Click
Here
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 114, D18104, doi:10.1029/2009JD011833, 2009
for
Full
Article
Impacts of climate change on historical locust outbreaks in China
Ge Yu,1 Huadong Shen,1 and Jian Liu1
Received 31 January 2009; revised 16 April 2009; accepted 4 June 2009; published 17 September 2009.
[1] To probe if greenhouse-effected climate warming strengthens severe locust outbreaks
that would cause continental-scale crop failures in China, we studied their statistical
relationships in history and examined the impacts of climate change on the long-term
locust outbreaks. According to analysis for interannual time series during the past
100 years, the most severe locust outbreak years were in the warm-dry years with
warm-dry summers and warm-wet winters in the Yellow River–Haihe River region,
northern China, and warm-wet years with warm-wet springs in the Yangtze River–Huihe
River region, southern China. Checking wavelet-analyzed variance series, these
interannual time scale synchronous changes with 2–10 periodicity years were 58–60%
among the total locust outbreak years of the past 1000 years. Locust outbreak correlation
analysis with decadal time scale temperature proxy and general circulation model–
simulated climate during the past 1000 years showed significant correlations in
warm-winter-half years and in warm-dry May–June and annual means in the northern
region (p < 0.05), where p is probability, and in warm years and warm-dry August–
September years in the southern region (p < 0.10), while these decadal time scale
synchronous changes with 20–110 periodicity years were 56–65% of the total locust
outbreak years of the past 1000 years. Historical records on ‘‘drought locust’’ were true, as
we found that the drought of spring-summer season in the northern region caused the
highest regional locust outbreaks, but it is not a sensitive factor for locust outbreaks in the
wet-humid Yangtze River region. Although warm winter condition is a key factor for
locust egg survival and preservation, it works well when winter temperatures reach
to 10° to 30°C in the northern region but is not a limiting factor for locust survival in
the warm southern region. We found that both interannual and decadal variability of
higher temperature changes have led to the highest locust outbreaks in the past 1000 years,
so as to suggest that greenhouse-effected climate warming would increase the severe
locust outbreaks in the area.
Citation: Yu, G., H. Shen, and J. Liu (2009), Impacts of climate change on historical locust outbreaks in China, J. Geophys. Res., 114,
D18104, doi:10.1029/2009JD011833.
1. Introduction
[2] The locust is one of the most serious pests that feed
upon crops, eating and gnawing the leaves and stems of
mainly Gramineae plants, one major family of popular
Chinese foods of cereal, wheat, corn, and rice. High-density
locusts often occur in summer and autumn in eastern China;
for example, one locust outbreak in summer-autumn of
1998 had a maximum density of 10,000 locust/m2 and
covered areas of 200 km2 in Hebei Province, 300 km2 in
Henan Province [Ren and Tang, 2003], and 450 km2 in
Shangdong Province [Zhang et al., 2006]. The locust outbreaks in Chinese history have greatly decreased crop
output, causing serious food shortage and famines; therefore
great attention has been paid in history and detailed records
can be found in much historical literature. The earliest
1
Nanjing Institute of Geography and Limnology, Chinese Academy of
Sciences, Nanjing, China.
Copyright 2009 by the American Geophysical Union.
0148-0227/09/2009JD011833$09.00
record can be found in The Book of Songs, a book already
in circulation in East-Zhou Dynasty (770 –476 B.C.), which
has introduced methods of burning bonfire to lure locusts to
their death and burning grass to destroy their eggs on farm
lands [Ni, 1998]. Emperor of Fu Jian (338 – 385 A.D.) sent
thousands of his soldiers to assist locals to erase locusts in
Yellow River regions in 382 A.D., which was recorded in
History as a Mirror (published during 1019 – 1086 A.D.),
the first historical annals in China. As locust outbreaks
caused serious food shortage and threatened the stabilization
of political power, likely ‘‘long years of locust outbreaks
caused the peasant uprising and the destruction of WestHan Dynasty’’ (recorded in History of East-Han Dynasty,
a book published circa 25– 220 A.D) [Ni, 1998], the first
law of locust control in China was issued in 1075 A.D. and
the second in 1182 A.D. by emperors of North-Song
Dynasty (960– 1127 A.D.) [Wu, 1951]. There have been
special books describing the locust ecology and control
methods, for example, Methods of Locust Capture issued in
1193 A.D. and Complete Methods of Locust Control published in 1857 A.D. [Zheng, 1990]. On the basis of the
D18104
1 of 11
D18104
YU ET AL.: HISTORICAL LOCUST OUTBREAKS IN CHINA
historical records, there were 796 disastrous locust outbreaks between 707 B.C. and 1935 A.D. [Chen, 1935],
during which smaller locust outbreaks occurred once every
2 – 3 years and severe outbreaks every 7 –8 years [Wu et al.,
1990].
[3] Attention has been paid to causes of locust outbreaks
since 2000 years ago. Lu [1986] has counted the times of
historical locust outbreak records with climate causes, from
a photocopy of The 25 Dynastic Histories of China [1935].
He found that there were 58 records related to temperature
change and 14 records related to precipitation change in a total
of 261 locust outbreak records since 841 B.C.; particularly,
many records directly named the pest as ‘‘drought locust’’ or
‘‘drought locust calamity.’’ The historical records reveal
that the locust outbreak years often occurred in a warm
spring year and/or a year after a warm winter, for example,
‘‘A winter season with spring temperature will lead to a
potential locust outbreak, and a spring season with summer
temperature will cause a locust disaster’’ (recorded in The
Book of Rites, circulation circa 770– 476 B.C.) [Ma, 1958],
and ‘‘locust eggs would die out in a heavy snow winter’’
recorded in Complete Collection of Pictures and Books of
Old and New Times (1650 – 1741 A.D.) [Zheng, 1990].
Modern biology can provide a scientific basis to explain
the phenomenon. The minimum temperature for locust
ovum growth is 16°C, but adult insects can tolerate up to
45°C [Wu et al., 1990], suggesting the pests are warmtemperature-loving species. Locust eggs can survive 15 days
under 10°C but only 1 day under 25°C, and the adult
insect needs an effective growth degree day of 210 day°C
[Wu et al., 1990], suggesting that the changes in winter and
annual temperatures are both responsible for locust growth.
