Tải bản đầy đủ

Xây dựng mô hình cảnh báo sớm h n khí tượng phục vụ công tác khai thác và quản lý tài nguyên nước vùng duyên hải miền trung tt tiếng anh

MINISTRY OF EDUCATION
AND TRAINING

MINISTRY OF AGRICULTURE
AND RURAL DEVELOPMENT

THUYLOI UNIVERSITY

NGUYEN THAI HA

ESTABLISHING THE EARLY WARNING MODEL OF
METEOROLOGICAL DROUGHT IN WATER RESOURCES
EXPLOITATION AND MANAGEMENT FOR THE CENTRAL
COASTAL REGION

Specialization: Water Resources Engineering
Code No: 9 58 02 12

SUMMARY OF DOCTORAL DISSERTATION

HANOI, 2019



This dissertation was completed at Thuyloi University

Supervisor 1: Assoc. Prof . Dr. Nguyen Dang Tinh
Supervisor 2: Prof. Dr. Nguyen Van Tinh

Reviewer 1:
Reviewer 2:
Reviewer 3:

The dissertation will be defended at the Council of dissertation evaluation, at:
Thuyloi University
175 Tay Son str., Dong Da, Hanoi
At
hour
date month
year

The dissertation can be read at:
- National Library
- Thuyloi University’s Library


INTRODUCTION
1. Urgency of the thesis
Drought is a natural phenomenon. It occurs due to rain shortage, high evaporation
rate, and overexploitation of water resources. Drought can happen in every part
of the world and every type of climate zone, and it is quite common in Vietnam,
especially the Central Coastal Region (CCR). The severity of a drought period
not only depends on the duration, intensity and space, but also the water usage
for crops and living activities.
Currently in Vietnam there have been many researches on drought. However,
those researches still have some limits such as using an unsystematic method of
research or no wide application in practice. In order to help managers, policy
planners and the local people be proactive in exploiting and managing water
resources, researches on drought are both important and necessary. For that
reason, the project “Establishing the early warning model of meteorological
drought in Water Resources Exploitation and Management for the Central
Coastal Region” was proposed for research.
2. Research objective
Assess drought development, build an early warning model of meteorological
drought for the purpose of exploitation and management of water resources in
the Central Coastal region.
3. Research subjects and scope
Research subjects: the meteorology, characteristic, intensity, trend and
distribution of drought. Scope of research: Central Coastal region.
4. Research content of the thesis
Analyze and assess drought development in the researched area; analyze drought
development in the researched area when ENSO is occurring; correlatively assess
drought of the researched area with Sea Surface Temperature (SST) and Southern
Oscillation Index (SOI) to choose input variables for the drought forecasting
model; Build a structure for the meteorological drought forecasting model using
statistical method (Adaptive Neuro-Fuzzy Inference System, ANFIS) and select
a suitable model for the researched area; Build an early warning model of
meteorological drought for the researched area using drought index forecasting
map.
1


5. Research method
Statistical analysis method; inheritance method; Non-Contiguous Drought Area
(NCDA) method; correlation analysis method; meta-analysis method; and
mathematical model method (ANFIS model).
6. Scientific and practical significance
This thesis will contribute to the scientific basis in using and selecting drought
index, building drought maps using the NCDA method and drought forecasting
method. The result of the thesis can be applied to build an early warning model
of meteorological drought for the researched area as well as others. By doing so
it helps managers, policy planners and the local people be proactive in exploiting
and managing water resources.
7. Thesis structure
The content of this thesis includes foreword, conclusion and 3 chapters:
Chapter 1: Summary of current drought researches and forecast
Chapter 2: Scientific basis and meteorological drought forecasting method for
the Central Coastal region,
Chapter 3: Building an early warning model of meteorological drought for the
Central Coastal region.
CHAPTER 1 SUMMARY OF CURRENT DROUGHT RESEARCHES
AND FORECAST

