Nat Hazards (2014) 72:701–710 DOI 10.1007/s11069-013-1030-2 ORIGINAL PAPER
Risk early warning of maize drought disaster in Northwestern Liaoning Province, China Qi Zhang • Jiquan Zhang • Chunyi Wang • Liang Cui Denghua Yan
•
Received: 1 December 2013 / Accepted: 31 December 2013 / Published online: 14 January 2014 Ó Springer Science+Business Media Dordrecht 2014
Abstract This study presents a methodology of risk early warning of maize drought disaster in Northwestern Liaoning Province from the viewpoints of climatology, geography, disaster science, environmental science, and so on. The study area was disaggregated into small grid cells, which has higher resolution than counties. Based on the daily meteorological data and maize yield data from 1997 to 2005, the risk early warning model was built up for drought disaster. The early warning crisis signs were considered from exogenous warning signs and endogenous warning signs. The probability of drought was taken as endogenous warnings sign, which was calculated by logistic regression model. Beside precipitation, wind speed and temperature were taken into consideration when assessing the drought. The optimal partition method was used to define the threshold of each warning grade. Take the year of 2009 as an example, this risk early warning model performed well in warning drought disasters of each maize-growing stage. Results obtained from the early warning model can guide the government to take emergency action to reduce the losses. Keywords Risk early warning Drought disaster Maize Northwestern Liaoning Province
Q. Zhang J. Zhang (&) School of Environment, Northeast Normal University, Changchun 130024, People’s Republic of China e-mail:
[email protected] C. Wang Chinese Academy of Meteorological Sciences, Beijing 100081, People’s Republic of China L. Cui Sina-Canada Resources and Environmental Research Academy, North China Electric Power University, Beijing 102206, People’s Republic of China D. Yan China Institute of Water Resources and Hydropower Research, Beijing 100038, People’s Republic of China
123
702
Nat Hazards (2014) 72:701–710
1 Introduction China is a great agricultural country with large population. Agro-meteorological disasters, such as drought and waterlogging, are the main restriction factors for agricultural production (Li et al. 2000). Maize is one of the main grain crops in China. Northeastern China is the world famous golden maize belt. The weather conditions are suitable for maize growth, but the frequently agro-meteorological disasters fluctuates the maize yield greatly. Drought brings larger proportion of yield losses than other agro-meteorological disasters according to statistical records. Northwestern Liaoning Province is located in the Songliao Plain, which is one of the major maize-growing regions in China. Drought is the major agro-meteorological disaster in this region, which accounts for more than 60 % of the total agro-meteorological disasters. Drought disasters happened much more frequently with widely range under global warming. Take the year of 2009 as an example, the drought lasted from spring to autumn, and it almost caused total crop failure. So it is important to study the potential maize yield losses caused by drought and make risk early warning for the purpose of maximumly reducing disaster losses (Zhang 2004; Wilhite et al. 2007). It also could help to adjust the medium and long-term distribution of agricultural activities so as to adapt to environmental change. The natural disaster risk assessment is defined as the assessment on both the probability of natural disaster occurrence and the degree of damage caused by natural disasters (Zhang et al. 2004). Risk early warning is the best way to reduce the yield losses caused by maize drought disaster. Risk early warning is first used in the field of economy, and now, it is widely used in finance, environment, socio-political emergencies, industrial hazards, etc. (Bian et al. 2003). It is an important process to carry out risk management. For drought disaster risk early warning, most researches focus on the meteorology drought and soil drought, and there are seldom researches aimed at drought disaster specially. The start point of drought disaster early warning research is to recognize and monitor drought. Researchers mainly focused on weather factors in drought monitoring. They use mathematical statistics to construct drought identification indexes, such as standardized precipitation index (SPI) (McKee et al. 1993), palmer drought severity index (PDSI) (Palmer 1965), and effective drought index (EDI) (Byun and Wilhite 1999). Palmer made PDSI based on soil water balance and evaluate agricultural drought from the view of the soil, and it is the foundation of agricultural drought research. Based on PDSI, he took into account water requirement to get the Palmer Crop Moisture Index (CMI). SPI is an index that precipitation is the only input data. SPI can be calculated at different time scales and is sensitive to drought. It is widely used in China (Zhang et al. 2013). Steven and Timothy (2003) compared these drought indices, and for different crops and areas, the most suitable index is different. In this study, we take Northwestern Liaoning Province as the study area. Based on the theories of natural disaster risk and disaster risk early warning, we build the risk early warning model of maize drought disaster in Liaoning Province. The grid GIS technology disaggregates northwestern Liaoning into small cells, and it has higher spatial resolution than administrative units. Finally, we got the risk early warning model by using logistic regression and OPM technology, which helps to make timely waning and decision and guiding the rescue activities to reduce potential yield losses.