[4] The locusts are mostly born and grow in floodplains,
lake overbanks, swamps, and coastal lowlands in eastern
China. These places are dry lands during low water level
and little rainfall and are submerged during high water level
and heavy rainfall. ‘‘Lake drying makes locust multiply and
outbreak,’’ recorded in An Agricultural Encyclopedia of
Ming Dynasty (1562– 1633 A.D.), ‘‘changes to low water
levels in ponds, lakes and rivers lead to locust growth
greatly,’’ recorded in Complete Methods of Locust Control,
and ‘‘rainstorms in spring and summer could cause locust
dying,’’ recorded in History of Song Dynasty (circa 960 –
1127 A.D.) have been recognized in Chinese history [Lu,
1986]. The records reflected the facts that locust eggs are
multiplied in wet soil but not in a fully submerged place,
and locust migration and dispersal need dry conditions, as
rainfall restrains wing vibration and flight. Modern bioecology has proven that the best soil for laying its eggs is the
soil with water content of 10– 20%, where locust eggs reach
to 200,000 – 400,000 grain/m2 [Wu et al., 1990]. As a locust
egg needs 5– 10 days to develop, a continuous 15 days of
submergence in water can drown 100% of locust eggs [Wu
et al., 1990]. Therefore, in terms of locust bioecology, the
lake/river/coast lands above water level are favorable places
for locusts to develop, while spring and summer dry
conditions lead to locust outbreaks.
[5] Studies on whether the locust outbreak variability is
consistent with climatic change have been undertaken in
China. Many studies indicated that, on the basis of the
records of the past 50 years, a warm winter year, i.e., higher
January temperature than average, a warm-summer-half
D18104
year, i.e., higher April – September temperatures than average, and a dry-summer-half year, i.e., lower April – August
precipitation than average, normally lead to locust outbreaks
in eastern China [Ma, 1958; Zhang and Chen, 1998; Kong
et al., 2003; Ren and Tang, 2003; Wang et al., 2006; Zhang
et al., 2006]. Locust outbreaks with relationship to past
climate changes were studied by Lu [1986], Zheng [1990],
and Zhang and Chen [1998], who found a basis for linkage
of locust outbreaks with warm and drought climate in
Chinese history. Zhang and Chen [1998] have also revealed
that locust outbreaks in northeast China (latitude north to
40°N) occurred in winter warm years during the middle of
the 12th, 13th, and 18th centuries. In terms of the locust
outbreaks forced by decadal time scale climate, Stige et al.
[2007] used a temperature proxy and a precipitation proxy
to correlate the past 1000-year locust outbreak series. The
precipitation proxy was reconstructed by tree ring records in
the western mountains of China, and the temperature was a
synthesis for whole China as sourced from multiple line
evidence of ice cap records, lake level records, historical
documents, etc. [Yang et al., 2002]. Their study has attempted to prove a hypothesis that locust outbreaks were in cold
and wet years during the past 1000 years [Stige et al., 2007].
[6] Regional studies in China, likely studies for regions
of provinces Liaoning, Hebei, Henan, Shandong, Anhui,
and Jiangsu [e.g., Ma, 1958; Kong et al., 2003; Ren and
Tang, 2003; Wang et al., 2006; Zhang et al., 2006], were on
interannual time scale for the past 50 years, while Stige et
al.’s [2007] study was on decadal time scale for the past
1000 years. These two types of studies on different time
scales have produced contradictory understanding of the
relationships between climate character and locust outbreaks: the former believed the highest locust outbreak
occurred in a warm year, while the later related it to cold
years.
[7] To do regional and seasonal analysis in both interannual and decadal time scales would resolve the issues. This
is based on two considerations of different regional climates
in northern and southern China.
[8] 1. The relative variability of annual precipitation is
much higher in the northern region (25 –50%) than in the
southern region (15 – 30%) [Liu et al., 1998], and the
variability is much higher in summer (200 15%) than
in winter (70 50%) [Wang et al., 2000; Zhong et al.,
2007]. Ma [1958], on the basis of 50-year climate gauging
data, has found that the higher the precipitation variability,
the more significant the locust outbreak. This discovery has
been proven by more studies later on [e.g., Wu et al., 1990].
Thus responses of locust outbreaks to climate could be
different from region to region.
[9] 2. Winter temperatures are mostly above the freezing
point in the southern region, but mostly below the point in
the northern region. Although locust eggs cannot survive
below winter temperature of 20°C, there is little chance of
consecutive occurrence of such a temperature for 5 days in
the southern region (latitudes between 35°N and 28°N).
Seasonal temperature as a limiting condition of locust
multiplication would be different between the northern
and southern regions.
[10] In this paper, to understand the impacts of climate
change on occurrence of the important environmental
events and to predict the severe locust outbreaks in an
2 of 11
YU ET AL.: HISTORICAL LOCUST OUTBREAKS IN CHINA
D18104
D18104
Figure 1. Map of study areas and information for locust and climate data. (a) The middle-lower reaches
of Yellow River (northern region) are marked by a light-blue area for the GCM grid boxes and blue dots
for the climate gauging stations. (b) The middle-lower reaches of Yangtze River (southern region) are
marked by light pink for the GCM grid boxes and red pluses for the climate gauging stations. Annual
index of locust outbreaks during the past 1000 years for the (c) northern region and (d) southern region
(digitized from Ma [1958]), in which coarse lines are 10-year mean series.
ultimate aim to avoid a continent-wide food shortage, we
applied both modern and historical records to analyze the
relationships of locust outbreaks and climate changes in
China, in which severe locust hazards occur in two major
areas: the middle-lower reaches of the Yellow River and
the Yangtze River (Figure 1). Compared with the past
1000-year locust index, we correlated climate data on two
time scales: (1) 150-year-long gauging climate on interannual time scale and (2) 1000-year-long reconstructing
climate on decadal time scale; as subsidiary data, simulation
results from a general circulation model (GCM) on both
interannual and decadal time scales can support the climate
proxy. Moreover, spectrum and wavelet analyses are used in
the present study in order to reveal relations of interannual
and decadal time scales.
2. Data and Methods
2.1. Locust Data and Climate Data
[11] The major species of locust that often forms major
outbreaks and causes serious regional crop failures in China
is Locusta migratoria manilensis. The major area of locust
outbreaks is in eastern China (Figure 1). Because of
different outbreak times and abundance, the area can be
divided into two regions of the middle-lower reaches of the
Yellow River (including Haihe River) and the middle-lower
reaches of the Yangtze River (including Huaihe River), in
the northern and southern regions of the eastern China,
respectively (Figures 1a and 1b). Historical literature from
China has recorded abundances and areas of the locust
outbreaks (integration in literature [Zhang, 2004]). A Chinese historian, S.-C. Ma, has delineated it as an annual
locust index based on the severity and affect area for eastern
China [Ma, 1958]. He has separated the area into the
northern region (the middle-lower reaches of the Yellow
and Haihe rivers) and the southern region (the middle-lower
reaches of the Yangtze and Huaihe rivers) (two curves of the
annual locust index in Figures 1c and 1d), and has combined the two into one of whole China.