1.1 Definition of drought
1.1.1 Definition and classification of drought.
Drought is a natural phenomenon. It can occur for one or many reasons, such as
rain shortage, high evaporation rate and overexploitation of water resources.
Drought appears in every part of the world and every type of climate, with
significant variation in characteristics from one region to another. According to
World Meteorological Organization (WMO). Drought is classified into 4 type:
(1) meteorological drought, (2) Agricultural drought, (3) Hydrological drought
and (4) Socioeconomic drought.
1.1.2 Drought indices
Drought index is usually a unique number representing the general drought
condition at the time of measurement. Drought index most suitable for the
2


researched area and research purpose is chosen. Below are some drought indices
commonly used around the world:
 Meteorological drought indices: Standardized Precipitation Index (SPI);
Sazonov Index (SaI); Standardized Precipitation-Evapotranspiration Index
(SPEI). Of the above meteorological drought indices, Vietnam mostly uses
SPI and SaI, whereas SPEI has only been proposed on 2010 and has yet to be
widely used.
 Agricultural drought indices: Root-Weighted Soil Moisture Index (RSMI),
Soil Moisture Anomaly Percentage Index (SMAPI); Palmer Drought Severity
Index (PDSI); Standardized Soil Moisture Index (SSI); Soil Moisture Index
(SMI). Of the above agricultural drought indices, PDSI is most widely used
both globally and in Vietnam
 Hydrological drought indices: Surface Water Supply Index (SWSI).
 Socioeconomic drought indices: Social Water Scarcity Index (SWSI).
1.1.3 Characteristics of drought
Drought periods usually differentiate from each other based on three
characteristics: intensity, duration and outspread.
1.2 Drought situation and researches around the world
1.2.1 Drought situation around the world
In recent decades, drought has been occurring around the world, causing
significant damages to the economy and negatively affecting people’s lives and
the ecosystem. Every year approximately 21 million hectares of land lose
economic productivity due to drought. In the last quarter of the century, the
number of people facing risks from drought on dry lands has increased by 80%.
More than 1/3 of the world’s land has been dried up, yet 17.7% of the world
population still lives on them. The damages caused by drought around the world
are detrimental to both people and properties.
1.2.2 Drought researches around the world
There have been many drought researches around the world. Due to the
complexity of this phenomenon, however, there has yet to be a common method
used for drought researches. In detecting, identifying, monitoring and warning
drought, researchers usually use drought indices as their main tool. Research
results show that no index has a clear advantage over others in all situations.
3


Therefore, choosing which drought index to apply depends on the specific
conditions of each area as well as its already existing monitoring data.
Currently the task of forecasting and warning are being executed using two main
method: (1) directly forecasting drought index using traditional statistical
forecasting model. This method is based on the correlation between drought
index and large scale circulation factors as well as characteristics of ENSO, etc.
Many researches have proved that the characteristics of ENSO are important
factors in detecting drought and can be used for forecasting; (2) the second
method is forecasting based on the forecasting results of climate model and
hydrological model. This method directly related to the ability to forecast climate
and hydrologic conditions which provides physical features of drought, such as
rain, temperature, flow and soil moisture level.
1.3 Drought situation and researches in Vietnam
1.3.1 Drought situation in Vietnam
Drought development in Vietnam is becoming increasingly complex. From
1985-2016, Vietnam suffered multiple droughts such as: severe drought in the
Central region and Southern Delta in 1992; Harsh drought in multiple provinces
of Central Vietnam during the winter-spring crops of 1994-1995; widespread
drought in 2002 and 2004-2005, especially in Northern Central, Southern Central
and Highland regions; water shortage, weak flow and historically low water level
of river systems throughout the country in 2009-2010 causing water shortage for
agriculture production; the severe widespread drought during the winter-spring
crops of 1997-1998 caused, when only counting agricultural properties, 5000
billion Vietnam dong in damages; and in 2014-2016, the El Nino phenomenon
caused drought and salinization. Drought has threaten both agricultural
production and human livelihood, with approximately 5.572 billion Vietnamese
dong total damage.
1.3.2 Drought researches in Vietnam
Drought researches in Vietnam in recent years mainly focused on 2 issues: (1)
General research on drought and its effects on livelihood, economy and society;
(2) Solution, prevention and damage mitigation against drought. In 2001, Nguyen
Duc Hau performed a research on detecting and applying drought index to assess
the influences of ENSO on drought and creating drought forecasting equations
for 7 climate zones in Vietnam. In 2007, Nguyen Van Thang managed to assess
4


