The Largest Act of Environmental Warfare in History

Steven Dutch, Professor Emeritus, Natural and Applied Sciences, University of Wisconsin - Green Bay 54311-7001



Although the destruction of oil fields in the Persian Gulf War of 1991 riveted public attention on war and the environment, this was neither the first nor the largest act of environmental warfare in history. In terms of loss of life and geographic area affected, probably the largest-scale act of environmental warfare in history was the breaching of the Huang He (Yellow River) levees by the Nationalist Chinese in June, 1938. Detailed information about this event has only recently become available in English and even today information is sketchy. The breach was an attempt to slow down the westward advance of the Japanese army south of the Huang He. It was marginally successful from a military standpoint but caused enormous loss of life from direct flooding, disease and famine. Official Chinese figures place the death toll at over 800,000.


Environmental warfare is ancient, but two elements have changed in the last century. First, technology has made it possible to affect the environment on unprecedented scales and in new ways. Whereas military disturbance of terrain was once limited to earthworks and shell craters, modern earthmoving technology has made it possible to construct massive strategic installations like the Panama Canal or the Interstate Highway System, or tactical barriers like the Maginot Line. Also, chemical weapons, the possibility of attacks on chemical or nuclear plants, and defoliants have created new dangers to the environment. Second, environmental terrorism, such as the events in Kuwait in 1991, is new, because concern about the environment has only recently become general enough to make environmental terrorism a credible political weapon.

Apart from small-scale defensive engineering and siege technology, environmental warfare in centuries past was mostly directed against food and water supplies. Examples of destruction of food supplies in U. S. history include Sheridan’s scorched earth campaigns during the Civil War in the Shenandoah Valley, Sherman’s campaigns in Georgia and the Carolinas, and the systematic eradication of bison herds to deprive Plains Indian tribes of sustenance. However, water is an even more critical need than food. Most places likely to endure siege sought to obtain a secure water source, and a common siege tactic was to dam or divert rivers to deprive a besieged city of water while opening a possible new avenue of attack. The theory of breaching of dikes to drown or impede an opponent was well known in early Chinese warfare (Sawyer, 2004). In 1938, one such action by the Nationalist Chinese army marginally impeded the invading Japanese army but caused almost a million Chinese deaths.

The purposes of this paper are threefold. First, I aim to provide additional details of this event, which is frequently mentioned in passing but rarely described in any detail. Second, I intend to summarize the state of information in English and to identify areas where sources conflict or are incomplete. The most comprehensive source in English is that of Lary (2001). The best map showing the extent of flooding is that of Todd (1949), who also presented by far the best account of the repair of the breach. Third, I hope to stimulate thinking about possible research topics for instructors and students who are in a position to access Japanese and Chinese sources on this event. Most published information was written by military observers, journalists or political scientists and almost nothing, apart from the account of Todd (1949), seems to have been written from the perspective of a geoscientist, civil engineer, or environmental scientist. Present-day Chinese Romanizations are used, with spellings and usages current in 1938 in parentheses.

Geologic Setting

The Huang He (Yellow River) originates on the northeastern margin of the Tibetan Plateau in Qinghai and flows 5464 km east to the Yellow Sea (Figure 1). It drains an area of 752,443 km2, with a discharge of 1822 m3/sec or an annual discharge of 57.5 km3 of water (Zhao, 1986). On the way to the sea, it describes a great square bend that encloses the Ordos Desert. The square bend is governed by normal faulting, a far-field effect of the collision between India and Eurasia (Lin and others, 2001). Prior to faulting, the Huang He probably flowed straight east across the south side of the bend. The present Weihe (Wei River) is a relic of the original course. The southern half of the square bend, and a large tract to the east, are thickly mantled in loess. This region is called the Loess Plateau and is the source of the high sediment load in the Huang He (Figure 2).