[12] Although most observations of temperature and/or
precipitation in China were started in 1950 A.D., some
stations have gauging records longer than 100 years; Beijing
Station is the earliest one, with gauging precipitation
records started in 1841 A.D. Within the middle-lower
reaches of the Yellow River and Yangtze River in eastern
China, there are 17 stations with both temperature and
precipitation records longer than 60 years [Climate Data
Section, 1990; Wang, 1994] (see the locations in Figure 1).
[13] Before the gauge records, the climate information
could be traced from historical documents, geological
records (e.g., lacustrine sediment, stalagmites, and ice cap
deposit), biological remains (e.g., sedimentary pollen and
tree rings), and archeological evidence. These climate
proxies have been obtained progressively during the past
20 years in China, but most of them are qualitative and
spatial data. For quantitative and time series data, we found
that there are annual temperature records from the ice cap of
the Tibetan Plateau [Thompson et al., 2000], tree ring
reconstructed precipitation from the mountains in western
China [Shao, 2004], a 10-year-interval temperature with the
3 of 11
D18104
YU ET AL.: HISTORICAL LOCUST OUTBREAKS IN CHINA
data sources mostly derived from Tibetan and the western
mountains [Yang et al., 2002], and recently completed
synthesis of 10-year-interval temperatures for 10 regions
of China [Wang et al., 2007]. For eastern China, the
phenological cold and warm events recorded in historical
documents have been used to reconstruct winter half-year
(October to March) temperature in 10-year intervals since
1500 A.D. and in 30-year intervals for the past 2000 years
[Ge et al., 2003]. On the basis of Ge et al.’s data, continuous
annual temperature series in a 10-year interval were analyzed [Wang et al., 2007], in which stalagmite data for
northern China and pollen data for southern China have
offset the lacks in Ge et al.’s data. The winter half-year
temperature and annual temperature cover the present study
area and provide the quantitative and time series data
available for us.
[14] Currently, monthly climate data of the past 1000 years
are hard to obtain from the proxies but can be provided by
paleoclimate modeling. A few of the past 1000 years of
climate experiments have been performed by global atmosphere-ocean coupled GCMs, for example, HADCM3
[Gordon et al., 2000] and ECHO-G [Legutke and Voss,
1999]. These climate simulations were driven by 1000-year
time-varying external forces including solar radiation, volcanic eruptions, and greenhouse gas concentrations (CO2
and CH4) [Collins et al., 2002; von Storch et al., 2004],
which can be used to examine basic features of the past
1000 years of climate change. We understand that GCM
results would not be independent materials as climate
baseline. In the present study, we use the 1000-year GCM
results as subsidiary data to support the 1000-year climate
proxy (because the climate proxy is 10-year-interval coarse
data, not available for interannual or for seasonal/monthly).
Because the ECHO-G simulations have been validated by
Chinese gauging climate data [Liu et al., 2005; Yu et al.,
2007], we know how far both robust and individual features
of the simulations are in agreement with observed climate
changes; likely ECHO-G has well simulated climate
changes during the Little Ice Age (circa 1650 –1850 A.D.)
for eastern China. Thus we downscaled the ECHO-G girded
data for the northern region and the southern region in the
present study (see grid box in Figure 1).
2.2. Correlation Analysis With Bootstrapping Test and
Spectrum Analysis
[15] Linear correlation was analyzed in each pair of the
annual locust series and the climate series including monthly,
seasonal (DJA for three winter months and June, July, and
August (JJA) for three summer months), and annual mean
temperature/precipitation in turn. We used departures (di)
for the locust index, temperature, and precipitation at time i
xmean. The
(xi) minus the means (xmean), i.e., di = xi
significant level was cut off by statistical confidence interval
(CI). For gauging climate data, we checked the frequency of
regional correlation of individual sites, and also examined
the regional correlation of multiple site means.
[16] In order to examine both temperature (Ti) and
precipitation (Pi) related to locust outbreak years, we did
multiple linear regression (MLR)_on between locust time
series (L) and the two climate variables, i.e., L = a1Ti +
a2Pi + b (i is 1, 2, . . . 12 months), where a1 and a2 are tow
variable’s slopes, which can indicate the variation directions
D18104
with the locust variability. The significant level was tested
by F test.
[17] In order to deduce uncertainties in the above parametric testing, we also use nonparametric testing. Biascorrected percentile bootstrapping test (B test) [Efron and
Tibshirani, 1993] was applied in the present study. Core
codes of the bootstrapping programs were from Hesterberg
et al. [2005]. We did random resamples following the
uniform distribution with sample size of N = 1000 for
10-year series and N = 10,000 for 1-year series, and
accepted the significant correlations at the test level. Final
results are overlapping parts of parametric testing (P) and
nonparametric testing (NP), equivalent to the logical AND
operation by using the function of minimum (P, NP).
[18] To understand the locust-climate relations between
interannual time scale and decadal time scale, we did
spectral analysis and wavelet analysis. Spectral estimation
can describe the distribution over frequency of the powercontained signal that can be buried in a noisy time domain
over long periods. Here we used fast Fourier transforms
(FFT) to exam the frequency components of signals in
1000-year-long locust index and temperature series.
[19] Computing the FFT at each time using only the data
within the window would solve the problem in frequency
localization but would still be dependent on the window
size used. Wavelet analysis has been attempted to solve
these problems by decomposing a time series into time/
frequency space simultaneously [Torrence and Compo,
1998]. We applied the wavelet analysis to find the varied
periods in the locust index series and temperature series.
The wavelet analysis was computed on the basis of a
program wavelet.m (C. Torrence and G. P. Compo, http://
paos.colorado.edu/research/wavelets).
[20] Moreover, the proxy temperature was reconstructed
from multiple lines of evidence, and it mixed more information. We use principal component analysis (PCA) to
transform a number of possibly correlated variables into a
smaller number of uncorrelated variables (i.e., principal
components). Since the first principal component accounts
for as much of the variability in the temperature data as
possible, we take the greatest variance of temperature to
correlate with locust index, obtaining significant results.
3. Results
3.1. Past 100 Years
[21] In the northern region, results of each site correlation
between locust index and monthly temperatures show that
the higher frequencies of significant positive correlations
are in February, April – June, August – September, and JJA, in
which frequency of negative correlation is void (Figure 2a).