the drought level, selected suitable drought indices and created a forecasting and
early warning technology for climate zones in Vietnam using hydro –
meteorological data and remote sensing materials to serve socioeconomic
development. In 2015, Nguyen Van Thang managed to create a suitable a drought
criteria set for monitoring and warning; built technologies and meta-statistical
meteorological drought forecasting procedures for the entire country using SPI;
built technologies and procedures for applying warning products of 8 different
models around the world in drought warning in Vietnam with up to 6 month in
advance.
For specific areas, Nguyen Trong Hieu (2000) and Nguyen Van Cu (2001)
determined the drought criteria, assessed the effects of drought, figured out
causes of desertification and suggested ways to prevent and combat drought and
desertification in 4 different provinces: Quang Ngai, Binh Dinh, Ninh Thuan and
Binh Thuan. Dao Xuan Hoc (2004) used the Sezonov index to survey and assess
drought in the CCR area. In 2005, Nguyen Quang Kim researched the drought
situation at the time and established a scientific basis for drought forecasting
procedure for South Central and Central Highlands regions; programmed drought
index calculation software and meteorological drought forecasting software
using SPI. In 2008, Tran Thuc assessed the drought level as well as water
shortage level of 9 provinces in South Central and Central Highlands regions. In
2014, Nguyen Luong Bang used SPI and SPEI in researching the influences of
ENSO on meteorological drought development in Khanh Hoa river basin. The
result of that research showed that SPEI is more fitting than SPI for assessing
drought development in river basins.
Most researches in Vietnam used statistical method for their drought forecast.
Nguyen Quang Kim (2005) used the multivariate linear regression model to
forecast drought using SPI for South Central and Central Highlands regions,
forecasting factors used are SOI, SST and terrain altitude at 500mb. Nguyen Van
Thang (2015) also used the multivariate linear regression model for SPI forecast
for 7 climate zones of Vietnam. Nguyen Luong Bang (2015) used ANFIS model
to forecast drought through SPI and SPEI for Khanh Hoa province with sea
surface temperature (SST) as inputting variable. Nguyen Van Thang (2015)
managed to build technologies and procedures for applying warning products of
8 different models around the world in drought warning in Vietnam with up to 6
month in advance, using SPI.
5


1.4 Overview of researched area
The CCR includes 14 provinces from Thanh Hoa to Binh Thuan. The researched
area usually suffers from drought due to influences of El Nino, especially from
Khanh Hoa to Binh Thuan.
1.5 Conclusion of chapter 1
Through overall research and analysis of drought forecasting and warning
models around the world as well as in Vietnam, this thesis has chosen the
research contents and methods as shown in the chart below:

Figure 1.3 Research content and methods chart
CHAPTER 2 SCIENETIFIC BASIS AND METEOROLOGICAL
DROUGHT FORECASTING METHOD FOR THE CENTRAL
COASTAL REGION
2.1 Current drought situation of the researched area
According to statistical data provided by the Department of Water Resources in
35 years from 1980 to 2014, the CCR has suffered many droughts, damaging
6


hundreds of thousands hectares of land and causing severe water shortage for
millions of people. The drought land/farm land ratio is at its highest in 1993,
followed by 2010, 2005, 1998, 1985 and 1988. In reality, however, 1988 suffered
the harshest drought with 180,836 hectares of drought land and 51,130 hectares
of land completely losing its crops. The most intense drought period of the
researched area was the summer-autumn crops from the end of June to the
beginning of September, the frequency of drought happening in large scale is
once every 5-9 years in the CCR.

Figure 2.3 Drought land/farm land ratio of CCR
2.2 ENSO introduction and necessary data to collect
2.2.1 ENSO introduction
ENSO is the interaction between the atmosphere and ocean in the Pacific Ocean
(PO) that results in changing climate. El Nino is the name of the abnormal
warming in STT in east-central Equatorial Pacific. In contrast to El Nino, the
phenomenon when the SST in east-central Equatorial Pacific gets abnormally
cold is called La Nina. The Southern Oscillation (SO) is what causes the air
exchange between the Eastern and Western hemisphere. SO is determined using
the differences in sea surface air pressure between Tahiti (17.5S; 149.6W)
Southeast of Pacific and Darwin (12.4S; 130.9E) Northwest of Australia.
To monitor ENSO activities, SSTA of Equatorial Pacific is used. An El Nino
cycle is a continuous period, with average SSTA of Nino3.4 (5oN-5S, 120oW170oW) equal or higher than 0.5oC, while a La Nina cycle is a continuous period
with average SSTA in 5 month of Nino3.4 equal or smaller than -0.5oC. From
1985 to 2014, there have been 9 instances of El Nino and 9 instances of La Nina.
The change in SST in the Pacific Ocean creates ENSO activities, which leads to
7


the anomaly of air masses as well as monsoon activities and equator-tropical
disturbance changing the weather of West PO, including the researched area
2.2.2 Necessary data to collect
1. Precipitation, temperature: Collected from 27 weather stations in CCR. Data
used in the research were from January-1985 to December-2014.
2. Sea Surface Temperature Anomaly (SSTA): SSTA of Nino3.4 was taken from
http://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_
v5.php from January-1985 to December-2014.
3. The Southern Oscillation Index (SOI): SOI data was taken from
https://www.ncdc.noaa.gov/teleconnections/enso/indicators/soi/ from January1985 to December-2014.
2.3 Meteorological drought forecasting method for the researched area
Content and method of meteorological drought assessment, forecasting and
warning for the researched area are demonstrated in the chart below:

Figure 2.8 Meteorological drought content and forecasting method
demonstration chart
8


2.3.1 Selection of drought indices
SPI is used in multiple researches for assessment and warning and the results
shows that SPI is suitable in many researched areas. Therefore, in this thesis, the
author will also choose SPI, as well as SPEI in addition (this index has only been
proposed on 2010 and haven’t been researched or used widely in Vietnam) to
compare and assess the suitability of these two indices in meteorological drought
assessment and early warning in the researched area.
 SPI: Proposed by McKee and his partners on 1993, SPI is calculated
based on gamma function distribution and long-term precipitation record in
random variable form with normal distribution.
 SPEI: Proposed by Vicente-Serrano on 2010, SPEI is calculated based
on log-logistic distribution probability with the result of precipitation minus
potential evapotranspiration over time in random variable form with normal
distribution.
In this thesis, SPI and SPEI will be used in different period of time (1 month and
3 months). SPI and SPEI for the 1 month period are coded as SPI1 and PSEI1
respectively, whereas SPI and SPEI for the 3 months period are coded as SPI3
and SPEI3 respectively.
Table 2.3 Drought classification using SPI and SPEI
SPI, SPEI
≥ 2,0
1,5÷1,99
1,0 ÷ 1,49
0.50 ÷ 0,99
-0.49 ÷ 0.49

Weather condition
Extremely wet
Very wet
Moderately wet
Slightly wet
Normal

SPI, SPEI
-0.50 ÷ - 0.99
-1,0 ÷ -1,49
-1,5 ÷ -1,99
≤ -2,0

Weather condition
Slight drought
Moderate drought
Severe drought
Extreme drought

In this thesis, the author has written the calculation program code for these
indices using programming language R.
2.3.2

Analysis and assessment of drought development in the researched
area using drought indices

This thesis uses the Non-Contiguous Drought Area (NCDA) method to analyze
and assess drought development in researched area. This method divides the
researched area into grids and treat the drought condition of each grid as
independent and unrelated to each other. To apply NCDA method for the CCR,
the execution steps are as follows: (1) Divide the researched area into 3,752 5x5
grids; (2) Interpolating the precipitation and temperature of each grid using the
9


Inverse Distance Weighted method; (3) Test the results of precipitation and
temperature interpolation; (4) Calculate SPI, SPEI in 1 month and 3 months
periods for each grid; (5) Apply NCDA method for drought assessment based on
space and duration using the right drought level for area ratio (level 1, 2, 3 and 4
are slight, moderate, severe and extreme, respectively).
2.3.3 Analysis of the influences of ENSO on drought development in the
researched area
Changes in SSTA and SOI in PO create ENSO activities, which leads to the
anomaly of air masses, causing El Nino. When El Nino appears, the precipitation
of the researched area will be lowered and the chance of drought is high.
Therefore, in this thesis, the author will research the influences of ENSO
activities on drought development, both in space and duration, in the researched
area. It will be done using the 2 following methods: (1) analyze drought
development in researched area during El Nino periods: (2) determine the
correlation between SSTA, SOI and SPI, SPEI using the correlation equation (for
both the 1 month and 3 months periods). Each chain of indices will be determined
the correlation with 12 different chains of SSTA and SOI (12 chains of SSTA
and SOI are SSTA and SOI chains with 1 to 12 months latency comparing
drought indices). The higher the absolute value of the correlation coefficient
(closer to 1), the closer the linear relationship between drought indices and SSTA
and SOI.
2.3.4 Construction of a meteorological forecasting model and propose a
suitable forecasting model for the researched area
Drought assessment and forecasting can be done using drought indices. After
assessing different drought forecasting models currently in use around the world
and in Vietnam, the author decided to use the drought forecasting model with the
following components:
1. Output variables (forecasting elements)
Output variables are SPI and SPEI in 1 month and 3 months periods. The output
result of SPI and SPEI are shown through 3 factors: forecasting duration (short
or long), severity and chance of occurrence (value based on drought classification
of drought indices).
2. Input variables (forecasting factors)
10