Flow in the Huang He is lowest during the winter months. There is a small flood (“peach blossom” flood; Todd, 1949) in late March and early April from snow melt upstream, a period of low water, then the principal floods occur during the summer rainy season from late June to October. The Huang He is the most heavily silt-laden large river in the world, with suspended solids approaching half the total volume during peak floods, although sediment load is small during low water (Todd, 1949). The Huang He carries about a sixth of the total sediment carried to the oceans by the world’s 21 largest rivers (Meie and Yunliang, 1991) and about 7% of total global sediment discharge. A substantial part of this extreme sediment load in recent centuries is anthropogenic, resulting from deforestation of the Loess Plateau. In recent decades the sediment discharge has decreased due to reforestation, sediment trapping by dams, diminished rainfall and increased diversion for agriculture (Shi and others, 2002; Wang and others, 2007; Xiubin and others, 2004).

Once the river exits the highlands, it flows across a gently sloping plain that is effectively a vast alluvial fan consisting of redeposited loess. The fan has built seaward until it envelops the Shandong Peninsula. The Huang He sweeps out a conical surface as it repeatedly fills in its channel and then jumps to a new location. Abandoned channels are frequently marked not by troughs but by low ridges, the result of natural and artificial levee-building that allowed the river bed to aggrade above the surrounding land (Xu, 1993). Chinese records document dozens of changes of course in the last 2,500 years (Figure 3; Zhao, 1986). The hilly Shandong Peninsula prevents the river from discharging in that direction, with the result that the Huang He tends to alternate between outlets north and south of the peninsula. The Huang He is unique among the world’s large rivers for its drastic changes of course. Although rivers with active deltas frequently experience changes in distributaries, no other large river experiences wholesale diversions 500 kilometers inland to new outlets hundreds of kilometers distant from the old outlet. A rough American equivalent in scale would be the Mississippi River being diverted below Memphis, Tennessee to outlets between Beaumont, Texas and Pensacola, Florida

The upper course of the Huang He is most vulnerable to change for a variety of reasons. First, the reduced gradient as the river exits the highlands leads to rapid sedimentation and channel aggradation. Second, diversions high on the flood plain result in the greatest changes of course, as opposed to more local changes from diversions at lower elevations. Not only does the upper part of the alluvial fan experience great natural diversions, it has also been the site of the most significant military diversions of the river. Lary (2001) describes the 1938 diversion as something never before done, but in reality there have been at least two other diversions. In 1128, the Sung Dynasty breached the levees at Kaifeng to thwart the invading army of the Jin Dynasty (Lamoroux, 1998). This maneuver may have blocked the invasion but also diverted the river to a southerly course, inadvertently relocating the border between the two dynasties and conceding a large tract of land to the Jin invaders.

In 1642, during a peasant revolt, the levees at Kaifeng were again breached. Accounts differ as to whether it was the government that caused the breach to block the attacking rebels, or whether the rebels caused the breach to overwhelm the city’s defenses. In any case, the city was severely damaged and there was great loss of life, estimated at about a quarter of a million. One cultural effect of this flood was the near-destruction of the ancient Jewish community in Kaifeng (Xin, 2003). The 1938 levee breach occurred in the same general area but farther west.

Prologue: The Second Sino-Japanese War

By the late 19th century, Japan had emerged as a world-rank military power and sought to expand its dominance to the Asian mainland (Dreyer, 1995; Morley, 1983). In 1895, after a short war with China commonly termed the First Sino-Japanese War, Japan occupied Taiwan and forced China to recognize the independence of Korea. Japan’s ambitions on the mainland brought it into conflict with Russia, which it defeated during the Russo-Japanese War of 1904-06. During World War I, Japan sided with the Allies, and after the war, asserted a claim on Germany’s former rights in China, thereby acquiring a number of garrisons and enclaves, notably the Shandong Peninsula. In 1931 a bomb exploded on a Japanese-owned railway in Manchuria. There is widespread suspicion that the incident may have been staged to give Japan a pretext for invading Manchuria. In any case, Japan quickly occupied Manchuria and set up the puppet state of Manchukuo in 1932. As ruler, the Japanese installed Henry Pu Yi, the last Manchu emperor, whose story is told in the film The Last Emperor (Bertolucci and Peploe, 1987). Real and staged anti-Japanese incidents in Shanghai after the occupation of Manchuria resulted in a short conflict that ended with Japan being permitted to station troops in Shanghai.