123
Nat Hazards (2014) 72:701–710
703
Fig. 1 Location of Northwestern Liaoning Province, China
2 Data and methods 2.1 Study area Liaoning Province located in northeastern China, between 458580 –428590 N and 119°160 –123°540 E (Fig. 1). Northwestern Liaoning covers an area of 6.8 9 104 km2, accounting for 47 % of the Liaoning Province. Maize is the most important grain crop in this area. This region is in the semi-arid continental and monsoon-controlled climatic zone with hot rainy summers and cold dry winters. Annual mean temperature is 7.2–8.3 °C, and the frost-free period is 135–150 days. Annual sunshine is 2,823–2,944 h, and annual total solar radiation is 5,719–6,050 MJ/m2. Annual precipitation is 450–700 mm, which concentrates in summer due to monsoon climate. 2.2 Theoretical basis 2.2.1 Natural disaster risk formation Natural disaster risk (I) refers to the potential human and socio-economic losses caused by natural variations. The potential adverse effects of natural disasters are considered as the products of hazard (H), exposure (E), vulnerability (V), and the capability of emergency response and recovery (R) (Fig. 2) (Zhang et al. 2006). The mathematical formula is as I = H \ E \ V \ R. 2.2.2 Theory of disaster risk early warning Disaster risk early warning is the vital component in risk assessment and management. As the disasters could bring about great losses to humanity and economy, it is important to warn and predict disasters so as to maximumly reduce the losses. The analysis processes of the early warning are as follows: determine warning condition ? find crisis sign ? analysis warning sign ? predict warning degree ? decision analysis (Yin et al. 1999). Warning condition is the target of early warning study, and it is the different conditions of warning. In this study, it is the probability of maize drought disaster. Crisis origins are the
123
704
Nat Hazards (2014) 72:701–710
Fig. 2 Formation principle of natural disaster risk
Fig. 3 Sketch map of early warning of disaster risk Endogenous waning sign
Crisis sign
Warning condition
Crisis sign
Exogenous warning sign
sources of the warning condition, and they are the factors that affect the happening of maize drought disaster and the socio-economic losses in this study. Warning signs are the signs of warning conditions, which are the comprehensive reflection of the crisis origins. The quantitative change in crisis origin is the warning sign. One crisis origin change may affect other crisis origins, so they presented as warning sign. It is essentially to consider the comprehensive affection of the disaster causing factors and the hazard-bearing body. So the warning sign could be divided into endogenous warning sign and exogenous warning sign. We summarized the theory of disaster risk early warning as Fig. 3. 2.3 Methods In this research, logistic regression method was used to build the predictable relationship between drought and the crisis signs. The basic function is as follows: P ¼ Pm=t ¼
expðb0 þ b1 x1 þ b2 x3 þ þ bk xk Þ 1 þ expðb0 þ b1 x1 þ b2 x3 þ þ bk xk Þ
ð1Þ
where x is the crisis sign; P is the severity of drought; b is the coefficient. First, we need to test samples to get the values of b. Then, we can use this function to predict drought by inputting crisis sings. In this research, the maize-growing season was divided into five growing stages. In each growing stage, a specific logistic regression model was built.
