and Tmin The correlation results of cumulative monthly observed precipitation (mm/day) and GCMs data show values in the range of 0.994 to 0.998 during this period Dimensional and dimensionless measures are also recommended in this study to evaluate the performance of the model these results suggest that the statistics are within the reasonable range between representative GCM models and

observation stations Besides, the calculation results from a refined index of agreement (dr),

indicate that the value of the BCSD model error, represented by MAE, is lower than the mean, implying that BCSD values can be reasonably used in the input of future climate/hydrology scenarios

Future scenarios are downgraded for climate variables (precipitation, T2m, Tmax, and Tmin) to detect the general trend for the period 1985-2100 in the CRB Accordingly, a profound warming trend is recorded with the annual average Tmax and Tmin both increasing at all future meteorological stations and consistent with the increasing trend in average temperature At the end of this century, SSP5-8.5 makes the worst assumption with increases in Tmax and Tmin of 3.3oC and 3.2oC respectively, significantly higher than the scenario SSP2-4.5 With regard to precipitation, the results showed an increasing trend at all SSPs in the near-future (the 2030s) and mid-future (the 2060s); while SSP5-8.5 showed the opposite trend with the decline of average annual rainfall in the distant future period (the 2080s) In general, these outcomes imply that the CRB is likely to be hotter in future periods, which may cause potential issues relating

to agricultural activities and water consumption

Impacts of Climate Change on Hydrology Processes in the Cau river basin

The projected changes in climate will have direct and indirect effects on the natural

environment as well as on human society, especially on hydrology and water resources In Chapter Five, we introduce a quantitative assessment of the changes in the flow regime of the

CRB under climate change impacts First, historical streamflow on the basin was simulated from topography, land cover, soil, and ground weather observations by the SWAT model Second, project streamflow on the basin by inputting climate change data under SSP scenarios over the twenty-first century into a well-validated SWAT model Finally, differences in flow regime between climate change scenarios and baseline period were analyzed

The calculation results of the water balance in climate change scenarios show that precipitation will increase (2-12%), while ET will decrease (2-7%), leading to an increase in runoff (9-34%) compared to the baseline period (1997-2013) In terms of precipitation, climate projections in both scenarios SSP2.45 and 5.85 show a significant upward trend in the middle

of the dry season or the wet season while the decreasing trend of rainfall occurs mainly at the

beginning of the dry season (December) or the end of the dry season (April) or the beginning

of the wet season (May) The values of ET are identical with an upward trend dominating, except for a decrease at the end of the dry season and the first half of the wet season For water yield, the flow in the future tends to increase in both wet and dry seasons Flow may decrease

at the end of the dry season (April) and the beginning of the wet season (May)

The discharge at the basin outlet in all scenarios will increase from 6 to 53% compared

to the baseline For the SSP2-4.5 scenario, the increased rate of streamflow is higher in the rainy season than in the dry season in the short and medium-term However, this value is stronger in the dry season than in the rainy season in the long term For the SSP5-8.5 scenario, streamflow acceleration is higher in the dry season than in the rainy season in all three periods of this century Annual flow tends to increase more strongly in two periods 2021-40 and 2041-60 than

in the 2080-99 period On the other hand, the impact of climate change will make streamflow higher in most of the dry season (from November to March of the following year), and the rainy season (July-October) However, the flow projections during the transition period between the dry and wet seasons (April - June) are heterogeneous

Prediction of runoff, as well as extreme flow, is important for water management in the CRB Accordingly, The changes in the high flow (Q5) under SSP2-4.5 showed an increasing tendency in three time periods (2021–40, 2041-60, and 2080–99) whereas, in the SSP5-8.5 scenario the high flow increases in the short term but decreases in the medium term and the long term For the low flow (Q95), the predicted value will decrease in the middle period for SSP2-4.5 and the last period under SSP5-8.5 The final content of this chapter covers the annual mean of surface runoff, soil water content, and ET at the sub-basin scale under SSPs over three periods (2021–2040, 2041-60, and 2080-99) Whereby, surface runoff increases in most of the sub-basins according to SSP2-4.5, especially in the upstream and midstream regions of the basin in the two periods 2021-2040 and 2041-2060 But it decreased in all sub-basins during this century according to SSP5-8.5 Soil water content and ET are projected to increase in most sub-basins under climate change scenarios However, the rate of rising of the two water balance components will higher in SSP5-8.5 than in SSP2-4.5

Conclusion

Generally, the results obtained are very encouraging, showing that hydrological processes in the Cau river have benefited from the SWAT model The findings of this study can assist water resource managers in making informed water use decisions, creating public policies that support appropriate use, in applying mitigation and prevention measures, to ensure water security in the basin and agronomically similar basins

Keywords: Cau river basin, SWAT model, CFSR, CMADS, CMIP6, BCSD method, streamflow, downscale, climate change

摘要

根据政府间气候变化专门委员会(IPCC, 2013)的数据，从 1880 年到 2012 年，全球地表的平均温度升高了 0.85°C，导致降水发生变化，并对水文过程产生了重大影响。研究发现，温度和降水的变化会影响流域的出水量、蒸散发(ET)、地表径流以及洪水的规模和频率。因此，建立水文模型以捕捉当前的水文过程并评估气候变化对水资源的影响对于预防和减轻这部分影响有着极为重要的意义。土壤水分评估工具(SWAT)是一种半分布模型，常被用于分析土地利用和气候变化对流域的流量、侵蚀、泥沙和水质的影响(Arnold et al.， 1998)。SWAT 作为一种强大的跨学科流域规模的建模工具，因其具有对用户友好的特性、广泛的应用能力，以及得到了良好的评估、推广和支持，已经得到了国际上广泛的认可。

联合国环境规划署(UNEP)最近的研究表明，越南是受气候变化影响最严重的国家之一，到 21 世纪末，越南的气温将上升约 1.3 至 4°C。在这种情况下，包括 Cau 河流域(CRB)在内河流的水源可能会受到负面的影响。Cau 河流域是越南北部的一条大河。令人惊讶的是，这一区域只在评估地表水水质现状的研究中才受到关注。因此，需要深入了解气候变化条件下 Cau 河流域水文过程的现状和变化趋势，为开展 Cau 河流域状态的可持续水资源管理提供依据。

探讨 CFSR 和 CMADS 数据用于越南北部 Cau 河流域水文气象研究的可能性

在第三章中，本文首次将 SWAT 模型和气候预报系统再分析(NECP-CFSR)下的中国气象同化驱动数据集(CMADS)这两种潜在的 GCP 与越南北部 CRB 的地面气象站数据进行了比较。之所以使用这些产品，是因为它们比其他产品具有更高的空间分辨率，且数据在研究区域内是公开的，它们包含了温度和降水数据，并可立即用于径流模拟。这是 CFSR 和 CMADS 相对于缺乏相关温度数据和非均匀时间尺度的卫星降水数据的一个主要优势。

SWAT 的主要输入数据包括 DEM、LULC、土壤属性和每日天气数据(包括位于集水区周围或覆盖集水区的格网和地面测量站)。为了确保输入数据性能评价和比较的一致性，本研究收集和处理了 2008 年 1 月 1 日至 2013 年 12 月 31 日间的数据。本研究