After occupying Manchuria, Japan proceeded to nibble away at adjacent Chinese territory until by 1937 they had hemmed in Beijing (Peking) on three sides, as well as creating Japanese-controlled provincial governments north of the Huang He. A skirmish at the Marco Polo Bridge, the last bridge connecting Beijing to the rest of China on July 7, 1937, escalated into full-blown invasion, and Japan occupied Beijing on July 29. Among the numerous references that chronicle the Second Sino-Japanese War and its continuation into World War II are Chang-tai (1994), Dorn (1974), Dreyer (1995), Fischer (1991), Hsiung and Levine (1992), Lary and McKinnon (2001), Li (1975), Lindsay (1975), Liu (1956), Long-hsuen and Ming-Kai (1972), Moser (1978), Pu-Yu (1974), Sih (1977) and Wilson (1982). By the end of 1937, Japan had occupied all of China north and east of the Huang He. The Chinese government was not in a position to challenge Japan north of the Huang He, but they struck back where they could, against Japanese forces in Shanghai. The Japanese launched an invasion of Shanghai in August, but unexpectedly stubborn Chinese resistance held out until November. The Japanese then pushed west to Nanjing (Nanking), which they captured in December. The campaign had been more difficult and lengthy than the Japanese expected, and the Japanese army vented its fury on the populace of Nanjing in the infamous Rape of Nanking (Chang, 1998).

Breaching the Dikes

In 1938 the Japanese consolidated their hold on the coast of China between Shanghai and the Huang He. A bitter campaign for the city of Xuzhou (Suchow) occupied the first five months of 1938. At one point Chinese forces scored their first major victory, but on the whole Japan won the campaign, though they were unable to envelop the Chinese. By May the Chinese were in retreat and the Japanese were advancing on Zhengzhou (Chengchow). Zhengzhou was a strategic rail junction where an east-west line parallel to the Huang He crossed a north-south line from Beijing to Wuhan (Hankow), the temporary capital. The Japanese strategy was to capture Xuzhou and Zhengzhou, then drive south along the rail line to Wuhan (Figure 4).

June is the start of flood season in China, and the levee breach came in early June while the water was still fairly low. The first press report of a breach was in the New York Times on June 8 (Abend, 1938). The Times account claimed the breach was made on June 6, but the location shown in the accompanying map is east of Kaifeng and well east of the location generally cited for the main breach. It is possible that this breach was a purely tactical move meant as a brief delaying action. Whether it contributed to the later massive flooding is unknown. In any case, the Times account describes the river as “abnormally low for this time of year,” and expressed doubt that the breach would seriously impede the Japanese because of the low water level.

Lary (2001) claims, based on Chinese accounts, that an unsuccessful attempt to breach the levee by explosives on June 4-6 failed to make a deep enough gap, and a second attempt on June 7 failed because the river was so low it did not reach the levee, but that a hand excavation on June 9 finally succeeded.

Following the breach, some in the Chinese government attempted to spin the event as a Japanese attack. Lary (2001) reports that the Chinese even filmed a staged Japanese attack on the levees. Durdin (1938) also quoted Chinese press accounts as blaming Japanese bombing for the breaks in the levees. The Japanese expressed concern that they would be blamed for the flood, although almost from the outset the press were unanimous that the Chinese had broken the levees. The timing and location of the break and the path of the resulting flood were so perfectly suited to impede the Japanese (even if they had little practical effect) that a Japanese role is virtually impossible. On June 17, the New York Times reported that Chinese in Shanghai (safely out of the path of the floods) were almost unanimous in expressing support for breaching the levees (New York Times, 1938b). Given the notorious inefficiency and disorganization of the Nationalist forces, it is entirely possible that some representatives were attempting to blame the Japanese while others simultaneously openly admitted that Chinese forces caused the flood.