On the basis of the regional mean series from nine stations
of the northern region, the locust series has significant
positive correlations in February –April, July, December,
January, and February (DJF), and annual mean (Figure 2c)
temperatures (R = +0.26 +0.36 with 95% CI and B test
with p < 0.05) and in December precipitation (R = +0.74
with 95% CI and B test with p < 0.05). Temperatures from
individual stations and the means reflect that the locust
outbreak was the highest in warm winter, spring-summer
years.
4 of 11
D18104
YU ET AL.: HISTORICAL LOCUST OUTBREAKS IN CHINA
D18104
Figure 2. (a and b) Frequency of monthly temperature correlations and (c and d) comparisons of
regional means of annual temperature with locust index during the past 100 years in northern region
(Figures 2a and 2c) and in southern region (Figures 2b and 2d). In legend, P-cor and N-cor are significant
positive and negative correlations at 95% confidence interval (CI), respectively, and No-cor is none of
significant correlation. Histograms of B resampling correlations between locust index and annual
temperature were plotted by correlation coefficients (x axis) versus the frequency (y axis) in Figures 2c
and 2d.
[22] In the southern region, higher frequency of significant correlations occurred in April, June, and November –
December, in which months there is not negative correlation
at all (Figure 2b). This is consistent with correlations from
the mean climate of eight stations, where significant temperature positive correlations occurred in March – May, and
annual mean (Figure 2d) (R = +0.26 +0.33 with 95% CI
and B test with p < 0.05). Precipitation analyses show that
there is no significant positive correlation but there is a
negative correlation in January (R = 0.28 with 95% CI
and B test with p < 0.05), suggesting that the highest locust
outbreak occurred in warm years with warm spring and dry
January in the southern region.
[23] Significant multiple correlations (F test and B test
with p < 0.05) between locust index and two climate
variables (temperature and precipitation) were listed in
Table 1. Results from the northern region show positive
slope temperatures but negative slope precipitations in
March – April, July, September, JJA, and annual mean,
and both positive slope temperature and precipitation in
May and DJF, suggesting that the highest locust outbreak
occurred in the warm and dry spring-summer years and
warm and wet winter years. This is a similar result to the
single temperature correlations above but shows a combination relationship between temperature and precipitation,
i.e., significant correlations with spring-summer warm and
dry years. In the southern region, temperatures and precipitation in March – April, June, and annual mean have pos-
itive slope correlations with locust index, but January has a
negative correlation.
3.2. Past 300 Years
[24] Correlations of the 10-year mean proxy annual
temperature with 10-year mean locust index (Figures 3a
and 3b) show a significant positive correlation and pass the
B test (p < 0.05) in the northern region (R = +0.60 with 95%
CI) and the southern region (R = +0.34 with 90% CI).
Correlations of locust index with proxy winter half-year
temperatures show a significant positive correlation and
pass the B test in the northern region (R = +0.49 with
95% CI) and the southern region (R = +0.31 with 90% CI).
[25] GCM-simulated 10-year mean temperature series
have significant positive correlations with the locust series
in May– June, September – October, JJA, and annual mean
(R = +0.47 +0.64 with 95% CI and B test with p < 0.05)
in the northern region (Figure 3c). When applying 90% CI
cutoff, there are positive correlations (R = +0.38 +0.46)
in May, July, September, December, and annual mean
(Figure 3d) in the southern region.
[26] MLR analysis shows that the significant correlations
occur in May– July and annual mean (p < 0.05) in the
northern region and in September and December (p < 0.10)
in the southern region (Table 1). They have all positive
significant slopes in temperatures but negative slopes in
June precipitation in the northern region and in September
and December precipitation in the southern region.
5 of 11
YU ET AL.: HISTORICAL LOCUST OUTBREAKS IN CHINA
D18104
Table 1. Significant Correlations Between Locust Index and
Monthly Climate Based on Gauging Observations From 1873 –
1959 A.D. and General Circulation Model Simulations From 1700 –
1950 and 1000 – 1950 A.D.a
Month
R2
F Value
Critical
F Value
Temperature Precipitation
Slope
Slope
3
4
5
7
9
DJFc
JJAd
Annual
Northern Region, 1904 – 1959, Annual Seriesb
0.119 3.571 3.162 (p < 0.05)
0.293
0.143 4.407
0.419
0.149 4.629
0.468
0.114 3.399
0.521
0.141 4.353
0.587
0.131 3.988
0.305
0.112 3.359
0.515
0.109 3.231
0.481
0.05
0.022
0.016
0.003
0.008
0.186
0.002
0.005
1
3
4
6
Annual
Southern Region, 1873 – 1953, Annual Series
0.097 4.169 3.114 (p < 0.05)
0.161
0.119 5.288
0.391
0.09
3.877
0.365
0.107 4.661
0.367
0.098 4.219
0.713
0.02
0.006
0.002
0.006
0.019
5
6
7
Annual
Northern Region, 1700 – 1950, 10-Year Mean Series
0.261 4.062 3.422 (p < 0.05)
0.986
0.474
0.262 4.083
0.984
0.423
0.36
6.462
1.163
0.493
0.26
4.039
1.283
0.058
9
12
Southern Region, 1700 – 1950, 10-Year Mean Series
0.191 2.719 2.549 (p < 0.10)
0.711
0.215 3.157
0.739
0.583
0.747
5
Annual
Northern Region, 1000 – 1950, 10-Year Mean Series
0.070 3.484 3.09 (p < 0.10)
0.168
0.064 3.191
0.836
0.108
6.144
8
11
Southern Region, 1000 – 1950, 10-Year Mean Series
0.069 3.422 3.09 (p < 0.10)
0.574
0.368
0.071 3.553
0.489
0.156
a
The results are also passed by bootstrapping test.
Region, period, and series.
c
December, January, and February.
d
June, July, and August.
b
3.3. Past 1000 Years
[27] We did not find any significant correlations between
locust index and proxy temperature in the 10-year mean
series even at 90% CI level. When using proxy temperatures
from 10 regions of China to do the PCA, we found major
changes of temperatures during the last 1000 years. Variance sums of the first three PCs in both the northern and
southern regions are greater than 80%, suggesting that we
may ignore 20% local impacts on the temperature
changes. Correlations between the annual temperature variance and locust index show significant positive correlations
(95% CI and B test with p < 0.05) in the northern region
(R = +0.31) for the last 1000 years, and in the northern region
(R = +0.43) and southern region (R = +0.32) for the periods
of years 1000 – 1500 A.D. (Figure 4). These results suggest
that decadal time scale locust outbreaks were closely linked
with the temperature changes to warm years.