使用 2012 版 ArcSWAT——ArcGIS 中的一个接口来执行由 CFSR_、CMADS_和 GMS_控制的模拟。测站数据的缺乏是世界各地的一个主要问题，包括 CRB。此外，基于网格监测数据的测站插值过程中可能存在一定的不确定性，因此评价仅限于包含在相应实测观测值的网格之间。因此，采用点到格网的方法进行气候方面的比较，在这种方法中，测量站直接与各自格网的值进行比较。

由 CFSR 得到的 Tmax 和 Tmin 的 CC 均值>0.92, 与此同时，CMADS>0.96。此外，MAE 在 0.95 ~ 2.47 之间，RMSE 在 1.27 ~ 2.85 之间，这表明 GCPs 与观测站温度变化

的一致性较好。虽然大多数台站的 PBIAS 为负值——这表明 CFSR 和 CMADS 数据都有低估 Tmax 和 Tmin 的倾向，但 CFSR 和 CMADS 在 CRB 水文气象研究中为替代 GMS提供了可能。两者的差异在于，CMADS 低估了实际降水，PBIAS 为-16.64%，而 CFSR高估了实际降水，PBIAS 为 99.2%。因此，CMADS 的 MAE 值远低于 CFSR，分别为5.7 mm/d 和 8.01 mm/d。此外，从 CFSR 数据得到的季节统计指标的分析结果显示，平均误差最大，MAE 和 RMSE 值都过大。正如预期的那样，在流域的像素尺度上，CFSR降水数据在大部分流域被高估，普遍值在 60%到 150%之间。相比之下，CMADS 的降水数据往往低估，平均 PBIAS 为-16%，但数据显示西部山区不同州的降水被低估，而在南部平原则略高。

在热带气候地区，如 Cau 河，降雨是径流量的主要来源，极大地影响其模拟结果。分析表明，GMS 和 CMADS 的降水数据比 CFSR 和 GMS 的降水数据具有更好的一致性。一般来说，基于 GMS 数据的 SWAT 模型在每日和每月的校准和验证期间最为适合。SWAT_GMS 在 Gia Bay 站模拟的流场效果为“良好”，NSE>0.79，R2>0.68。使用SWAT_CMADS 进行的模拟往往会低估观测到的流量，PBIAS 值在-16.19%到-19.35%之

在月尺度上是“令人满意的”。最后，SWAT_CFSR 不适合 CRB 盆地的流动模拟，根

CFRS 的温度数据与其他 GMP 的降水数据相结合不会造成任何差异，主要是因为这些数据高估了实际降水值。

由于第 3 章的研究重点是评估水文研究中重新分析数据的可能性，因此本节保留

了 GMS 控制模型中修正的参数，在 SWAT 模型中使用月尺度的 CMDAS 天气数据集(以下简称 SWAT 模型，使用 CMADS 气象数据和校准的 GMS、SUC-CG 参数)。在校准(2009-2011)和验证(2012-2013)期间，利用 GCM、CMADS 和 SUC-CG 在 Gia Bay 水物站的模拟结果对观测流量进行验证。根据每月时步的推荐性能评级记录结果为“良

进行校准, PBIAS 值达到-5.47 和-9.3% (相比之下，CMADS 分别为-16.19 和-19.35%)。流量在 8 月达到峰值，这与 GMS、CMADS 甚至是 SUC-CG 一年中降雨量最多的时期一致。然而，在流量的峰值和退化阶段，SUC-CG 的模拟结果比 CMADS 更好。汛期(5

~ 10 月)的流量曲线比 CMADS 更接近水文站，表明 SUC-CG 模拟的结果优于常规CMADS 模拟。尽管与 SWAT_CMADS 一样有低估实际流量的倾向，但 SWAT_SUC-

CG 有更好的 PBIAS 值。分析表明，与传统策略相比，该方法显著提高了模型的性能。如果对参数进行校准以提高模型的性能，这种方法为水文研究重新分析数据的潜力提供了一种新的解决方案。如果模型输入，特别是降水变量，在应用于水文研究(如CFSR)之前得到验证，则会使建模者对模型输出更有信心。

气候预测的降尺度

第四章的研究重点是气候变化项目中全球尺度的气象数据降尺度到高分辨率局部尺度数据。全球气候模型(GCMs)是定量评估气候变化影响的有力工具。然而，由于GCM 的粗水平分辨率(通常在 100-300 公里)不足，其产出不足以为地方到区域尺度的需要提供准确的信息。偏差校正空间分解方法 BCSD (Wood et al 2004)被认为是最可靠、最有效的方法之一，在世界许多地区的各种气候相关影响评估研究中被广泛采用。在本研究中，我们首先对 BCSD 方法进行了详细的描述，以方便未来的用户，这在以前的研究中没有很好的文献记录。

首先，通过基于 T2m、日 Tmax、Tmin 和降雨量数据，将观测台站数据插值到0.1°x 0.1°的网格数据集(以下简称 OBS)。新创建的网格化 OBS 数据集将在本研究中进一步用于校正 GCM 数据的偏差，并在 CRB 中估计未来的气候模式。然后，将 BCSD降尺度过程分为偏置校正(BC)和空间分解(SD)两个主要阶段，将 GCM 数据从 1°× 1°的中间分辨率转化为 0.1°× 0.1°的目标分辨率。

利用 1985-2014 年地球观测资料，对 5 1、EC-Earth3、GFDL-ESM4、HadGEM-GC31-LL、MPI-ESM1-2-HR)进行了降尺度。从散点图中可以看出，BCSD 在该流域观测站所有 GCM 的月平均温度和月平均降水输出是相似的 T2m、Tmax 和 Tmin 的精度大多高于 0.99。月累计观测降水(mm/day)与GCMs 数据的相关性在 0.994 ~ 0.998 之间。本文还建议采用量纲测度和无量纲测度来评价模型的性能，结果表明模型与观测站的统计量在合理范围内。此外，根据改进的一致性指数(dr)的计算结果表明，以 MAE 为代表的 BCSD 模型误差低于平均值，表明BCSD 值可以合理地用于未来气候/水文情景的输入。

个代表性的大气环流模式(CNRM-ESM2-对气候变量(降水、T2m、Tmax 和 Tmin)的未来情景进行降级，以检测 CRB 中1985-2100 年期间的总体趋势。未来所有气象站的年平均 Tmax 和 Tmin 均有所增加，并与平均气温的增加趋势一致，呈现出明显的变暖趋势。在本世纪末，SSP5-8.5 情景最为糟糕，Tmax 和 Tmin 分别增加 3.3℃和 3.2℃，显著高于 SSP2-4.5 情景。降水方面，近期(2030 年代)和中期(2060 年代)的 SPS 均呈增加趋势；而 SSP5-8.5 在遥远的未来期(2080 年代)，随着年均降水量的减少呈现相反的趋势。总的来说，这些结果意味着CRB 在未来时期可能会更热，这可能会导致与农业活动和用水有关的潜在问题。