Contemporary press accounts of the flood individually look authoritative, but when viewed collectively, generally are based on second hand reports of varying reliability, frequently confuse breach locations, and contain a good deal of exaggeration, extrapolation, and conjecture. For example, news reports of June 15 cited Japanese claims that 150,000 civilians had already drowned (New York Times, 1938a, Times of London, 1938a), but reports the next day (Robertson, 1938a, Times of London, 1938b) stated the Japanese largely retracted those estimates. Another report of June 17 (New York Times, 1938a) minimized the direct danger of the floods, and reports of June 28 (New York Times, 1938c) and July 4 (Robertson, 1938b) both doubted that there would be large loss of life. A report of June 16 (Robertson, 1938a) speaks of a “wall of water,” though that comment may be a headline composed by an editor. The New York Times report of June 15 that claimed 150,000 drowned also claimed “the river broke through a fifty-mile stretch of weakened dikes on its south bank,” clearly a conflation of reports of smaller breaches in several locations. Nevertheless many of the inconsistencies can be sorted out.

One of the most detailed press accounts (New York Times, 1938a) on June 15, 1938, mentions numerous levee breaks in the 100 or so kilometers west of Kaifeng. The report further claims that the Japanese had impressed thousands of Chinese workers into service to plug the breaches but that the floods were out of control and that Chinese forces were firing on the repair parties. A report the next day (Robertson 1938a) contained the interesting observation that water in the Huang He downstream of the breaches had reversed course and was flowing northwest. It is even possible that some breaches could have been the result of bombing or artillery fire. It appears that enough water was diverted to lower the water level in the old channel downstream from the breaks and allow backflow. By the time the levees were repaired in 1947, there was a single very wide gap carrying the entire flow (Todd, 1949). Reports on June 17 (New York Times, 1938c, Times of London, 1938c) stated that the initial gaps had coalesced into a single gap a quarter of a mile (400 meters) wide through which had diverted about 90 per cent of the Huang He’s flow, that the Japanese had decided it was hopeless to try to contain the flood, and that the former course of the river east of the breaks was almost dry.

A press report on June 16 (Robertson, 1938a) quoted Japanese sources as saying that flood waters were advancing at about five miles (8 kilometers) per hour (2.2 m/sec), and mentioned flood waters had advanced fifty miles (80 kilometers). A map published on June 19 (Todd, 1938) shows the flooded area extending about 170 miles (270 kilometers) southeast down two tributaries of the Huai River.

The flood initially spread southeast down the Go and Guo rivers to the Huai River (Figure 5). From there it flowed northeast to Hongze Lake. Hongze Lake is actually a reservoir, dammed in 1578 at a time when the Huang He flowed southeast (Xu, 1993). Its original function was to store Huai River water to be released in an artificial flood after the annual Huang He floods to flush the Huang He channel of sediment. Initially, impounding the flood in Hongze Lake served to back up the incoming flood waters, but Hongze Lake soon overflowed into the Grand Canal, which in turn overflowed until the flood waters finally reached the sea.