Input variables are SSTA and SOI data with high correlation coefficient with SPI
and SPEI, which also combine with SPI and SPEI in different time stages
previously. Details about type of variables and input variables of the forecasting
model will be demonstrated in Figure 2.4 below
3. Forecasting method
The ANFIS model, proposed by J.S.R. Jang and his partners on 1997, was used
to build forecasting models for SPI and SPEI with different forecasting factors to
find a suitable forecasting model for the researched area. ANFIS is based on an
open interface system and is trained by an algorithm originated from neural
network theory and has the following structure:

Figure 2.11 Structure of ANFIS model
Output value of SPI or SPEI is calculated as follows:

w1 f1  w2 f 2

w1  w2
w ( x , y , z ) f 1  x , y , z   w 2 ( x, y , z ) f 2  x , y , z 
 1
w1 ( z, y, z ) w2 x, y, z 

SPI or SPEI = f  x, y, z  

(2-41)

With the mixed mathematical algorithm of ANFIS model, the input data is
divided into 2 sets: network training process data and testing process data.
4. Construxtion of meteorological drought forecasting models for the researched
area
 Structures of forecasting models
Structures of forecasting models (variables types, number of input variables) are
shown in detail in Figure 2.13.
Network training data is from 1985-2009. Testing data is from 2010-2014.
11


Figure 2.13 Structure of forecasting models
 Assessment of forecasting quality
The assessment of forecasting quality of the models during the network training
process and testing process is done using 3 coefficients: RSR (RMSEobservations standard deviation ratio); CORR (Correlation Coefficient); and E
(Efficiency), also called Nash Index. A forecasting model has good quality when
its CORR and E values are closer to 1.0 and RSR value is closer to 0.0. In
addition, RSR and E values must meet the criteria of WMO’s assessment
standard as shown in this table:
Table 2.4 Criteria for forecasting quality assessment
Ranking
Very good
Good
Pass
Fail

RSR
0 ≤ RSR ≤ 0.5
0.5 < RSR ≤ 0.6
0.6 < RSR ≤ 0.7
RSR > 0.7

12

E
0.75 < E ≤ 1
0.65 < E ≤ 0.75
0.50 < E ≤ 0.65
E ≤ 0.5


The meteorological drought forecasting models for the researched area are
programmed on Matlab software. The block diagram of the meteorological
drought forecasting program is shown in the following Figure:

Figure 2.14 Block diagram of forecasting program
13


2.4 Conclusion of chapter 2
(1) The CCR is one of the regions with frequent droughts. Drought is most
intense during the summer-autumn crops from the end of June to the beginning
of September. In this research, the author chose two meteorological drought
indices: SPI and SPEI to assess, forecast and warn drought for the researched
area (2) Drought development assessment method for the researched area during
the El Nino period and method of analyzing the correlation between SSTA &
SOI and SPI & SPEI were used to assess the influences of ENSO on drought
development in the CCR; (3) Meteorological forecasting method is the ANFIS
model with output variables being drought indices (SPI & SPEI) from previous
time period; (4) Five forecasting models with different variables and input
variables were created. The forecasting results of these models were compared
and assessed using 3 coefficients, which are RSR, CCRR and E.
CHAPTER 3 BUILDING AN EARLY WARNING MODEL OF
METEOROLOGICAL DROUGHT FOR THE CENTRAL COASTAL
REGION.
3.1 Drought development of the researched area based on space and
duration
3.1.1

Analysis of the testing result of the precipitation and temperature
interpolation

The IDW interpolation method is reliable enough to interpolate the precipitation
and temperature to serve drought index calculation of the researched area.
3.1.2 Drought development of the researched area based on duration
The drought development of the researched area based on duration using SPI and
SPEI are shown in the following Figures:

Figure 3.5 Drought development of the CCR according to SPI1 and SPEI1
14


Figure 3.6 Drought development of the CCR according to SPI3 and SPEI3
The drought development of CCR using the aforementioned indices show 1988,
1993, 1998, 2005, and 2010 suffering from long drought period, matching with
drought periods in reality. SPEI, however, show the drought development of the
researched area more accurately compared to SPI, and SPEI3 is more accurate
than SPEI1.
3.1.3 Drought development of researched area based on space
The Drought area ratio according to SPI, SPEI and actual agricultural production
of the researched area are shown in the following Figure:

Figure 3.11 Drought area ratio according to SPI3, SPEI3 and actual agricultural
production
Drought area ratio according to SPEI is higher than SPI in most years, with only
one year where it’s the reverse. The drought area ratio according to SPEI3 is the
highest, followed by SPI3, SPEI1 and finally SPI1. The drought frequency of
indices are shown in the following Figure:
15


Figure 3.12 Drought frequency (%) of different indices based on space
3.2 Influences of ENSO on the drought development of researched area
3.2.1 Drought development of researched area during ENSO periods
The drought development of the researched area according to SPI and SPEI
during ENSO periods are shown in the following Figures:

Figure 3.27 SPI1 and SPEI1 values during ENSO periods

Figure 3.28 SPI3 and SPEI-3 values during ENSO periods
During El Nino periods, drought always happened in researched area, but the
start-time of drought periods is later than El Nino periods. Drought level reflected
16


by SPEI3 of the researched area during El Nino was higher compared to other
indices.
3.2.2 Assessment of the correlation between SSTA, SOI and SPI, SPEI
The average correlation coefficient results between SSTA, SOI and SPI, SPEI in
all grids of the researched area are shown in table 3.4.
The correlation between SSTA and SPI, SPEI is at its highest when the SSTA
data is 3 months ahead of SPI and SPEI (3 months latency), whereas the
correlation between SOI and SPI, SPEI is at its highest when SOI data is 2 months
ahead of SPI and SPEI (2 months latency).
Table 3.4 Correlation coefficient between SSTA, SOI and SPI, SPEI
Correlation coefficient between SSTA1 and SPI1, SPEI1
Latency
1
2 (*) 3 (*) 4 (*) 5 (*)
6
7
8
9
10
(month)
SPI1
-0.14 -0.157 -0.167 -0.161 -0.146 -0.119 -0.108 -0.094 -0.089 -0.073
SPEI1
-0.14 -0.157 -0.167 -0.161 -0.146 -0.119 -0.108 -0.094 -0.089 -0.073
Correlation coefficient between SSTA3 với SPI3, SPEI3
Latency
1
2 (*) 3 (*) 4 (*) 5 (*)
6
7
8
9
10
(month)
SPI3
-0.256 -0.269 -0.274 -0.266 -0.244 -0.212 -0.183 -0.156 -0.134 -0.118
SPEI3
-0.256 -0.269 -0.274 -0.266 -0.244 -0.212 -0.183 -0.156 -0.134 -0.118
Correlation coefficient between SOI1 and SPI1, SPEI1
Latency
1 (*) 2 (*) 3 (*) 4 (*)
5
6
7
8
9
10
(month)
SPI1
0.162 0.169 0.143 0.144 0.138 0.088 0.085 0.063 0.058 0.045
SPEI1
0.162 0.169 0.143 0.144 0.138 0.088 0.085 0.063 0.058 0.045
Correlation coefficient between SOI3 and SPI3, SPEI3
Latency
1 (*) 2 (*) 3 (*) 4 (*)
5
6
7
8
9
10
(month)
SPI3
0.278 0.303 0.292 0.266 0.238 0.213 0.17 0.124 0.096 0.079
SPEI3
0.278 0.303 0.292 0.266 0.238 0.213 0.17 0.124 0.096 0.079

11

12

-0.062 -0.064
-0.062 -0.064

11

12

-0.109 -0.102
-0.109 -0.102

11

12

0.042 0.037
0.042 0.037

11

12

0.071 0.06
0.071 0.06

Note: * Correlation has reliability rate higher than 90%
3.3 Construction of an early warning model of meteorological drought for
the researched area
3.3.1 Assessment of the results of drought forecasting model for researched
area
The statistical results (CORR, E and RSR) of SPI and SPEI forecasting models
are shown in the table below:
17


Table 3.5 Assessment of drought index forecasting result of models
SPI1
Network training
process
Model
CORR E
RSR
M1
0.33 0.11 0.94
M2
0.38 0.15 0.92
M3
0.56 0.31 0.83
M4
0.82 0.67 0.58
M5
1.00 1.00 0.06
SPI3
Network training
process
Model
CORR E
RSR
M1
0.53 0.28 0.85
M2
0.58 0.34 0.81
M3
0.76 0.57 0.65
M4
0.90 0.81 0.44
M5
1.00 0.99 0.08