气候变化对 Cau 河流域水文过程的影响

预测的气候变化将对自然环境和人类社会产生直接和间接的影响，特别是对水文和水资源。在第五章中，我们介绍了气候变化影响下 CRB 流态变化的定量评估。首先，利用 SWAT 模型根据地形、土地覆盖、土壤和地面天气观测数据模拟流域的历史径流。其次，通过将 21 世纪 SSP 情景下的气候变化数据输入到一个经过充分验证的

SWAT 模型中来预测流域的流量。最后，分析了气候变化情景与基线期之间的流态差异。

气候变化情景下的水平衡计算结果表明，与基线期(1997-2013 年)相比，降水将

增 加(2-12%)，ET 将 减少(2-7%)，导致径 流量增加 (9-34%)。降 水、气候预测情景SSP2.45 和 5.85 在旱季和雨季显示显著的上升趋势，降雨量的减少趋势主要发生在旱季(12 月)的开始、旱季的结束(4 月)或雨季的开始(5 月)。ET 值则基本保持一致，除枯水期末和雨季上半期出现下降外，均以上升趋势为主。就产水量而言，未来的流量在雨季和旱季都趋于增加。在旱季结束(4 月)和雨季开始(5 月)，流量可能会减少。

与基线相比，在所有情景中，流域出口处的流量将从 6%增加到 53%。对于SSP245 情景，在中短期内，雨季的径流增加速率高于旱季。然而，从长期来看，旱季的这一数值要高于雨季。对于 SSP585 情景，在本世纪的所有三个时期，旱季的径流加速都高于雨季。与 2080-99 年相比，2021-40 年和 2041-60 年期间的年流量增长更为强劲。另一方面，气候变化的影响会使旱季(11 月至次年 3 月)和雨季(7 - 10 月)的径流增加。旱季和雨季(4 - 6 月)过渡期间的流量预测是不一致的。

径流和极端流量的预测对于 CRB 的水资源管理非常重要。因此，在 SSP245 情景下，高流量(Q5)的变化在三个时间段(2021-40 年、2041-60 年和 2080-99 年)呈现增加趋势，而在 SSP585 情景下，高流量在短期内增加，但在中长期内减少。对于小流量(Q95)， SSP245 的中间时段和 SSP585 的最后时段预测值会降低。本章的最终内容包括三个时段(2021-2040 年、2041-60 年和 2080-99 年)SSPs 下的子流域尺度的年平均地表径流、土壤水分和 ET。根据 SSP245 分析，大部分子流域的地表径流均呈增加趋势，特别是在 2021-2040 年和 2041-2060 年两个时段，流域上游和中游区域的地表径流均呈增加趋势。但根据 SSP585 的数据，本世纪以来各子盆地均呈下降趋势。在气候变化情景下，大部分子流域土壤水分和蒸散量均呈增加趋势。而 SSP585 的两种水分平衡组分的上升速率高于 SSP245。

结论

总的来说，本研究得到的结果是令人鼓舞的，它表明 Cau 河的水文过程适用于SWAT 模型。这项研究的结果可以帮助水资源管理者做出有益的用水决策，制定用水的公共政策以保障必要的使用，实施缓解和预防措施，以确保该流域和农业条件类似流域的水安全。

尺度、气候变化

TABLE AND CONTENTS

ABSTRACT i

摘要 vi

TABLE AND CONTENTS x

LIST OF FIGURES xiii

LIST OF TABLES xvi

LIST OF ACRONYMS AND ABBREVIATIONS xvii

Chapter One: Introduction 2

1.1 Backgrounds 2

1.2 Problem Statement 4

1.3 Research Objectives 5

1.4 Thesis Organization 6

Chapter Two: Literature Review and Data Sources 9

2.1 Introduction 9

2.2 Hydrological cycle and review of Vietnam’s water resources 9

2.2.1 Hydrological Cycle Review 9

2.2.2 Catchment Hydrology 11

2.3 Catchment Hydrologic Modelling 12

2.3.1 Importance of Hydrologic Models 12

2.3.2 Hydrologic Modeling 12

2.3.3 Why did choose SWAT Applications in Hydrologic Assessment in the CRB? 14

2.4 The Soil and Water Assessment Tool (SWAT) Model 15

2.4.1 General Description of The Model 15

2.4.2 Modeling Approach and Structure 16

2.4.3 Model Calibration and Uncertainty Analysis 18

2.4.4 Overview of alternative climate products for SWAT modelling 20

2.5 Future Climate Model 22

2.5.1 Climate Change Impacts on Hydrology 22

2.5.1.1 ET 23

2.5.1.2 Surface Runoff 23

2.5.1.3 Streamflow and Flood 24

2.5.2 Future Emissions Scenarios for Climate Changes 24

2.5.2.1 Overview of climate models 24

2.5.2.2 CMIP6 and Future emission scenarios 25

2.6 Study Area and Data Source for SWAT Model 27

2.6.1 Description of the study area 27

2.6.2 SWAT Model Input for the CRB 28

2.6.2.1 Digital Elevation Model (DEM) 28

2.6.2.2 Land use/land cover data 28

2.6.2.3 Climatic and Hydrological Data 29

2.7 Conclusion 31

Chapter Three: StreamFlow Simulation used CFSR and CMADS Data 33

3.1 Introduction 34

3.2 Materials and methods 36

3.2.1 Location of the study area 36

3.2.2 Model input data 37

3.2.2.1 Digital Elevation Model (DEM) 37

3.2.2.2 Soil type and characteristics 38

3.2.2.3 Reanalysis datasets 38

3.2.2.4 GMS Data 39

3.2.3 Model calibration, validation and sensitivity analysis 41

3.2.3.1 The calibration method and approach 41

3.2.3.2 Uncertainty and model performance indices 41

3.2.4 Evaluation indicators 42

3.2.4.1 Index evaluates temperature and precipitation 42

3.2.4.2 Flow indicators 43

3.3 Results and Discussions 45

3.3.1 Temperature/Precipitation validation 45

3.3.1.1 Compare CFSR and CMADS temperatures using GMS data 45

3.3.1.2 Temporal distribution of Precipitation 47

3.3.1.3 Comparison in the spatial scale (pixels) 48

3.3.2 Precision of precipitation event detection 50

3.3.3 Evaluate the ability to capture extreme weather events 51

3.3.4 Parameter sensitivity analysis and calibration 54

3.3.5 Peformance of simulate streamflow 55

3.3.6 Calibration and validation of the SWAT model against parameters calibrated from GMS and CMDAS weather data 57

3.3.7 Discussion 59

3.4 Conclusion 62

Chapter Four: Projections of Future Climate Change over the Cau river basin Using the

BCSD Downscaling Method 64

4.1 Introduction 64

4.2 Data and methods 66

4.2.1 Study area and Data Sources 66

4.2.2 The BCSD approach 69

4.2.3 Additive and Multiplicative Methods of Change Factors 71

4.2.4 The Innovative-Şen Trend Analysis Method 72

4.2.5 Evaluation of Model Performance 73

4.3 Results and Discussions 74

4.3.1 Correlation between GCMs and Data Observed In Situ 74

4.3.2 Future changes in annual precipitation and temperature 77

4.3.3 Changes in Future Seasonal Maximum/Minimum Temperature and Precipitation 81

Chapter Five: Impacts of Climate Change on Hydrology Processes in Cau river basin 87 5.1 Introduction 87