123
Nat Hazards (2014) 72:701–710 Table 1 SPI values and drought and flood levels
705
SPI values
Drought and flood levels
C2.0
Extremely wet
1.5 to 1.99
Severely wet
1.0 to 1.49
Moderately wet
-0.99 to 0.99
Near normal
-1.0 to -1.49
Moderately dry
-1.5 to -1.99
Severely dry
B-2
Extremely dry
The understanding that a deficit of precipitation could impact groundwater, reservoir storage, soil moisture, snowpack, and stream flow led McKee et al. to develop the SPI in 1993. The SPI was designed to quantify drought at multiple time scales, and precipitation is the only required inputting. Table 1 shows the drought and wet levels detected by SPI. In this study, SPI was used to detect the happening of drought as the input of P in function 1. Fisher described optimal partition method (OPM) (also known as least-squares Fisher method) in 1958, which described an algorithm for calculating optimal partitions of onedimensional data sets. The goal of OPM described in the one-dimensional case is the minimization of the same objective sum of squares distance (SSD). A partition of the data points that minimizes SSD is called the least-squares partition. To find this partition, Fisher proves that it has to be a contiguous partition (Fisher 1958). That is, if x, y, z are three data points ordered such that x \ y \ z and x and z are in the same cluster, then y is also in this cluster. It was used to partition the threshold of warning grade in this study.
3 Results 3.1 System construction for maize drought disaster risk early warning Based on disaster risk early warning theory, it is necessary to include the direct cause of disasters, but also consider the status of hazard-bearing body when building the risk early warning model. The warning sign indicators of the early warning model were selected from endogenous warning sign and exogenous warning sign. Each warning sign indicator affected by different crisis signs (Table 2). The crisis signs are in different units, and they need to be scaled into unit-less scores by using the following linear scaling function: 0
Xij ¼ 10
Xij Ximin Ximax Ximin
ð2Þ
0
and where Xij and Xij refer to the unscaled and scaled values of indicator i in county j. Xmax i Xmin refer to the maximal value and minimum value of the indicator. i 3.2 Risk early warning index of maize drought disaster (MDEWI) The risk early warning index of maize drought disaster (MDEWI) is a comprehensive evaluation of exogenous warning sign (P) and endogenous warning sign (Dg).
123
706
Nat Hazards (2014) 72:701–710
Table 2 Indicators of risk early warning of maize drought disaster in Northwestern Liaoning Province Crisis sign
Warning sign indicator
Warning sign
Consecutive days without rain Dr (day)
Probability of maize drought (Pm/t)
Endogenous waning sign (P)
XE1 planting area of maize (ha)
Maize exposure factors (E)
XV1 maize yield per unit area (kg/ha)
Maize vulnerability factors (V)
Exogenous warning sign (Dg)
Wind speed of the growth stages Ws (m/s)
Risk early warning index of maize drought disasters (MDEWI)
Temperature of the growth stages Tm (°C) Precipitation of the growth stages Pr (mm) Relative humidity of the growth stages Hu (%) Soil relative humidity of the growth stages Su (%)
XV2 ratio of water supply and demand for maize (%) XR1 well (number) XR2 buffer of reservoir (km)
Capability of emergency response and recovery (R)
XR3 ratio of effective irrigation area (%)
MDEWI ¼ P Dg
ð3Þ
The establishment of exogenous warning sign (P) was based on logistic regression model (Cui et al. 2010). The drought was determined by the SPI index. Pm=t ¼ prob N ¼ 1=Um=t expðb0 þ b1 ðDr Þ þ b2 ðWs Þ þ b3 ðTm Þ þ b4 ðPr Þ þ b5 ðHu Þ þ b6 ðSu ÞÞ ð4Þ ¼ 1 þ expðb0 þ b1 ðDr Þ þ b2 ðWs Þ þ b3 ðTm Þ þ b4 ðPr Þ þ b5 ðHu Þ þ b6 ðSu ÞÞ Define variable N = 1 as SPI [ 0 at location m and time t, otherwise N = 0. The model we used to estimate Pm/t was a logistic regression. Um/t is the collection of the values of the explanatory. b is the coefficient. Historical data from 1997 to 2005 were used to calculate the values of b. The significance of the relative humidity and soil relative humidity is 0.037 and 0.631, which exceed the acceptable error boundary after the calculation by SPSS software. It means that these two variables have little influence on drought. The following four variables were used to build the logistic regression model. The acceptable error boundary of each variable is less than 0.05. Therefore, it has statistic meaning. ð5Þ Logit pm=t ¼ 0:707 0:046ðDr Þ 0:076ðWs Þ þ 0:029ðTm Þ 0:023ðPr Þ Endogenous warning sign (Dg) is closely related to maize exposure factors (E), maize vulnerability factors (V), and resistant ability to drought disaster (R). The formula of risk early warning for maize drought disaster (MDEWI) based on the above analysis is defined as follows: MDEWI ¼ P Dg