[28] We did not find significant correlations of locust
index with GCM-simulated monthly temperature/precipitation
in the 10-year mean series, whether monocorrelation or
multiple correlations. However, correlations between the
10-year mean GCM temperature variance and 10-year mean
D18104
locust index show significant positive correlations in May
(R = +0.242) in the northern region and in August (R =
+0.243) in the southern region (95% CI and B test with p <
0.05). Multiple correlations show similar results: that the
significant warm-dry changes in May and annual mean in
the northern region (F test and B test with p < 0.05) were
correlated with locust outbreak years (Table 1), while a
warm-dry change in August and a warm-wet change in
November in the southern region only passed test level p <
0.10 by F test and B test.
[29] Results of FFT analysis show that there are several
cycle years of locust series as the same as the proxy
temperature series. Both cycle years above 95% CI null
continuum curves are 80, 40, 27, 22, and 21 years in the
northern region (Figure 5a) and 107, 80, 49, 40, 30, and 20
years in southern region (Figure 5b), reflecting that frequency of locust outbreaks has synchronous changes with
the temperature in decadal time scale cycles.
[30] Examining the 10-year interval cycle years by wavelet analysis (Figures 5c – 5f), we found the significant cycle
years in two regions concentrated in 20 to 110-year periodicities, consistent with the FFT analysis. More specifically,
wavelet analysis provided a way to test if the locust
outbreak cycles respond to climate change in timing.
According to variance series at the 20 – 110 periodicity
years by wavelet computing, there are five periods that
are significantly matched (95% CI) between locust outbreak
years and temperature changes during 1210 – 1230, 1280–
1310, 1470 – 1510, 1670– 1760, and 1800 – 1850 A.D. in the
northern region (gray areas in Figures 5c and 5e). These
overlapping periods are 240 years and take 56.2% of the
total 427 years of locust outbreaks. For the southern region,
there are four matched periods during 1040 –1130, 1280–
1330 and 1490 – 1520 and 1610 – 1630 A.D. (gray areas in
Figures 5d and 5f). These matched periods are 230 years,
totaling 64.6% of the total 356 years of locust outbreaks.
[31] We further examined interannual relationships between locust index and GCM monthly temperatures during
the past 1000 years. FFT and wavelet analyses show that
there are significant periods (95% CI) at 2– 8 years, at 12–
16 years and at approximately 30 years that commonly
occurred both in the locust series and annual temperature
series (figures not shown). According to variance series at
the 2- to 10-year periodicity by the wavelet computing,
there are eight matched periods with 10 successive years
between locust series and temperature series, during 1444–
1472, 1481 – 1493, 1505 – 1518, 1582 – 1592, 1684 – 1696,
1809– 1821, 1927– 1939, and 1942 – 1953 A.D. (Figure 4c).
These overlapping years are 252 years, totaling 59.7%
locust outbreak years. In the southern region, nine matched
periods with 10 successive years occurred during 1429–
1440, 1444 – 1472, 1481 – 1493, 1505 – 1518, 1585 – 1592,
1684 – 1696, 1809 – 1821, 1927 – 1939, and 1942 – 1956
A.D. (Figure 4d). There are 205 matched years, totaling
57.6% locust outbreak years.
4. Discussion and Summary
[32] There are 427 locust outbreak years in the total
960 years (1000 – 1959 A.D.) in the northern region, and
356 locust outbreak years in the total 954 years (1000–
1953 A.D.) in the southern region. Because of differences of
6 of 11
D18104
YU ET AL.: HISTORICAL LOCUST OUTBREAKS IN CHINA
Figure 3. (a and b) Comparisons of locust outbreaks of the past 300 years on decadal time scale with
proxy temperatures and (c and d) with GCM simulated temperatures of the annual mean in the northern
region (Figures 3a and 3c) and in the southern region (Figures 3b and 3d). Coarse lines are 30-year
running averages in Figures 3a and 3b. Histograms of B resampling correlations between locust index and
temperature as Figure 2.
Figure 4. (a and b) Decadal time scale correlations of proxy temperature variance and (c and d)
interannual time scale correlations of GCM-simulated temperature variance with the same time scale
locust index of the past 1000 years. Two pairs during 1000 –1500 A.D. have significant positive
correlations with 95% CI. In Figures 4a and 4c, interannual series were cut off by 95% CI baseline for
clear presentation purpose. Figures 4a and 4c are the northern region, and Figures 4b and 4d are the
southern region. Histograms of B resampling correlations between locust index and temperature as
Figure 2.
7 of 11
D18104
D18104
YU ET AL.: HISTORICAL LOCUST OUTBREAKS IN CHINA
D18104
Figure 5. (a and b) Comparison of fast Fourier transform spectral power spectrum of 10-year mean
locust index series and proxy temperature time series, (c and d) wavelet power spectrum contours for the
proxy temperature, and (e and f) the locust index. Figures 5a, 5c, and 5e are the northern region, and
Figures 5b, 5d, and 5f are the southern region. The wavelet contours were plotted by time series (x axis)
versus periodicity year (y axis), in which 95% significance contour is the black dotted line, and the black
arc solid line indicates cone of influence. Anything below the arc is dubious. Gray areas show
overlapping significant periods of locust outbreak years with temperature changes.
data types and time scales in climate series, we did eight
pair sets of monocorrelations and multiple correlations with
locust outbreak time series to check climate change impacts.
A summary diagram was plotted for the locust outbreak
correlations with changes in monthly/season/annual mean
of temperatures and precipitation (Figure 6). Correlate
coefficients and linear regression slopes in Figure 6 can
indicate directions of positive or negative relationships.
Although there is some uncertainty in the statistics, we
can obtain more robust features by F test and B test ( p <
0.05). In the northern region, locust outbreak years
responded to warm year, warm winter, and warm-summer-
half year during the past 300 years in 10-year mean series
(Figure 6a), and responded to warm year, warm winter, and
warm March – May, July, and September during the past
100 years in interannual series (Figure 6c). Negative June
precipitation change during the past 300 years (Figure 6a)
and negative March – April, July, September, and annual
precipitation change during the past 100 years (Figure 6c)
are significantly correlated with locust outbreak years. In
100 year series of the southern region (Figure 6d), locust
outbreak years responding to warm year and warm springsummer year still stand, except for a colder January
temperature response. Dry January and wet spring were
8 of 11
D18104
YU ET AL.: HISTORICAL LOCUST OUTBREAKS IN CHINA
D18104
Figure 6. Summary diagrams for locust outbreak years significantly correlated with temperature and
precipitation (a) during 1950– 1000 A.D. (p < 0.05) in the northern region, (b) during 1950 – 1000 A.D.