5.2 Material and Methods 88

5.2.1 Study area 88

5.2.2 SWAT model 89

5.2.3 Input Data 90

5.2.4 Methodology 93

5.3 Results and Discussions 94

5.3.1 Model calibration and validation 94

5.3.2 Impact of climate change on flow regime 96

5.3.2.1 Water balance components 96

5.3.2.2 Streamflow and surface runoff 98

5.3.2.3 Peak flows 99

5.3.2.4 Hydrological extremes 100

5.3.3 Impact of climate change on sub-basin scale hydrology 102

5.4 Conclusion 104

Chapter Six: Conclusions and Recommendations 107

6.1 Conclusions 107

6.2 Thesis New Findings 109

6.3 Recommentdations and Future Work 110

BIBLIOGRAPHY 113

PUBLICATIONS 124

LIST OF FIGURES

Figure 1.1: Framework of the research: (a) Present input data; (b) Future input data; (c) Process in the SWAT model, and (d) Simulation results 7 Figure 2.1: Pictorial representation of hydrological cycle (Source: http://water.usgs.goc, accessed on 25/03/2022) 10 Figure 2.2: Classification of hydrological models (Source: Singh, 1988; Xu, 2002) 14 Figure 2.3: Components of the hydrologic balance of the SWAT model (Adapted from Arnold et al., 1998) 16 Figure 2.4: Schematic respresentation of a HRU hydrologic cycle (Source: Neisch et al., 2001) 17 Figure 2.5: SWAT Model Components and Inputs (Modified from Neitsch et al., 2002) 18 Figure 2.6: Shared Socieconomic Pathways and CMIP6 Scenarios 26 Figure 2.7: CMIP6 includes future CO2 emission scenarios as well as historical CO2emissions 26 Figure 2.8: Study region map of the CRB: (a) Digital elevation model (DEM) and monitoring networks, (b) Land use map 29 Figure 2.9: Conceptual framework of SWAT model and its setup 30 Figure 3.1: Study region map of the Cau River Basin: (a) digital elevation model (DEM)

and (b) land use map 40

Figyre 3.2: Box plots of daily maximum (a-d) and minimum (e-h) temperatures from CFSR and CMADS at the BacKan, DinhHoa, ThaiNguyen and BacNinh meteorological stations 46 Figure 3.3: Spatial distributions of the correlation coefficient (CC) on the daily (a-b) and monthly (c-d) scales and of MAE (mm/month) (e-f) and PBIAS (%) (g-h) on the monthly scale in the Cau River basin over the period from 2008-2013 49 Figure 3.4: Occurrence frequencies (a) and relative contributions (b) of daily-scale rainfall thresholds obtained from the CFSR, CMADS and GMS data for the period 2008-2013 51 Figure 3.5: Number of days when hot weather occurred in the period 2008-2013 54 Figure 3.6: Observed streamflow and simulations performed using the GMS-, CFSR-, and CMADS-driven models at the daily scale over the CRB 56 Figure 3.7: Observed streamflow and simulations obtained using the GMS-, CFSR-, and CMADS-driven models at the monthly scale over the CRB 57

Figure 3.8: Calibration (1981-1992) and validation (1993-2002) of the SWAT model with simulations based on ground-based meteorological stations (GMS) and models controlled by CMADS and SCU-CG (SWAT model using CMADS's meteological data and Calibtrated parameters of GMS) uses monthly observed flow at Gia Bay hydrological station 58 Figure 4.1 Geographical location of the CRB and meteorological stations collected data Tmax, Tmin and rainfall 67 Figure 4.2 The maijor steps of BCSD downscaling 69 Figure 4.3: The schema of the BCSD algorithm applied in this study 70 Figure 4.4: Method to evaluate future scenarios by applying the multiplicative and additive change factors 72 Figure 4.5: GCMs monthly Tmax, Tmin and mean temperature and accumulated precipitation from 1985 to 2014 at the Thai Nguyen climate station in the CRB 75 Figure 4.6: Future trends of (a) SSP2-4.5 and (b) SSP5-8.5 Tmax and Tmin mean annual temperatures (1985–2100) 77 Figure 4.7: Future trends of (a) SSP4.5 and (b) SSP8.5 total mean annual rainfall (1985-2100) 77 Figure 4.8: Projected changes in annual average Tmax and Tmin temperatures in the 2030s, 2060s, and 2080s under SSPs scenarios 78 Figure 4.9: Projected changes in average annual rainfall in the 2030s, 2060s, and 2080s under SSPs scenarios 79 Figure 4.10: Box-and-whisker plots of the precipitation changes of the GCMs during the 2030s, 2060s and 2090s under SSPs 4.5 and 8.5 in the CRB, Vietnam 80 Figure 4.11: Innovative-Şen trend plots for rainfall comparing the near-, middle-, and far-future periods to the reference period under multiple scenarios 81 Figure 4.12: Change in seasonal and annual temperature for three time periods under SSP2-4.5 and SSP5-8.5 83 Figure 4.13: Seasonal and annual change in rainfall for three time periods under SSP2-4.5 and SSP5-8.5 83 Figure 5.1: Geographic location of the CRB 89Figure 5.2: DEM map (a), 2005-land cover map (b), 1995-soil map (c) of the CRB 92 Figure 5.3 Hydrometeorological monitoring network on the CRB 93 Figure 5.4: Flowchart diagram of methodology 93Figure 5.5: Simulated and observed monthly discharge at the GiaBay station for the calibration period (1998–2005) 95 Figure 5.6: Simulated and observed monthly discharge at the GiaBay station for the validation period (2006-2013) 96

Figure 5.7: Annual changes of rainfall, ET, water yield in CRB under climate change scenarios compared to baseline period (%) 97 Figure 5.8: Monthly changes of rainfall in CRB under climate change scenarios compared to baseline period (%) 97 Figure 5.9: Monthly changes of ET in CRB under climate change scenarios compared

to baseline period (%) 98 Figure 5.10: Monthly changes of water yield in CRB under climate change scenarios compared to baseline period (%) 98 Figure 5.11: Seasonal and annual changes of streamflow at the CRB outlet under climate change scenarios compared to baseline period (%) 99 Figure 5.12: Seasonal and annual changes of surface runoff in the CRB under climate change scenarios compared to baseli ne period (%) 99 Figure 5.13: Monthly changes of streamflow at the CRB outlet under climate change scenarios compared to baseline period (%) 100 Figure 5.14: Flow duration curves of the CRB at the basin outlet under climate change scenarios during (a) 2021–2040, (b) 2041-60, and (c) 2080-99 periods 101 Figure 5.15: Spatial distribution of average annual changes of surface runoff at the sub-basin scale in the CRB under climate change scenarios compared to baseline period (%) 102 Figure 5.16: Spatial distribution of average annual changes of soil water at the sub-basin scale in the CRB under climate change scenarios compared to baseline period (%) 103 Figure 5.17: Spatial distribution of average annual changes of ET at the sub-basin scale