SPEI1
Network testing
Testing process
Testing process
process
Model
CORR
E
RSR
CORR E
RSR CORR E
RSR
0.26
0.06 0.97
M1
0.38 0.14 0.92
0.45 0.13 0.93
0.29
0.05 0.97
M2
0.46 0.22 0.89
0.35 0.09 0.96
0.36
0.12 0.94
M3
0.63 0.40 0.77
0.40 0.14 0.93
0.46
0.21 0.89
M4
0.83 0.68 0.56
0.48 0.18 0.91
0.16 -0.01 1.01
M5
1.00 1.00 0.04
0.55 0.21 0.89
SPEI3
Network training
Testing process
Testing process
process
Model
CORR
E
RSR
CORR E
RSR CORR E
RSR
0.52
0.24 0.87
M1
0.60 0.36 0.80
0.77 0.32 0.82
0.60
0.28 0.85
M2
0.64 0.41 0.77
0.78 0.36 0.80
0.42
0.12 0.94
M3
0.76 0.57 0.65
0.78 0.46 0.73
0.63
0.38 0.79
M4
0.91 0.82 0.42
0.82 0.68 0.56
0.60
0.34 0.81
M5
1.00 0.99 0.07
0.74 0.53 0.69

The SPEI forecasting model M4 has the best result, meeting WMO standard
(Table 2.4). The comparison result between forecasted and actually calculated
SPEI3 of the network training process and testing process using the M4 model
for the researched area are shown as follows:

Figure 3.32 Comparing SPEI3 forecasting results using M5 for the entire area
3.3.2 Selection of an index for meteorological drought early warning for the
researched area
According to the drought development assessment results based on space and
duration using SPI and SPEI in section 3.1, SPEI’s reflection of drought
development is more suitable for the actual drought situation of the researched
area than other indices. In addition, the research result in section 3.3.1 shows that
18


drought forecasting using SPEI3 also gave the best results. Therefore, in this
thesis, the author chose SPEI3 as the index for meteorological drought
forecasting.
Table 3.6 Drought levels and meteorological drought early warning levels using
SPEI3
SPEI3
≥ - 0.49

Climate
condition
No drought

-0.50 ÷ - 0.99

Slight
drought

-1.0 ÷ -1.49

Moderate
drought

-1.5 ÷ -1.99

Severe
drought

≥ -2.0

Extreme
drought

Warning level
Level 1: Rain shortage warning. Make sure to conserve water if
SPEI3 at different time of the month and/or recent months are
also at this level.
Level 2: Moderate drought warning. Make sure to conserve
water and prepare prevention works. Be extra careful if SPEI3
at different time of the month and/or recent moths are also at
this level.
Level 3: Severe drought warning. Water conservation and usage
limit solutions are required, especially if SPEI3 at different time
of the month and/or recent months are also at this level. Check
prevention procedures. If water supplies (reservoir, surface
runoff, ground water) get low, stopping water supply for least
important households is allowed.
Level 4: Extreme drought warning. Depending on the condition
of surface water supply, stop water supply for least important
households and/or apply water usage limit regulation. May need
to start relief campaigns.

3.3.3 Selection of a drought forecasting model for the researched area
In this research, the author propose forecasting models for the next 6 months of
monitoring period in the researched area using 2 types of forecasting (1)
forecasting the first 2 months (January and February of 2015) and (2) forecasting
the next 4 months (March to June of 2015). The parameters for the 2
aforementioned forecast are shown in the following table:
Table 3.7 Forecasting models’ parameters for the next 6 months
Model
Input parameter
Short drought
SOI3 (10/2014), SOI3 (11/2014), SSTA3 (10/2014),
M4
SSTA3 (11/2014), SPEI3 (12/2014)
SOI3 (11/2014), SOI3 (12/2014), SSTA3 (11/2014),
M4
SSTA3 (12/2014), SPEI3 (1/2015)
Long drought

19

Output
SPEI3
(1/2015)
SPEI3
(2/2015)


M4
M4
M4
M4

SOI3 (12/2014), SOI3 (1/2015), SSTA3 (12/2014),
SSTA3 (1/2015), SPEI3 (2/2015)
SOI3 (1/2015), SOI3 (2/2015), SSTA3 (1/2015),
SSTA3 (2/2015), SPEI3 (3/2015)
SOI3 (2/2015), SOI3 (3/2015), SSTA3 (2/2015),
SSTA3 (3/2015), SPEI3 (4/2015)
SOI3 (3/2015), SOI3 (4/2015), SSTA3 (3/2015),
SSTA3 (4/2015), SPEI3 (5/2015)

SPEI3
(3/2015)
SPEI3
(4/2015)
SPEI3
(5/2015)
SPEI3
(6/2015)