123
ð6aÞ
Nat Hazards (2014) 72:701–710
707
Table 3 Weights of exogenous warning signs Warning sign
Warning sign indicator
Weights
Crisis sign
Exogenous warning sign
Maize exposure factors (E)
0.3410
XE1 planting area of maize (ha)
1
Maize vulnerability factors (V)
0.4452
XV1 maize yield per unit area (kg/ha)
0.5987
XV2 ratio of water supply and demand for maize (%)
0.4013
XR1 well (number)
0.4767
Capability of emergency response and recovery (R)
P ¼ Pm=t ¼
0.2138
XR2 buffer of reservoir (km)
0.2616
XR3 ratio of effective irrigation area (%)
0.2616
expðb0 þ b1 x1 þ b2 x2 þ þ bk xk Þ 1 þ expðb0 þ b1 x1 þ b2 x2 þ þ bk xk Þ Dg ¼
Weights
EðX Þ Po ð X Þ 1 þ Rð X Þ
ð6bÞ ð6cÞ
Eð X Þ ¼ w1 XE1
ð6dÞ
V ð X Þ ¼ w1 XV1 þ w2 XV2
ð6eÞ
Rð X Þ ¼ w1 XR1 þ w2 XR2 þ w3 XR3
ð6fÞ
where Pm/t refers to the possibility of the drought. Pm/t [ [0, 1]. The higher the value is, the greater possibility the drought is. w is the weight of E(X), V(X), and R(X), and they were calculated by analytic hierarchy process (AHP) method (Table 3). 3.3 Spatial distribution of crisis origin Grid GIS technology disaggregates northwestern Liaoning into 873 grid cells. These grid cells have both the form of raster data and the attribute of vector data. Kriging spatial interpolation in the ArcGIS software was used in endogenous warning sign spatial distribution analysis. Kriging spatial interpolation uses spatial relationship between sample points and some unreadily available properties to interpolate other spatial attributes. The main factors that affect the precipitation distribution are the latitude, longitude, and topographical features. Therefore, elevation, latitude, and longitude were considered in the interpolation of precipitation. Take the precipitation in maize heading stage in 2009 as an example to interpolate (Fig. 4). Co-Kriging is used to distribute the exogenous warning signs. The main factor that affects exogenous warning sign is the population. Take the counties’ relevant economic data as the sample data, population and the number of residents as collaborative data. y = f(P0, Re), where P0 is the population, Re is the number of residents. 3.4 Threshold determination for each early warning grade In this paper, the threshold of the early warning grade is determined by the OPM. It can give a quantitative and objective classification and make a scientific result. The maizegrowing season was divided into five stages (Table 4). There are six counties in
123
708
Nat Hazards (2014) 72:701–710
Fig. 4 Interpolation of average daily precipitation in the heading stage in 2009
Table 4 Division of maize-growing stages in Northwestern Liaoning Province Growing stages
Sowing
Seedling
Jointing
Heading
Maturing
Time
Mid April– late April
Early May– mid June
Late June– mid July
Late July– early August
Mid August– mid September
Fig. 5 The plot of error function and classification number of optimal partition for the MDEWI in Northwestern Liaoning Province
northwestern Liaoning. The data in each county at five stages from 1997 to 2005 were used as samples to calculate MDEWIs. So there are 270 values of MDEWIs to determine the thresholds of warning grade by using OPM. Figure 5 is the error function plot. Small
123
Nat Hazards (2014) 72:701–710
709
Table 5 The optimal partition points and F test value of the MDEWI in Northwestern Liaoning Province Item
Variable number (n)
First partition point (T1)
Second partition point (T2)
Third partition point (T3)
F (3, n k - 1)
F 0.05 (3, n - k - 1)
MDEWI
270
0.89
3.04
8.27
1,040.48
2.66
Table 6 Threshold of MDEWI in Northwestern Liaoning Province Alarm
None
Yellow
Orange
Red
Range
0–0.89
0.89–3.04
3.04–8.27
C8.27
Fig. 6 Risk early warning of maize drought disaster at different growing stages in 2009
classification with lower function error is the principle when determining the number of classifications. Four classifications are the best choice in this research. Tables 5 and 6 show the range of each classification. 3.5 Model validation Take the year of 2009 as an example, the previous model was used to make risk early warning for maize drought disaster in northwest Liaoning. The values of MDEWIs in each grid were calculated. Then, the warning grade of each grid cell was determined according to Table 6. Figure 6 shows the continuous change in warning grade at different maize-growing stages. Orange and red warning areas expanded gradually from north to south as the growing stage move on. Northwestern Liaoning Province suffered the worst drought in
123
710
Nat Hazards (2014) 72:701–710
2009 during the past 50 years, and the drought almost caused crop failure. The Liaoning Provincial Meteorological Bureau issued red and orange warning alarm. This model showed good dynamic and spatial early warning capabilities.