in the CRB under climate change scenarios compared to baseline period (%) 103

LIST OF TABLES

Table 2.1: The criteria used for evaluating the SWAT model performance ratings for simulating flow at a daily time scale 20 Table 2.2: Percent of land-use/land-cover classification in the CRB 29 Table 2.3: List of input data on the CRB 30 Table 3.1: Percent of land-use/land-cover classification in the Cau river basin 38 Table 3.2: The input datasets used for the meteorological assessment and hydrological simulation in this study 39 Table 3.3: The criteria used for evaluating the SWAT model performance ratings for simulating flow at a monthly time scale 44 Table 3.4: Statistical indicators used to evaluate temperature (maximum, Tmax/minimum, Tmin) in the CFSR and CMADS data in the CRB 46 Table 3.5: Continuous statistical indicators of the CFSR and CMADS rainfall data on the CRB from 2008 to 2013 48 Table 3.6: Statistics on the total number of cold and damaging days at meteorological stations in the CRB in the period 2008-2013 53 Table 3.7: Sensitivity values of the parameters used for flow simulations by GMS_, CFSR_, and CMADS_ using the SWAT model in the CRB 55 Table 3.8: Statistical indices obtained during the calibration and validation periods of the streamflow simulations with GMS-, CFSR-, and CMADS-driven models 56 Table 3.9: Evaluate the performance for the SWAT model driven by ground-based meteorological station (GMS) weather data during the calibration and validation phase and compare with the results from the CMADS and SCU-CG-driven SWAT model with observed flow 59Table 4.1: Information on the GCM-CMIP6, and SSPs availability used in this study 68 Table 4.2: The values of MAE and MBE between observed and predicted daily series during validation Tmax, Tmin, and R for Tmax and Tmin temperature and rainfall 75 Table 5.1: Input data types of SWAT model 90 Table 5.2: Area of 2005-land cover types in CRB 92 Table 5.3: Area of 1995-soil units in CRB 99 Table 5.4: General performance ratings for recommended statistics 94 Table 5.5: List of the model calibration parameters 95 Table 5.6: The statistics quantitative indices of SWAT model performance in the calibration and validation periods 96 Table 5.7: Relative changes in extreme (high and low) flows, as well as median flow, under climate change scenarios in the CRB 101

LIST OF ACRONYMS AND ABBREVIATIONSAcronyms Corresponding Meaning

"G" "Very good"

"S" "Satisfactory"

"U" "Unsatisfactory"

"V" “Very good”

95PPUs 95 Percent Predictive Uncertainties

AR4 4th Assessment Report

ARS-USDA The United States Department of Agriculture

BCSD The Bias Correction Spatial Disaggregation method

CC The Correlation Coefficient

CMADS The China Meteorological Assimilation Driving Datasets for the

SWAT Model CMIP5-6 Coupled Model Intercomparison Project Phase 5-6

CMORPH The Climate Prediction Center Morphing

CRB The Cau river basin

CRU Climatic Research Unit

CSI Critical Success Index

CHIRPS The Climate Hazards Group Infrared Precipitation

DEM Digital Elevation Model

ERA European Centre for Medium-Range Weather Forecast ReAnalysis ERSI the Environment Rating Scales Institute

ET Evapotranspiration

FAO The Food and Agriculture Organization

FAR False Alarm Ratio

GCMs The General Circulation Models

GCPs Gridded Climate Products

GLUE Generalized Likelihood Uncertainty Estimation

GMPs Gridded Meteorological Products

GIS Geographic Information System

HRUs Hydrologic Response Units

IPCC Intergovernmental Panel on Climate Change

LAPS/STMAS The Local Analysis and Prediction System/Space-Time Multiscale

Analysis System LULC Land Use / Land Cover

MAE Mean Absolute Error

MCMC Markov Chain Monte Carlo

MORNE The Ministry of Natural Resources and Environment, Vietnam MUSLE Modified Universal Soil Loss Equation

NCAR National Center for Atmospheric Research

NECP-CFSR The National Centers for Environmental Prediction-Climate

Forecast System Reanalysis NOAA National Oceanic and Atmospheric Administration

NRC The National Research Council

NSE Nash Sutcliffe efficiency coefficient

OAT One-At-a-Time

ParaSol Parameter Solution

PBIAS Percentage Bias

PERSIANN-CDR

The Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks-Climate Data Record

PET Potential Evapotranspiration

POD Probability of Detection

R2 Coefficient of Determination (or R quared)

RCPs Representative Concentration Pathway

RMSE Root-Mean-Square Error

SAC-SMA Sacramento Soil Moisture Accounting model

SCS The Soil Conservation Service

SWAT The Soil & Water Assessment Tool model

Chapter One

Introduction

Chapter One: Introduction

1.1 Backgrounds

Water is considered the most important resource and has a direct influence on life processes on Earth Increased human population and coupled with inadequate management of natural resources, among other factors, are aggravating problems related

to uses (including agriculture, irrigation, hydropower, and tourism) and water quality [1,2] In the tropical monsoon regions, where yearly variations of precipitation and there are distinct seasonal differences as in northern Vietnam, these problems are even greater

In the last decades of the 21 st century, humanity has witnessed the appearance

of extreme events due to climate changes, such as clod, heat, droughts and floods According to a report by the Intergovernmental Panel on Climate Change (IPCC) [3] the global average surface temperature warmed by 0.85 °C from 1880 to 2012 year, being the 21st century the warmest period, causing changes in precipitation and considerably impacting hydrological processes in the river basin Higher temperature induces a higher amount and intensity of precipitation which affects hydrology [4–7] Variations in precipitation are responsible for influencing streamflow trends in different regions across the World [8–13] Meanwhile, changes in precipitation patterns directly affected the water yield, ET, and surface runoff in the region [14–17], increased precipitation also led to increases in magnitude and frequency of floods (e.g., [11,18,19] Therefore,

in order for prevention and mitigation actions to be taken, understanding the current processes and related impacts of climate change on water resources is extremely important

Recent studies by the United Nations Environment Programme (UNEP) and the IPCC indicate that Vietnam is one of the countries most affected by climate change The IPCC’s Special Report on Emission Scenarios (SRES) of IPCC indicates that the air temperature will increase by 1,3 to 4°C by end of the 21st century throughout Vietnam [3,20] Under the recorded climate changes over time, the characteristics of temperature and precipitation have fluctuated, leading to significant changes in the regional hydrological components Studies on climate change trends as well as flows based on climate change scenarios have been carried out but mainly in river basins in South Vietnam [21–23] Under these circumstances, water sources in rivers including the Cau river basin, a large river in northern Vietnam may be adversely affected Surprisingly, this area has only been recognized for studies in the direction of assessing the current state of surface water quality [24–26], with no independent studies investigating

hydrological responses to climate change and potential climate change scenarios across the Cau river basin (CRB) Therefore, a thorough understanding of the current status and changing trends of hydrological processes, especially under changing climate conditions in river basins of various scales, from which for develop sustainable water resources management in the state