From a military perspective the levee breach was a marginal success. Few Japanese troops were killed by the flood although some equipment was lost. Those few Japanese or Chinese units unfortunate enough to have been caught on the wrong side of the flood were annihilated (Dorn, 1974); little quarter was given on either side in this war. Japanese forces, using first makeshift rafts (Times of London, 1938b) and later, military pontoons (Times of London, 1938c), mounted operations to rescue refugees. They also bombed marooned Chinese units (Times of London, 1938d). Meanwhile, Chinese units, wading through shallow flood waters, recaptured a number of villages (Times of London, 1938e). As late as June 21, the Japanese were still rescuing stranded troops (Times of London, 1938f). The Chinese did buy enough time to withdraw their forces and move their capital to Chungking, west of the Yangtze River gorges. However, the Japanese did push on to Zhegzhou and eventually captured Wuhan in October, 1938. Instead of advancing south down the rail line as initially intended, they merely moved west up the Yangtze to capture Wuhan and were barely even delayed beyond their original plan. The floods may have aided the Japanese nearly as much as the Chinese by allowing them to regroup in safety and move south along the east side of the flood zone without having to fear a flank attack (New York Times, 1938d). Almost immediately after the floods began, guerillas intensified their activities (Times of London, 1938f). After the floods receded and the Huang He settled into a new course, the Japanese were unable to reassert control over the flooded region, and parts of it came under guerilla control while some areas descended into banditry. Lary (2001) notes that resentment at both the Japanese and the national government made the flooded region a prime recruiting area for the Communists.

Japan in 1938 was not yet at war with the United States or Europe, and Western journalists had access to both Chinese and Japanese sources. However, nobody appears to have monitored the progress of the flood from beginning to end. Western observers viewed the flood as an immense human tragedy, but they were daily confronted with human tragedies much closer at hand, and the flood was only a side show in the much larger drama of the war. Japanese military authorities would not have viewed civilian relief as a principal part of their mission, though they did mount rescues, as noted above. They were severely taxed to maintain their offensive, and in any case their effective control amounted mostly to narrow corridors along key rail lines. Chinese central government had collapsed in the flood area, although local government often remained functional in areas bypassed or nominally occupied by the Japanese. There appear to be no contemporary overall first-hand accounts of the flood in English.

The Casualty Total

In disasters of this magnitude even the most careful statistics are only estimates. Lary (2001) reprints official figures totaling 844,000 fatalities (Figure 6). Many people certainly drowned. There is almost no high ground in the flooded area except low ridges marking abandoned river channels (Xu, 1993). Artificial refuges like railroad embankments and buildings might serve, except that the mud brick used for building in the plain disintegrates quickly in floods. Since the floods permanently rerouted the Huang He, the flooding would persist for a long time in some places. After days of isolation, deaths from hunger and exposure start to occur among the most vulnerable segments of the population: the old, the sick, or the very young, while disintegration of mud brick causes many refuges to collapse. Water borne diseases would certainly have been a major cause of death since there would be a complete lack of potable water in the flood zone. The floods came at harvest time, destroying crops and burying them under silt. Worse yet, the breach came just before the start of the normal Huang He flood season, so that many localities would get a brief respite, only to be flooded anew, while areas that escaped the initial flooding would be flooded later. Lary (2001) is probably correct in noting “The Chinese strategists seem not to have given thought to how far the flood would go, or how devastating it would be.”

The magnitude of the disaster was due to the fact that the population at risk was enormous. The most graphic way to get an appreciation of the number of people affected by the flood is to view imagery of the region. The average distance between villages today is less than a kilometer. Even allowing for the fact that China’s population is roughly four times what it was in 1938, it is not hard to imagine how a massive flood in such an area could cause great casualties. The huge population may well account for the radically different perspectives of journalists, some predicting a huge death toll while others stated that most people escaped. If 20 million people were in the danger zone and a million died, that would be a vast human catastrophe, yet it would also mean that 95 per cent of the people survived, a macabre illustration of how perspective matters.