4 Discussions and conclusions This study presents a new risk early warning framework for maize drought disaster. The early warning crisis signs were considered from exogenous warning sign and endogenous warning sign. It is very comprehensive. Logistic regression model was used to calculate the endogenous warnings sign, the probability of drought. Beside precipitation, wind speed and temperature were taken into consideration when assessing the drought. Take the year of 2009 as example, this risk early warning model performed well in warning drought disaster risk in each maize-growing stage. The early warning results can guide the government to take emergency action to reduce the losses. Nevertheless, this study has some deficiencies. Such as, the data we can get are limited to make the model work much better, because of the multiple data input it is hard to assess the model uncertainties. Acknowledgments This study is supported by the National Key Technology R&D Program of China under Grant No. 2011BAD32B00-04, the National Grand Fundamental Research 973 Program of China under Grant No. 2010CB951102, the National Natural Science Foundation of China under Grant Nos. 41071326, 40871236, 41201550, and 41371495, and the National Scientific Research Special Project of Public sectors (Agriculture) of China under Grant No. 200903041.
References Bian J, Lin N, Tang J (2003) Precaution theory and study on alkaline desertification of soil. J Agro-Environ Sci 22(1):29–36 Byun HR, Wilhite DA (1999) Objective quantification of drought severity and duration. J Clim 12:2747–2756 Cui L, Zhang JQ, Liu XP et al (2010) Logistic regression-based prairie fire hazard prediction in case of Hulunbuir grassland. J Saf Environ 10(1):173–177 Fisher WD (1958) On grouping for maximum homogeneity. J Am Stat As 53:789–798 Li KR, Chen YF, Huang CY (2000) The impacts of drought in China:recent experiences. Routledge Publishers, UK, pp 331–347 McKee TB, Doesken NJ, Kleist J (1993) The relationship of drought frequency and duration to time scales. Eighth conference on applied climatology, Anaheim, California, pp 179–184 Palmer WC (1965) Meteorological drought. Weather Bureau, Research Paper No. 45 Steven MQ, Timothy NP (2003) An evaluation of agricultural drought indices for the Canadian prairies. Agric For Meteorol 118:49–62 Wilhite DA, Svoboda M, Hayes MJ (2007) Understanding the complex impacts of drought: a key to enhancing drought mitigation and preparedness. Water Resour Manag 21:763–774 Yin CP, Chen JX, Lu MZ (1999) Establishment of natural resources development and utilization early warning system. Ecol Econ 5:23–26 Zhang JQ (2004) Risk assessment of drought disaster in the maize-growing region of Songliao plain, China. Agric Ecosyst Environ 102(2):133–153 Zhang JQ, Okada N, Tatano H, Hayakawa S (2004) Damage evaluation and regionalization of agrometeorological disasters in the maize-growing region of Songliao plain, China: case study area of Lishu county of Jilin province. Nat Hazards 31(1):209–232 Zhang JQ, Okada N, Tatano H (2006) Integrated natural disaster risk management: comprehensive and integrated model and Chinese strategy choice. J Nat Disaster 15(1):76–91 Zhang Q, Zhang JQ, Yan DH, Wang YF (2013) Extreme precipitation events identified using detrended fluctuation analysis (DFA) in Anhui, China. Theor Appl Climatol. doi:10.1007/s00704-013-0986-x
123