3.3.4 Construction of meteorological drought early warning models for the
researched area
From the results and content in the above sections of the research, the author
propose the following early warning model of meteorological drought for the
researched area:

Figure 3.42 Early warning model of meteorological drought for the researched
area
This early warning model of meteorological drought includes 3 main blocks: (1)
Monitoring and collecting data; (2) Calculating and handling data; and (3) Meta20


analysis of monitoring data and forecasting results to create maps and assessment
board on drought situation and drought development forecast.
3.3.5 Maps and data on meteorological drought early warning for the
researched area
1. Warning maps for the first 2 months (January and February of 2015)
January/2015

January/2015

Figure 3.43 Forecasting and warning maps according to SPEI3 for January and
February of 2015
2. Warning maps for the next 4 months (March to June of 2015)
March/2015

April/2015

March/2015

April/2015

Figure 3.44 Forecasting and warning maps according to SPEI3 for March,
April, May and June of 2015
21


3.4 Conclusion of chapter 3
(1) SPEI’s reflection of drought development was more suitable for the actual
drought situation of the researched area than SPI. Durations, scales, intensities
and frequencies of drought in the CCR according to SPEI were also higher then
SPI.
(2) When El Nino was happening, drought would appear in the researched area
and lasted for many months, however its start-times were usually later than El
Nino’s start-times. SSTA and SOI chains with big correlation to SPI and SPEI
chains would be chosen as input variables for meteorological drought forecasting
models.
(3) SPEI3 forecasting result using M4 had the highest quality and met the criteria
of forecasting quality assessment standard. Drought index chosen for drought
level assessment and early warning model of meteorological drought was SPEI3.
2 months short drought forecasting model and 4 following months long drought
forecasting model were chosen.
(4) An early warning model of meteorological drought for the researched area
was built with 3 main blocks: monitoring and collecting data; calculating and
handling data; and Meta-analysis of monitoring data and forecasting results. At
the same time meteorological drought early warning maps for the researched area
using SPEI3 from January to June of 2015 was created.
CONCLUSION AND RECOMMENDATION
1. Achieved results of the thesis
Drought can occur on every parts of the world and in all climate zone, including
CCR. In order to have affective solutions for exploitation and management of
water resources, the meteorological drought early warning task is essential.
Therefore, this thesis has solved and achieved the following results:
(1) From 1985-2014, the CCR suffered many droughts, 5 of them were extreme
droughts. Drought occurred during all 3 crops and caused damages to millions
hectares of land. The most intense period was the summer-spring crop from the

22


beginning of June to the end of September, and occurred in large scale once every
5 to 9 years.
(2) Drought developments according to SPI and SPEI show that 1988, 1993,
1998, 2005 and 2010 all suffered drought that lasted for many months, matching
with the actual drought periods which lasted for a long time during all 3 crops.
However, SPEI’s reflection of drought development was more suitable for the
actual drought situation of the researched area than other indices.
(3) The influences level of El Nino to drought development of the researched
area was quite high. Whenever El Nino occurred a drought would appeared in
the researched area and lasted for many months. However its start-times were
usually later than El Nino’s start-times.
(4) When the SST of Nino3.4 increased (positive SSTA) and SOI in negative,
they both affected the drought development of the researched area, especially
when the SST of El Nino3.4 increased (SSTA>= 0.5oC, leading to El Nino), there
were significant chances a drought would occurred in the researched area but
usually 2-3 months later.
(5) The SPEI3 forecast result with SSTA3 and SOI3 chains 2-3 month earlier and
SPEI3 1 month earlier (M4 model) as input variables gave the most reliable
result, meeting the criteria of forecasting quality assessment standard.
(6) Meteorological drought early warning maps according to SPEI3 from January
to June of 2015 were built using short drought forecasting model (1-2 months)
and long drought forecasting model (3-6 months). They were created with ANFIS
network with M4 structure as demonstrated in Table 3.6.
(7) An early warning model of meteorological drought for the researched
area was built with 3 main blocks: monitoring and collecting data; calculating
and handling data; and Meta-analysis of monitoring data and forecasting results.
This model is demonstrated in detail in Figure 3.42.
2. Summarizing the new conclusions of the Dissertation
(1) Determining the relationship among ENSO, Sea Surface Temperature
Anomalies (SSTA), and the Southern Oscillation Index (SOI) with
meteorological drought events in the Central Coast Region (through SPI and
23


Tài liệu bạn tìm kiếm đã sẵn sàng tải về

Tải bản đầy đủ ngay

×

×