( p < 0.05) in the southern region, (c) during 1950 – 1850 A.D. (p < 0.05) in the northern region, and
(d) during 1950– 1850 A.D. (p < 0.10) in the southern region. Mono correlation was plotted by the
coefficients marked by dots, squares, and rhombuses for different times, and multiple correlations plotted
by the slopes marked by bar.
significantly correlated with locust outbreak years. There is
less robust feature during the past 300 years, showing a
warm-dry summer-autumn linked with locust outbreak in
the southern region (p < 0.10).
[33] Generally speaking, the highest locust outbreak years
occurred in warm-wet winter and warm-dry spring-summer
in the northern region and warm and wet spring years in the
southern region. There is less robust feature in the southern
region (p < 0.10), showing a warm-dry summer-autumn
linked with locust outbreak years. We can see that locust
outbreak responding to seasonal climate change on a
decadal time scale is generally consistent with that on an
interannual time scale.
[34] Without doubt, locust outbreaks were closely linked
with warm winter because with its favorable conditions,
locust eggs can be well preserved, and locust outbreaks
were strongly linked with warm spring and summer because
during it, as an optimum season, locusts can greatly multiply and disperse. In fact, it is true that the locust outbreak
frequency of April – September was 85.8% among the
12 months according to 544 historical outbreak records
[Zheng, 1990], and a relationship of warm winter and warm
spring-summer with severe locust outbreaks has been found
for a long time in Provinces Henan, Hebei, Shangdong,
Anhui, Jiangsu, and Liaoning of eastern China [Ma, 1958;
Zhang and Chen, 1998; Zheng, 1990; Kong et al., 2003;
Ren and Tang, 2003; Wang et al., 2006; Zhang et al., 2006].
Here we confirmed their studies and did more work on
monthly analysis, larger area examinations, and longer time
period correlations. Thus we believe that the greenhouseeffected climate warming in the future would strengthen or
increase the locust outbreaks. Actually, locust outbreak
records in the agriculture statistics annals [Statistics Annals
Committee of Agriculture in China, 2001, 2002, 2003,
2004] show a marked increase since 2000. Sum of the
disastrous area in provinces Shangdong, Hebei, Henan, and
Liaoning (all in the northern region) is 19,360, 22,000,
18,280, and 19,200 km2 for 2000, 2001, 2002, and 2003,
respectively, which were increased 34% – 83% compared to
the mean of 1995 – 1999 [Statistics Annals Committee of
Agriculture in China, 1996, 1997, 1998, 1999, 2000].
[35] Study of the relationship of locust outbreaks to
precipitation changes leads us to recognize that precipitation
condition is one important climate factor impacting locust
outbreaks. Our results are consistent with previous studies.
Zhang et al. [2006] correlated summer locust outbreak
significantly with lower June precipitation and autumn
locust outbreak significantly with lower July precipitation
in Shangdong Province; lower April precipitation has led to
summer locust outbreaks in Henan Province [Kong et al.,
2003], and lower May – June precipitations resulted in
summer locust outbreaks in Liaoning Province [Wu et al.,
1990].
[36] Compared with the northern region, locust outbreaks
with precipitation relationship in the southern region were
less robust (test level at p < 0.10). We think that because
regional precipitation is much higher in the southern region
than in the northern region, more locust outbreaks occurred
in the northern (427 times) than in the southern region
(356 times), as wet conditions greatly restrains locust flying
and dispersing [Ma, 1958; Zheng, 1990]. Thus, historical
records on ‘‘drought locust’’ were true, as we found the
drought of spring-summer season in the northern region
caused the highest regional locust outbreaks, but it is not
9 of 11
D18104
YU ET AL.: HISTORICAL LOCUST OUTBREAKS IN CHINA
obvious in the wet-humid southern region. Hence we think
that precipitation change is not a sensitive factor for locust
outbreak in the Yangtze River region.
[37] For the negative January temperature correlation in
the southern region, we also think it is a regional feature.
There is no record of consecutive 5-day temperatures in
January lower than 20°C in this region, where the temperature condition would wipe out the locust eggs. For
example, on the basis of the past 50 years of observation
from Bengbu Meteorological Station (32.95°N and
117.37°E), the monthly mean of minimum January temperature is 2.2°C and the monthly minimum is 7.5°C,
where it is located near the northern boundary of the
southern region. Hence, although warm winter condition
is a key factor for locust egg survival and preservation, it
works well when winter temperature reaching to 10 to
30°C in the northern region, while January temperature
almost above the freezing point in the southern region is not
a limiting factor for locust survival.
[38] High-resolution climate proxy and GCM modeling
for the past 1000 years are still a challenge for us because of
the uncertainties [Crowley, 2000; Bauer et al., 2003; Wang
et al., 2007]. The locust outbreak index of 1700 – 1000 A.D.
correlated with the temperature proxies or with simulations
did not show any correlations (R2 < E-2 in the northern and
southern regions). However, there are obvious relationships
of locust outbreak with the variability, trends, and periods of
climate. The decadal time scale series analysis presented
significant relationships showing that the highest locust
outbreak years were in warm years commonly in the
northern and southern regions. However, it shows some
difference in the two regions: the highest locust outbreak
years were in warm-winter-half years with warm-dry May
in the northern region, but warm-winter-half years with
warm-dry August in the southern region. These synchronous changes with 20– 110 periodicity years were 56– 65%
among the total locust outbreak years during the past 1000
years.
[39] Our decadal time scale series analyses are different
from those of Stige et al. [2007]. They believed the decadal
mean locust outbreak is highest during cold and wet
periods. We trust their study method but doubt the data
they used. Outbreaks of the locust caused serious crop
failures mainly in the middle-lower reaches of Yellow River
and Yangtze River, East Plains, and Coast Plains in lowlands of eastern China. S.-C. Ma concentrated on the areas
and deleted some records from the northern – western China
when he developed the index [Ma, 1958]. In the work of
Stige et al., only one temperature curve of whole China
[Yang et al., 2002] was used to correlate the locust outbreaks. By doing this, they unconsciously enlarged the
covering areas of the locust outbreak. Actually, the reconstructed temperature of the past 1000 years from Yang et al.
[2002] used in their study was derived mainly from ice
cores in the Tibetan Plateau, tree rings in western mountains
of China, and lake sediments also in the Tibetan and other
western areas of China, while the historical locust outbreak
records from the western regions are not available so far. In
fact, they correlated the climate of western China with
locust outbreaks of eastern China. Moreover, using one
precipitation curve of Qinghai (located at 75°E– 105°E
and elevation >3000 m above sea level (asl)) to correlate
D18104
locust outbreaks in eastern China (105°E – 120°E and
elevation <1000 m asl) would diminish the differences of
geomorphologies between the plateau and the lowlands and
upset the rainfall patterns between the western semiarid
climate and the eastern monsoon climate.