Graham (1999) compared several formulas for predicting loss of life from flooding, especially dam failure. He identified three critical factors: violence of the flood, length of advance warning and level of understanding of the risk downstream. Graham grouped flood severity into three classes: high, where everything is swept away, medium, where trees and some buildings remain, and low, where most buildings survive. Over most of the 1938 flood zone, we can estimate that flood severity was medium in most of the deepest-flooded areas and low in shallowly flooded areas. We can presume risk understanding downstream was low. If there is no warning, as was the case for most of the flooded area in 1938, Wright suggested assuming a fatality rate of 15 per cent for medium severity floods and one per cent for low severity. Fatality rates suggested for predicting a range of scenarios range from 3 to 35 per cent in medium severity areas and 0 to 2 per cent in low severity areas. Lary (2001) presented a table of casualties by county for the 1938 flood area. Fatality rates in Henan Province, closest to the breach, were about ten per cent of the at-risk population with an actual range by county of 2 to 25 per cent. In Jiangsu, on the coast and farthest from the breach, fatalities were about 3 per cent of the at risk population, with fatalities by county ranging from less than 1 to 6 per cent. Thus the fatality rates are quite consistent with Wright’s predicted rates, and generally on the low side, even though his figures were not designed to be used on floods of this magnitude. The total population of the counties tabulated by Lary is about 13 million, so 800,000 deaths would amount to a fatality rate of 6 per cent.

A perspective on the casualty toll is provided by a more recent and far better documented flood, the Banqiao flood of 1975 (Si, 1998). This flood was precipitated by an extreme rainfall event that resulted in the failure of 62 dams along the Ru River, a tributary of the Huai (although uninvolved in the flooding of 1938). The flood was more abrupt, violent and deeper than the 1938 inundation, but warnings were attempted (although the storm disrupted communications) and military personnel were deployed immediately afterward for relief operations. Nevertheless, nine days after the flood a million people were still isolated by flood waters. The fatalities have been estimated at 26,000 from drowning and 145,000 from epidemics and famine, making it the worst dam failure in history. Six million buildings were damaged or destroyed, and 11 million people affected by the flooding. This disaster makes the estimated casualties of 1938 plausible.

Floods along the Huang He are so devastating because of the large at-risk population and the flat terrain offering no refuge. Other regions with a similar combination of flat land, poor infrastructure for warning and evacuation, and extremely dense at-risk population include Bangladesh and the Irrawaddy Delta of Myanmar. Floods in both areas due to cyclone storm surges have resulted in disasters costing hundreds of thousands of lives. The Mekong Delta, though it has not experienced such a large flood in modern times, may offer similar potential for disaster. The grim equation is: large at risk population plus flat terrain plus poor infrastructure and communications equals huge death tolls from flooding. Restoring the Huang He

Although relief efforts in China following World War II were plagued by rampant corruption at all levels (Kerr, 1965), it appears that repairing the Huang He levees was a high enough national priority to be accomplished reasonably quickly. Fortunately, O. J. Todd, an engineer with long experience in China, was a leader in the diversion project and left a detailed and richly illustrated account. Since the repairs were described by a highly qualified observer, it is possible to say much more about the restoration of the Huang He to its original course than it is about the original diversion. The summary below is largely based on Todd’s (1949) account.

Restoring the flow of the river was of high priority because the diversion had taken large areas of agricultural land out of use in the inundated area, while simultaneously robbing the areas along the former course of the river of water. Restoring Chinese agriculture to productivity would not only benefit China but remove a great burden from relief agencies.

In 1944, China asked the newly formed United Nations for assistance in repairing the levees. The first step in the repair was an inspection of the levees on the former river course downstream from the break. The inspectors found, to their considerable relief, that the levees were generally sound, so that repairs on the main break could begin promptly, while damage downstream was simultaneously repaired. (Todd makes no mention of other levee breaks, although his map of 1949, presented in modified form as Figure 5, clearly implies at least one other.) By the time repairs began, the river had entrenched channels along its new course, so that of the original 1600 meter breach, 1000 meters was dry. The repair strategy called for an earthen embankment to repair the dry gap and a rock fill dam for the 600 meters occupied by the diverted river channel. An additional problem was that the former bed of the river had been colonized by squatter settlements of refugees and peasants seeking additional land. These settlers were understandably upset about being dislocated again. East-central China and Shandong had been hotbeds of Communist guerilla activity during the war, and agreements with Communist forces had to be worked out for compensation to displaced squatters and for repairs to levees in Communist controlled area (Lieberman, 1946; Durdin, 1947). Shipments of supplies were sometimes delayed by civil warfare between Communist and Nationalist forces.