Since limited hydrological information is available in many watersheds, the advent of adequate hydrological assessment tools is needed to provide flexibility to deal with poor quantity and quality input data [27] Within this context, studies focused on the understanding of hydrological processes are relevant to the knowledge of the different hydrological variables representing an important aspect of the adequate management of water resources [28] In this way, studies from different basins have implemented and improved multiple instruments related to water resources, including the hydrological models [29,30]

Hydrological modeling is an important tool to support management and making on water resources Information is often not publicly available because of the incomplete understanding of the overall flow and motion processes of the soil-atmosphere-biosystem that can be found in these models According to Abbaspour et al [1], hydrological models are important for the planning of water resources in meeting the diverse demands, helping in their sustainable use The current modeling philosophy requires that the hydrological models to be achieved is: be clearly described and the processes of calibration, validation, and sensitivity and uncertainty analysis are inherent

decision-to the model

The Soil Water Assessment Tool (SWAT), a semi-distributed, continuous-time, process-based hydrology and water quality model [31] was developed by Dr Jeff Arnold and his team at the Agricultural Research Service of the United States Department of Agriculture (ARS-USDA) to analyze the impacts of land-use changes on discharge, erosion, sedimentation, and water quality in gauged and ungauged watersheds [32] The SWAT is a versatile model that considers different hydrological and agronomic components and has been applied by many governmental and private institutions, as well as universities and other institutions interested in supporting decision-making in the management of water resources [30] An extensive number of analyzes have already been carried out with SWAT worldwide, including studies on climate change as reported

by [22,33–35]

In this thesis, we hypothesize that modeling is an important tool to support management and decision-making on water resources Specifically, the SWAT model can efficiently simulate the hydrological processes occurring in the CRB, and possible discrepancies in the results obtained can be minimized by the calibration and validation procedures from the tools of this model The study also assumes that, after the calibration of the model, inferences related to the hydrological processes under different future climate changes across the basin can be made based on the outputs of the simulation scenarios provided by the model

1.2 Problem Statement

Since the hydrologic cycle is a complex process that is influenced by inter-linked environmental factors, hydrological models are crucial in the assessment of flow processes on Earth In this regard, the soil and water assessment tool (SWAT) is a semi-distributed model that can simulate continuous-time impacts of complex environmental factors on fresh water at different scales of study (e.g catchment or river basin scales) Applying the SWAT model requires different types and spatially explicit data These include topographic parameters, soil data, land use land cover, and weather parameters

In addition, SWAT requires time series data of streamflow, erosion, and chemical loadings, from ground-based point and non-point sources to provide for model calibration, validation, and uncertainty analysis The problem is that in watersheds such

as the CRB, hydrological data have historically been severely limited Published studies indicate that the major limitation of the development of hydrology is the lack of high-quality observations [36–38] It includes heterogeneity and data scarcity increase with catchment size, often affected by the sparse density of ground observation stations and economic resources limited; in addition, severe climate and hydrological events are infrequent but cause damage also make the values of these factors seriously decline [39,40]

It is recognized that general circulation models (GCMs) are greatly supportive of the assessment of potential climate change impacts on multiple sectors on the global scale Unfortunately, the horizontal resolution of GCMs cannot meet the requirements

of most local impact studies as it is typically in the range of 250 to 600 km Besides, the direct application of GCMs requires considerable computational resources, thereby being unfeasible for most developing countries including Vietnam Hence, many dynamical and especially statistical downscaling methods have arisen to overcome these key disadvantages of GCMs Determine the superiority of statistical versus statistical,

and/or statistical versus dynamical downscaling models is apparent that one of the most common themes of downscaling application Therefore, the construction of future daily climate data at individual meteorological stations by applying the statistical downscaling approach to primary inputs for assessing the impacts of climate change on water resources at the river basin scale is a very urgent need

These issues raise questions for the thesis to address the following innovative aspects related to the fit of the SWAT model, using alternative data sources for terrestrial meteorological observations to conduct current and future flow studies including: (a) What is the lack of data sources of ground measurement observations and reliability assessment of grid climate data commonly used in SWAT models?

(b) Ability to integrate between grid climate data sources (mainly temperature and precipitation aspects) in SWAT models with parameters corrected from actual station data?

(c) Applicability of the downscaling scale reduction model on a river basin in Vietnam with GCM-CMIP6-SSPs integrated scenarios?

(d) Uncertainty of the catchment scale reduction model when using general circulation models (GCMs)? It is recommended to use different GCM outputs based on experiments with CMIP6-SSPs in the development of future climate change scenarios

(1) Verify the representative ability of the two reanalysis datasets (CMADS and CFSR) were interpolated to regular meteorological stations in the CRB, and their respective interpolations were compared with gauge observations (OBS) according to a set of meaningful diagnostic statistics

(2) The spatial interpolation distributions of annual average precipitation and maximum/minimum temperature from these three data sources in CRB were compared,

as well as their monthly average precipitation and maximum/minimum temperature at the regular meteorological stations

(3) Investigate the effects of some extreme weather events (such as very cold, damaging cold, strong sun, and scorching hot events) and study a very heavy rain layer (R100mm/day) based on regulations in Vietnam to suit the conditions of tropical monsoon climate

(4) Implement the process of downgrading GCMs based on the BCSD method combined with reverse interpolated observation station data to create new OBS grids on the catchment scale From there as a basis for validation with the future climate data set (period 2021-2100)

(5) The difference and uncertainties of projected impacts associated with CMIP6 structure and SSPs scenarios were estimated and compared quantitatively; (6) Finally, the two reanalysis datasets, as well as future weather data of the GCM-CMIP6 and OBS were used as inputs for runoff simulation in the hydrological modeling framework aimed specifically at changes in various components of the water balance including precipitation, transpiration, total water production, soil moisture, and river discharge

GCMs-It is worthy to note that this thesis focuses on the natural cycles of the hydrologic processes only, i.e anthropogenic activities that have a significant effect on water resources (such as the LULC change, groundwater abstraction, and managed groundwater recharge) will not be considered in the analysis

1.4 Thesis Organization

This work is presented in six chapters

- Chapter 1 gives the introduction to the study In this chapter, the background, problem statement, and objectives of the study are clearly stated as well as the scope and organization of the thesis

- In Chapter 2, a general literature review is provided, including the water cycle and hydrological models at the CRB scale In the first part of the chapter, all the theories and practical perspectives related to hydrological models and descriptions of the effects

of climate change on the spatial and temporal variation of hydrological processes are presented The next part of Chapter 2 deals with the application of SWAT on CRB

including the process of model setup and configuration, parameterization, sensitivity analysis, model calibration and validation This preparation is intended to obtain basic data information on feature classes and is ready for further analysis in the next consecutive chapters

- A summary of the results obtained from the SWAT model simulation is the subject of Chapter 3 These results include temperature/precipitation validation as well

as the distribution of two reanalyzed datasets, CFSR and CMADS; the sensitivity analysis, the calibration and validation of the model comparing the measured and simulated flow in different flow gauges Chapter 4 presents downscaling technique by the BCSD method in the Cau river basin The final results of the study are the investigations of the impacts of climate change on water resources in the CRB, which are presented in Chapter 5 The examinations of the implications of climate change on water resources in the CRB, which are reported in Chapter 5, are the study's final conclusions Chapters 3, 4, and 5 follow an article format where each of the chapters is considered a stand-alone study