[40] To prove the above arguments, we did correlations of
10-year mean series of locust index with decadal temperature proxy for all China (10 region means) and eastern
China (seven region means) according to geography delineation of Wang et al. [2007]. We found no correlations for
1000-year series except for 300-year series: the locust index
from the northern region is significantly positively correlated (p < 0.05) with temperatures from all China (R = +0.63)
and from eastern China (R = +0.68). Locust index from the
southern region is significantly positively correlated ( p <
0.05) with temperatures from all China (R = +0.38) and
from eastern China (R = +0.40). Relationships between
whole China/eastern China climates and locust outbreak
index cannot return negative correlations. These results
prove that temperature series used by Stige et al. [2007]
are from western China, but are not from whole China nor
from eastern China, which are areas the locust outbreak
records were derived from.
[41] Furthermore, observations reveal that outbreaks of
the locust in eastern China normally occur in the seasons of
summer half year when the locust eggs were preserved in
the previous warm winter seasons [Wu et al., 1990]. This
relationship of seasonal variability has been confirmed in
the present study, since our results with 1-year series and
10-year mean series were very consistent. Using only
annual temperature and annual precipitation to undertake
correlations would ignore the seasonal variability totally,
and consequently could hardly draw a true deduction for
understanding the climate impacts. We would like to emphasize that, regardless of interannual or decadal time scales
of the data sets, study of such type of environmental events
should produce consistent relationships in terms of dynamics. In terms of Asian monsoon climatology, a warm year in
eastern China is mainly caused by higher winter temperature (variance >50%), and a dry year, caused by lower
summer precipitation (variance 40– 65%), and vice versa
[Ren, 1992]. These features are also found in longer climate
period in centennial and millennium time scales [Yu and
Qin, 1997; Yu et al., 2003]. Thus it is very difficult to
imagine that locust outbreak could behave in an opposite
way between interannual time scale and decadal time scale,
in terms of mechanisms.
[42] We understand that there were some geographical
gaps and blind points to establish the past 1000-year climate
series. It is premature to deduce a conclusion on lowfrequency variability by using the above data set with such
a low resolution. By doing this, one can only receive a very
ambiguous understanding of the locust dynamics. Nowadays, high-resolution climate data sets of temperature and
precipitation for the past millennium are ongoing in China.
The less-uncertainty data and simulations allow us to
reexamine the issues if and how severe disasters of locust
outbreaks respond to greenhouse-effected climate change.
These are promising points for future research.
[43] Acknowledgments. Financial support for this work was provided
by Key Directional Program of the Knowledge-Innovation Project of the
10 of 11
D18104
YU ET AL.: HISTORICAL LOCUST OUTBREAKS IN CHINA
Chinese Academy of Sciences: High-resolution lake records and the
environmental hydrology simulations during the past 3000 years in China
(KZCX2-YW-338-2) and by the Nanjing Institute of Geography and
Limnology: Data and modeling comparisons for dynamic simulations of
paleoclimate (CXNIGLAS-2006-07). The authors would like to thank the
anonymous reviewers for their insightful comments.
References
Bauer, E., M. Claussen, V. Brovkin, and A. Huenerbein (2003), Assessing
climate forcings of the Earth system for the past millennium, Geophys.
Res. Lett., 30(6), 1276, doi:10.1029/2002GL016639.
Chen, J. X. (1935), Historical locust records in China (in Chinese), report,
Insect Bur. of Zhejiang Prov., Hangzhou, China.
Climate Data Section of Beijing Center of National Meteorology Bureau of
China (1990), China climate data (in Chinese), report, Beijing.
Collins, M., T. J. Osborn, and S. F. B. Tett (2002), A comparison of
variability of a climate model with paleotemperature estimates from a
network of tree-ring density, J. Clim., 15, 1497 – 1515, doi:10.1175/15200442(2002)015<1497:ACOTVO>2.0.CO;2.
Crowley, T. J. (2000), Causes of climate change over the last 1000 years,
Science, 289, 270 – 277, doi:10.1126/science.289.5477.270.
Efron, B., and R. J. Tibshirani (1993), An Introduction to the Bootstrap,
Chapman and Hall, New York.
Ge, Q. S., J. Zheng, X. Fang, Z. Man, X. Zhang, P. Zhang, and W.-C. Wang
(2003), Winter half-year temperature reconstruction for the middle and
lower reaches of the Yellow River and Yangtze River, China, during the
past 2000 years, Holocene, 13(6), 933 – 940.
Gordon, C., C. Cooper, C. A. Senior, H. Banks, J. M. Gregory, T. C. Johns,
J. F. B. Mitchell, and R. A. Wood (2000), The simulation of SST, sea ice
extents and ocean heat transports in a version of the Hadley Centre
coupled model without flux adjustments, Clim. Dyn., 16, 147 – 168,
doi:10.1007/s003820050010.
Hesterberg, T. C., D. S. Moore, S. Monaghan, A. Clipson, and R. Epstein
(2005), Bootstrap methods and permutation tests, in Introduction to the
Practice of Statistics, 5th ed., edited by D. P. Moore and G. P. McCabe,
chap. 14, Freeman, San Francisco, Calif. (Available at www.insightful.
com/Hesterberg/bootstrap).
Kong, H. J., W. S. Lu, G. Q. Lu, X. D. Wang, and J. J. Wang (2003), Effect
of drought and higher temperature on the outbreaks of the Oriental
migratory locust in Henan Province (in Chinese), J. Nanjing Inst.
Meteorol., 26, 516 – 524.
Legutke, S., and R. Voss (1999), The Hamburg atmosphere-ocean coupled
circulation model ECHO-G, Tech. Rep. 18, pp. 1 – 88, Ger. Clim. Comput.
Cent., Hamburg.
Liu, J., H. Storch, X. Chen, E. Zorita, J. Zheng, and S. Wang (2005),
Simulated and reconstructed winter temperature in the eastern China
during the last millennium, Chin. Sci. Bull., 50(24), 2872 – 2877.
Liu, M. G., Y. F. Han, Z. X. Zhang, L. Z. Zhao, and Q. G. Zhan (1998),
Precipitation variance of China, in Atlas of Physical Geography of China
(in Chinese), p. 43, China Cartogr. Press, Beijing.
Lu, R. J. (1986), Preliminary study on historical locust plagues in China (in
Chinese), Agric. Archaeology, 1, 311 – 324.