To assist the diversion of the Huang He to its former course, 40,000 laborers dug channels in the dry riverbed. The channels were 3 to 16 kilometers long, 15 to 20 meters wide and 1.5 to two meters deep. The loess substrate was a blessing since there were no rocks, roots, or buried stumps to impede digging. Up to 350,000 people were engaged in the repair work at peak periods. Workers were paid with daily rations of about 1100 grams of flour. State of the Literature and Questions for Future Research

Thanks to the internet and recent historical research using Chinese sources, the availability of information on the 1938 flood has improved dramatically. Nevertheless, some aspects of the disaster are known in considerable detail, while others are far sketchier.

The military situation in June, 1938 and the strategic situation that inspired the decision to breach the levees are well documented. Lary (2001) furnished the most detailed data currently available on casualties, and Lary and Mackinnon (2001) have documented the effects of the Second Sino-Japanese War on Chinese politics and subsequent history. Although the 1938 flood is peripheral to their main interests, they have shed valuable light on it. Todd’s (1949) paper documents the repair of the levees and restoration of the former course of the Huang He in considerable detail.

The location of the dike breach is known, but the exact events surrounding the breach are not well documented, and different sources offer conflicting accounts. The overall extent of the flood is reasonably well known, but the exact extent, the time of flood crest at various locations, and the lateral spread of the flood waters with time are not. There are as no readily accessible detailed accounts of the flood at specific localities, nor eyewitness accounts other than by journalists. By now, 70 years after the flood, there are few surviving witnesses who were adults at the time, and if they are not located and interviewed soon their accounts will be permanently lost. By July 1938, press accounts indicated that flood waters were receding, and the event disappeared from world awareness. Thus there are no detailed accounts of the recession of the flood, accounts of famine or epidemic afterward, and no accounts of how survivors coped and rebuilt in the wake of the flood. Data on weather and the effects of natural rainfall and later flooding on the disaster are also not readily available.


Figure Captions

Figure 1.

Relief of northeastern China. Shorelines and hydrography for all figures are from the CIA World Data Bank II (Pape, 2004). Relief is generalized from GTOPO30 (U.S. Geological Survey, 1996). Geographic names from Times Atlas of the World (Times Books and Collins Bartholomew Ltd., 2007).

Figure 2.

Loess deposits of the North China Plain (after Zhao, 1986), with principal geographic localities also labeled.

Figure 3.

Diversions of the Huang He (after Zhao, 1986). Also shown is the Ru River and Banqiao Dam.

Figure 4.

Military situation leading up to the levee breach. After a bitter campaign to capture Xuzhou, the Japanese pressed westward to capture the rail junction at Zhengzhou, with the objective of driving south to the temporary capital at Wuhan. As they approached Kaifeng, the Chinese breached the levees. Major water bodies are horizontally hatched; Japanese occupied areas are vertically hatched. Railroad and military data based on Dorn (1974).

Figure 5.

Extent of flooding as shown by Todd (1949, with permission of the American Geographical Society). In recognition of Todd’s contribution to the historical record and his craftsmanlike drafting, Todd’s map is presented as originally published apart from the addition of a latitude and longitude grid and modernization of key geographic names.

Figure 6.

Fatalities plus refugees as percent of the population, based on data of Lary (2001). Rivers are parallel lines. Heavy solid lines are province boundaries. Fine solid lines are boundaries of counties (CIESIN, 1990). Present boundaries are shown and may not coincide with 1938 boundaries. The data show generally decreasing impact on the populace downstream. Unfortunately, the data for Anhui do not separate refugees and fatalities.