- The final chapter, which is Chapter 6, deals with the general discussion and recommendations for future studies

Figure 1.1 Framework of the research: (a) Present input data; (b) Future input data; (c)

Process in the SWAT model, and (d) Simulation results

Chapter Two

Literature Review and

Data Sources

Chapter Two: Literature Review and Data Sources

2.1 Introduction

Water is a very important resource for various aspects of life on Earth and is always scarce for many reasons In areas such as the CRB in northern Vietnam, where the tropical climate differentiates into rainy, and dry seasons quite clearly, and is strongly influenced by climate change, water is a scarce commodity Besides, the scarcity is worsened by the increasing demand for water due to demographic pressure, rate of economic development, urbanization, and water pollution [3,29,41,42] Therefore, sustainable water management should be a prerequisite for the CRB, understood as ensuring the optimum, and wise use of water resources without compromising the needs of future generations

Catchment hydrological models are hypotheses of the dynamic water balance at the catchment scale and have become useful tools in water resources planning and management, providing a capability to analyze the quantity and quality of streamflow from routinely measured climate data and catchment characteristics (e.g land use, soils and slope) The products derived from models provide useful information for reservoir system operations, water distribution systems, groundwater development and protection, surface water and groundwater conjunctive use management, water use and a range of water resources activities [43]

In this chapter, we describe the modeling approaches in general catchment hydrology, criteria and the selection of a suitable hydrological model for use in the data-limited the CRB of Vietnam was implemented Finally, the importance of hydrologic models, setup, calibration and uncertainty analysis is considered by taking the SWAT

as a typical example

2.2 Hydrological cycle and review of Vietnam’s water resources

2.2.1 Hydrological Cycle Review

The Water cycle, also known as the Hydrologic cycle, is a fundamental concept in hydrology and is among a number of cycles operating in nature, such as the nitrogen cycle, the carbon cycle, and other biogeochemical cycles The National Research Council (NRC-USDA, 1982) defines the hydrologic cycle as “the pathway of water as

it moves in its various phases to the atmosphere-the Earth-the land-the ocean, and back

to the atmosphere” Water is present in the cycle in three states: solid, liquid, gas, and it

is continuous, which determines whether the process has a beginning or an end It is

necessary to study the hydrologic cycle because water is an important input in many economic activities and is essential for the survival of life on Earth A description of the water cycle is shown in Figure 2.1

Figure 2.1 Pitorial representation of hyrdologic cycle (Source: http://water.usgs.goc,

accessed on 25/03/2022) Shiklomanov [44] called the exchange of water among the oceans, land, and the atmosphere “the turnover.” There are three major subsystems of the hydrologic cycle, and that are readily identified The oceans are the largest reservoir and source of water, while the land is the main user of the water, and the atmosphere functions as a storage, circulation, and distribution of moist water Water availability at a particular place changes with time because of changes in supply and consumption Water leaves land area through ET, streamflow, interflow, and groundwater flow ET and precipitation are the processes that take place in the vertical plane, while streamflow, interflow, and groundwater flow occur mostly in the horizontal plane In this endless circulation of water, the glaciers, and snowpacks are replenished, the quantity of river water is replenished, and its quality is restored From the point of view of the utilization of water, the land phase of the hydrologic cycle is the most important

Global water use by different sectors is mainly by agriculture (which uses up to 70% of the available water), followed by industrial consumption (19–20%) and direct human consumption (10–11%) [45] In Vietnam, about 55% of water is used for agriculture [3,46] Similarly, industrial water use (including mining, power generation, and other industrial activities) accounts for about 7 to 10% Use in human activities in rural and urban areas constitutes 22–27% of water usage Of the total freshwater use in

Vietnam, 57% comes from surface water (rivers, dams, lakes, etc.), 29% from groundwater, and the remaining 14% from reuse of return flow [47]

Precipitation is the major source of runoff and is also the most important contributing factor to the variation of the scarcity of water resources in Vietnam The mean annual precipitation ranges between 1.300 and 1.500 mm, which is higher than the worldwide average [48] Of this amount, about 25% will be converted to runoff, more than 30% to groundwater recharge, and almost half will be lost as evapotranspiration [48] MORNE (2005) estimated up to 45-50% of precipitation would be lost as evapotranspiration every year in Vietnam Numerous sources (e.g [33,47,49]) report that in this century, Vietnam faces a water supply crisis not only due to the more complex change in rainfall than in the past, but also to climate change, an expanding economy, and water pollution, while the growing population also puts pressure on freshwater resources

2.2.2 Catchment Hydrology

A catchment (or watershed) is a hydrological unit that has been characterized and utilized for natural resource planning and management as a physical, biological, socioeconomic, and political entity [50] Simply put, it is considered a geographical area through which water flows over the surface of the land and is concentrated in a body of water (stream, river, lake, or ocean) Environmental studies that are affected by the movement of water along the land surface, such as environmental pollution from point and non-point sources, soil degradation, and ecosystem functioning as a whole, should

be based on a catchment approach (NRC, 1999) This is because the surface and surface water flow in the catchment eventually pass through the same common outlet

sub-As such, any downstream environmental, economic, and social impacts are also a result related to upstream responses We need to consider the impacts on the whole basin because every upstream process ends up downstream In other words, all the physical, chemical, and biological processes in a catchment are highly integrated (NRC, 1999)

Vietnam’s water resources policy, law, and strategies are based on the approaches

of integrated catchment management [46] Accordingly, two river basin management organizations were established for the two Mekong river systems and the Sesan – Srepok river basin (IMHEN, 2020) At the 4th meeting, the project members signed the

"Agreement on cooperation in the protection and sustainable exploitation in the Cau river basin" The major role of these catchment management agencies is to develop catchment management strategies that are intended to provide integrated planning, rules,

and regulations for sustainably managing water resources In general, a basin is a single area suitable for the process of research, use, and management planning of its land, water, and ecological resources.

2.3 Catchment Hydrologic Modelling

2.3.1 Importance of Hydrologic Models

The hydrological cycle refers to the continuous circulation of water in the relationship between the Earth and the atmosphere While the general theory of these processes is quite easy to grasp, they are not easily understood and not quantified in detail [2] To solve this problem, abstraction was required and various types of hydrological models were created

Hydrological models are simplified systems that characterize real hydrologic processes Thus, a system can be considered as a collection of interacting or interrelated components forming an overall system The structure of the model and the implementation process allows the users to manipulate the variables/parameters of the system easily and help in understanding the interactions between variables that make up complex systems [51] Hydrological modeling is considered a bridge between theoretical knowledge and practice or the real world [52] As a result, hydrologic models are valuable instruments for studying hydrologic processes at the catchment, regional, and global sizes

It is impossible in practice to measure everything that we want to know about catchments due to various reasons such as high catchment heterogeneity and limitations

in measurement methods, and the fact that the methods are laborious, time-consuming, and costly to implement Due to such limitations and the need to extrapolate both spatial and temporal information on catchments, hydrologic models have prime importance Catchment hydrologic models assist in gaining a better understanding of important hydrologic processes and of how changes in the catchment affect these processes [2] Catchment hydrologic models also provide hydrologic data that assist in the prediction

of potential future impacts of land use and climate change on water resources [2,42] These will again assist us during important decisions regarding catchment hydrology including but not limited to water-table management, wetland restoration, irrigation water management, streamflow restoration, water quality evaluation, and flood forecasting and management