Ma, S.-C. (1958), The population dynamics of the Oriental migratory locust
(Locusta migratoria manilensis) in China (in Chinese), Acta Entomoogica
Sin., 8, 1 – 40.
Ni, G. J. (1998), Plague of locusts and the controlling in Chinese histories
(in Chinese), Hist. Educ., 6, 48 – 51.
Ren, C. G., and T. C. Tang (2003), Analysis of population dynamics and
trends of future outbreaks of the Oriental migratory locust in Hebei (in
Chinese), Entomological Knowledge, 40(1), 80 – 82.
Ren, M. E. (Ed.) (1992), Elements of the Physical Geography of China (in
Chinese), 430 pp., Commer. Press, Beijing.
Shao, X. M. (2004), Treering-recorded precipitation changes for the last
1000 years in Delingha region, Qinghai Province, Chin. Sci., D Ser.,
34(2), 145 – 153.
Statistics Annals Committee of Agriculture in China (1996), The Statistics
Annals of Agriculture in China (in Chinese), China Agric. Press, Beijing.
Statistics Annals Committee of Agriculture in China (1997), The Statistics
Annals of Agriculture in China (in Chinese), China Agric. Press, Beijing.
Statistics Annals Committee of Agriculture in China (1998), The Statistics
Annals of Agriculture in China (in Chinese), China Agric. Press, Beijing.
Statistics Annals Committee of Agriculture in China (1999), The Statistics
Annals of Agriculture in China (in Chinese), China Agric. Press, Beijing.
D18104
Statistics Annals Committee of Agriculture in China (2000), The Statistics
Annals of Agriculture in China (in Chinese), China Agric. Press, Beijing.
Statistics Annals Committee of Agriculture in China (2001), The Statistics
Annals of Agriculture in China (in Chinese), China Agric. Press, Beijing.
Statistics Annals Committee of Agriculture in China (2002), The Statistics
Annals of Agriculture in China (in Chinese), China Agric. Press, Beijing.
Statistics Annals Committee of Agriculture in China (2003), The Statistics
Annals of Agriculture in China (in Chinese), China Agric. Press, Beijing.
Statistics Annals Committee of Agriculture in China (2004), The Statistics
Annals of Agriculture in China (in Chinese), China Agric. Press, Beijing.
Stige, L. C., K.-S. Chan, Z. Zhang, D. Frank, and N. C. Stenseth (2007),
Thousand-year-long Chinese time series reveals climatic forcing of
decadal locust dynamics, Proc. Natl. Acad. Sci. U. S. A., 104(41),
16,188 – 16,193, doi:10.1073/pnas.0706813104.
Thompson, L. G., T. Yao, E. M. Thompson, M. E. Davis, K. A. Henderson,
and P. N. Lin (2000), A high-resolution millennial record of the south
Asian monsoon from Himalayan ice cores, Science, 289, 1916 – 1919,
doi:10.1126/science.289.5486.1916.
Torrence, C., and G. P. Compo (1998), A practical guide to wavelet analysis, Bull. Am. Meteorol. Soc., 79(1), 61 – 78, doi:10.1175/15200477(1998)079<0061:APGTWA>2.0.CO;2.
von Storch, H., E. Zorita, and J. M. Jones (2004), Reconstructing past
climate from Noisy data, Science, 306, 679 – 682, doi:10.1126/
science.1096109.
Wang, H., S. P. Zhang, and W. Jin (2006), Occurrence regularity and
management strategy of Locusta migratoria manilensis (in Chinese),
Henan Agric. Sci., 6, 75 – 78.
Wang, S. W. (1994), Diagnostic studies on the climate change and variability for the period of 1880 – 1990 (in Chinese), Meteorol. Sinica, 52(3),
261 – 273.
Wang, S. W., D. Y. Gong, J. L. Ye, and Z. H. Chen (2000), Seasonal
precipitation series of eastern China since 1800 and the variability (in
Chinese), Acta Geogr. Sin., 55(3), 281 – 293.
Wang, S. W., X. Y. Wen, Y. Luo, W. J. Dong, Z. C. Zhao, and
B. Yang (2007), Reconstruction of temperature series of China for the last
1000 years, Chin. Sci. Bull., 52(23), 3272 – 3280, doi:10.1007/s11434007-0425-4.
Wu, F. Z. (1951), East-Asia-Locust in China (in Chinese), 2 pp., Shanghai
Yongxiang Publication, Shanghai, China.
Wu, F. Z., , S. J. Ma, and H. F. Zhu (Eds.) (1990), Oriental migratory locust,
in Encyclopedia of Agriculture in China (in Chinese), pp. 73 – 78, China
Agric. Press, Beijing.
Yang, B., A. Braeuning, K. R. Johnson, and S. Yafeng (2002), General
characteristics of temperature variation in China during the last two millennia, Geophys. Res. Lett., 29(9), 1324, doi:10.1029/2001GL014485.
Yu, G., and B. Qin (1997), Holocene temperature and precipitation reconstructions and monsoonal climates in eastern China, Paleoclim. Data
Modell., 2, 1 – 32.
Yu, G., B. Xue, J. Liu, and C. Chen (2003), LGM lake records from China
and analysis of the climate dynamics, Global Planet. Change, 38, 223 –
256, doi:10.1016/S0921-8181(02)00257-6.
Yu, G., J. Liu, and B. Xue (2007), Dynamical Paleoclimate Simulations (in
Chinese), 337 pp., High Educ. Press, Beijing.
Zhang, D. E. (Ed.) (2004), Integrations of China Meteorological Records of
the Past 3000 Years (in Chinese), 3666 pp., Fenghuang Press, Nanjing,
China.
Zhang, D. E., and Y. L. Chen (1998), Paleoclimate inferred from the Chinese historical records of the northern boundary of migratory locust (in
Chinese), Quat. Sci., 1, 12 – 19.
Zhang, L. S., J. Z. Liu, and Q. N. Liu (2006), Studies on the mechanisms of
yearly severe outbreaks of the Oriental migratory locust in Shandong
Province (in Chinese), China Plant Prot., 26, 41 – 43.
Zheng, Y. F. (1990), Analysis of locust hazards in history of China (in
Chinese), Agric. Hist. China, 4, 38 – 50.
Zhong, Z., Y. J. Hu, and J. Z. Min (2007), Sensitivity of interannual and
interdecadal precipitation variability over China to spatial scale (in
Chinese), Chin. J. Geophys., 50(5), 1330 – 1336.
J. Liu, H. Shen, and G. Yu, Nanjing Institute of Geography and
Limnology, Chinese Academy of Sciences, 73 East Beijing Road, Nanjing
210008, China. ([email protected])
11 of 11