2.3.2 Hydrologic Modeling

In general, representing a complex system in a way that is simplified for the user

is the task of the models There is a wide variety of models to represent the complex hydrologic dynamics of the earth system Various hydrologic models can be classified

as categories as described by Singh [53]

- Empirical models, often known as black-box models, take a statistical look at the system's input-output interactions Such models often rely on the link between input and output, with parameters calibrated using observed hydrometeorological records, and hence do not enable a physical knowledge of the system's behavior The unit hydrograph model is a well-known black-box model among them Empirical models may be particularly effective in terms of the scope of calibration data since the model's mathematical skills are underpinned by an implicit grasp of the physical system Extrapolation beyond the calibration range, on the other hand, is often discouraged since the implicit knowledge may no longer be true Furthermore, many black-box models are linear, but real-world hydrological systems are non-linear, hence extrapolation may be valid Some practical issues, like predicting the impacts of land use and climate change

on hydrologic responsiveness, cannot be solved with black-box models

- Physical rules are considered in significantly simplified versions in conceptual models (also known as gray-box models) A conceptual model is a description of a hydrological system that takes into account the modeler's knowledge of the physical, chemical, and hydrologic circumstances This model may be thought of as a link between a physical-process model and a model based on a conceptual framework The Stanford watershed model, the HBV model [54], and the Sacramento Soil Moisture Accounting (SAC-SMA) model are also part of this category [55]

- Because they have a logical structure comparable to the real-world system, theoretical or physics-based models are appropriate for representing hydrological processes [2] The models make use of measurable state variables that are functions of both time and space These models assume that the subbasins are homogeneous (that is, that the regions are split into a grid net and that water flows from one grid point to another as it drains through the basin) and then compute flow contribution from individual subbasins The TOPMODEL is an example of a semi-distributed hydrologic model [56] and the Soil and Water Assessment Tool (SWAT) [57] or examples of distributed hydrologic models are the SHE [58], and the Institute of Hydrology Distributed Model (IHDM) [59]

Figure 2.2 Classification of hydrological models (Source: Singh, 1988; Xu, 2002)

2.3.3 Why we choose SWAT Applications in Hydrologic Assessment in the CRB?

Given the presence of a large number of hydrological models, a frequently asked question is: "Which model is most appropriate for a particular problem?" To answer this issue, it's preferable to advise which of the aforementioned model types is most suited

to a specific hydrological situation, based on data sources and availability Based on the objectives of the study, and the research results have been widely published especially

in Vietnam, the SWAT model was selected in this study because of its advantages:

 Ready to use for basins of different sizes and complexity This is a distributed model at the basin scale that uses daily/monthly/yearly scale time steps and performs multitasking such as water resource management, groundwater flows, water quality assessment inflows with no point source or limited data [32] The SWAT model

semi-is based on physical processes with necessary inputs for setup and operation such as data

on weather, topography, land use, etc in the basin This model has also been widely applied to analyze climate change effects on hydrological processes using future climate projections, applied on basins with different areas

 Links to GIS and multi-connectivity The SWAT has proven to be a very versatile and adaptable tool for investigating a range of hydrologic and water quality problems The availability of the model and its applicability through the development

of geographic information system (GIS) based interfaces, together with its easy linkage

to sensitivity, calibration and uncertainty analysis tools, has contributed to the popularity

of SWAT in global research Moreover, technological advancements have enabled extensive networking regarding the use of SWAT, including access to web-based

documentation, user support groups, a SWAT literature database, regional and international conferences, and targeted development workshops [57,60]

 Its presentation of the main processes describes the water balance of the basin The SWAT was developed to predict the impacts of land management practices on water resources, sediment, and agricultural chemical yields in large, complex watersheds [31,57] In addition, the use of Penman-Monteith for evapotranspiration has been recommended by FAO [61], along with other methods including Hargreaves-Samani and Priestley Taylor [62]

 SWAT has received international acceptance as a robust interdisciplinary catchment-scale modeling tool This model has been extensively used worldwide, as documented by multiple reviews [29,40,63–66], and over 4300 SWAT-related publications [40] Synthesized studies from more than 70 commonly used models show that SWAT is the most common model applied in practice in water resource management, land use and water quality as well as in ecology [65,67,68]

In summary, the user-friendliness of SWAT, its vast application capacity, and the fact that it is well-evaluated, well-promoted, and well-supported have all contributed to its favored selection in hydrometeorological investigations

2.4 The Soil and Water Assessment Tool (SWAT) Model

2.4.1 General Description of The Model

The SWAT is a hydrological and sedimentological model that was developed by

Dr Jeff Arnold and his team at the Agricultural Research Service of the United States Department of Agriculture (ARS - USDA) to analyze the impacts of land-use changes

on water runoff, sediment yield and water quality in large ungauged watersheds [32] This versatile model considers different hydrological and agronomic components and has been applied by many governmental and private institutions, as well as universities and other institutions interested in supporting decision-making in the management of water resources [30] The SWAT classification is a semi-conceptual, semi-distributed, continuous-time, process-based hydrology and water quality model that uses a daily time step and multiple hydrologic units to simulate different physical processes within the study area

This model has been deployed and used since the 90s of the last century and improved with different versions including SWAT 2012 and SWAT+ Among the most significant improvements of the model, one that we can highlight is the association with

the Geographic Information System (GIS), such as the ArcGis™ (Environmental Systems Research Institute – ESRI 1998) When using ArcGIS software, SWAT becomes ArcSWAT [31] The last update of the SWAT model (version 2012) is available for download on the official website of the model: http://swat.tamu.edu/

2.4.2 Modeling Approach and Structure

Equation 2-1 presents the hydrological cycle based on water balance in the SWAT model: [32]:

𝑆𝑊𝑡 = 𝑆𝑊0 + ∑𝑡 𝑖=1(𝑃 - 𝑄𝑠 - 𝐸𝑇 - 𝑊𝑠 - 𝑄𝑔𝑤) (2-1) where: SWt is the final soil water content measured at time t (mm) and SW0 is the initial soil water content measured in time t (mm); t is time (days), P is precipitation obtained

in time t (mm), Qs is a surface runoff value in time t (mm), ET is the actual evapotranspiration in time t (mm), Ws is percolation determined in time t (mm) and Qgw is the baseflow measured in time t (mm) [69]

Figure 2.3 Components of the hydrologic balance of the SWAT model (Adapted from

Arnold et al., 1998) The SWAT-Arcgis interface is used to partition a catchment into several sub-catchments using an inputted digital elevation model (DEM) This process is done automatically allowing the user to select different limits for delimiting the sub-catchment area [31] The guiding processes are presented in the documents of the author team developing the model [57]

The definition of HRUs was continued after the catchment delineation process was completed, and the definition of HRUs is also done in the SWAT2012 interface Spatial