Contr llingHurricanesCan hurricanes and other severe tropicalstorms be moderated or deflected?By Ross N. HoffmanMASSIVE HURRICANE with awell-developed eye, as seenfrom the space shuttleAtlantis in November 1994.COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.
Every year huge rotating stormspacking winds greater than 74 miles per hour sweepCOURTESY OF NASA/CORBISacross tropical seasand onto shorelines—often devastating large swaths of territory.When these roiling tempests—called hurricanes in the Atlanticand the eastern Pacific oceans, typhoons in the westernPacific and cyclones in the Indian Ocean—strike heavily populatedareas, they can kill thousands and cause billions of dollarsof property damage. And nothing, absolutely nothing, standsin their way.But must these fearful forces of nature be forever beyond ourcontrol? My research colleagues and I think not. Our team isinvestigating how we might learn to nudge hurricanes onto morebenign paths or otherwise defuse them. Although this bold goalprobably lies decades in the future, we think our results showthat it is not too early to study the possibilities.To even consider controlling hurricanes, researchers willneed to be able to predict a storm’s course extremely accurately,to identify the physical changes (such as alterations in air temperature)that would influence its behavior, and to find ways toeffect those changes. This work is in its infancy, but successfulcomputer simulations of hurricanes carried out during the pastfew years suggest that modification could one day be feasible.What is more, it turns out the very thing that makes forecastingany weather difficult—the atmosphere’s extreme sensitivity tosmall stimuli—may well be the key to achieving the control weseek. Our first attempt at influencing the course of a simulatedhurricane by making minor changes to the storm’s initial state,for example, proved remarkably successful, and the subsequentresults have continued to look favorable, too.o.To see why hurricanes and other severe tropical storms maybe susceptible to human intervention, one must understand theirnature and origins [ see box on next two pages]. Hurricanesgrow as clusters of thunderstorms over the tropical oceans. Lowlatitudeseas continuously provide heat and moisture to the atmosphere,producing warm, humid air above the sea surface.When this air rises, the water vapor in it condenses to formclouds and precipitation. Condensation releases heat—the solarheat it took to evaporate the water at the ocean surface. Thisso-called latent heat of condensation makes the air more buoy-COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.SCIENTIFIC AMERICAN 69
ANATOMY OF A HURRICANESome scientists believe that they may be able to weaken or movehurricanes onto less dangerous tracks by altering the initial physicalconditions (the air temperature or humidity, for example) in thecenter of the storm or even in the surrounding areas. To succeed,they need to make accurate and detailed forecasts of hurricanes.Here are the outlines of how these powerful storms arise.Stormrotation 6Developing thundercloudsIncoming air3Rising air5Tropical oceanRainRising airHurricanes start to form when tropical oceans release heat and water into theatmosphere, producing large amounts of warm, humid air above the surface(1). Warm air rises, and as it does so, the water vapor in it condenses to formclouds and rain (2). This condensation produces heat, causing air in thedeveloping thunderclouds to climb still farther (3).21CondensationLow-pressure zone 4Converging airThe release of heat above the tropical seas creates a surface low-pressurezone, where additional warm, moist air from the outer perimeter converges(4). This continuous movement into the burgeoning thunderstorm shifts hugeamounts of heat, air and water skyward (5). This upward transfer and releaseof heat further enhance the convergence of surrounding air toward thegrowing storm center, which starts to circulate under the influence of theearth’s rotation (6). The process continues apace, strengthening andorganizing the storm.ant, causing it to ascend still higher in aself-reinforcing feedback process. Eventually,the tropical depression begins to organizeand strengthen, forming the familiareye—the calm central hub aroundwhich a hurricane spins. On reachingland, the hurricane’s sustaining source ofwarm water is cut off, which leads to thestorm’s rapid weakening.Overview/Taming Hurricanes■ Meteorological researchers are simulating past hurricanes using sophisticatedweather-forecasting models that closely reproduce the complex internalprocesses crucial to the development and evolution of severe tropical storms.■ The work confirms that these massive, chaotic systems are susceptible tominor changes in their initial conditions—for instance, the air temperature andhumidity near the center of the storm and in the surrounding regions.■ Using complex mathematical optimization techniques, the researchers arelearning what modifications to a hurricane could weaken its winds or divert itfrom populated areas.■ If these theoretical studies are ultimately successful, they should pointthe way toward practical methods for intervening in the life cycle of hurricanesto protect life and property.Dreams of ControlBECAUSE A HURRICANE draws muchof its energy from heat released when watervapor over the ocean condenses intoclouds and rain, the first researchers todream of taming these unruly giants focusedon trying to alter the condensationprocess using cloud-seeding techniques—then the only practical way to try to affectweather. In the early 1960s a U.S.government-appointed scientific advisorypanel named Project Stormfury performeda series of courageous (or perhapsfoolhardy) experiments to determinewhether that approach might work.Project Stormfury aimed to slow thedevelopment of a hurricane by augmentingprecipitation in the first rain band outsidethe eye wall—the ring of clouds andhigh winds that encircle the eye [see “Experimentsin Hurricane Modification,” byR. H. Simpson and Joanne S. Malkus;Scientific American, December 1964].They attempted to accomplish this goalby seeding the clouds there with silver iodideparticles dispersed by aircraft, whichwould serve as nuclei for the formation ofice from water vapor that had been supercooledafter rising to the highest, coldestreaches of the storm. If all went as envisioned,the clouds would grow morequickly, consuming the supplies of warm,moist air near the ocean surface, thus replacingthe old eye wall. This processDAVID FIERSTEIN70 SCIENTIFIC AMERICAN OCTOBER 2004COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.
Hurricane911107Eye8EyewallDescending airAs the storm intensifies, an eye—a calm, low-pressure hub—typically forms(7). The eye is encircled by a ring of clouds and high winds called the eye wall(8). The storm has become a hurricane. At the same time, the rising air, nowheated and having lost much of its moisture, can rise no further, because thestratosphere acts like a lid above the hurricane. Some of this dry air falls intothe eye (9) and between the cloud bands (10), while the remainder spiralsaway from the storm center and descends (11). Meanwhile large-scale aircurrents nearby steer the hurricane along its path.would then expand the radius of the eye,lessening the hurricane’s intensity in amanner akin to a spinning skater who extendsher arms to slow down.The Stormfury results were ambiguousat best. Meteorologists today do not expectthis particular application of cloud seedingto be effective in hurricanes because,contrary to the early beliefs, the stormscontain little supercooled water vapor.Chaotic WeatherOUR CURRENT STUDIES grew out ofan intuition I had 30 years ago when Iwas a graduate student learning aboutchaos theory. A chaotic system is one thatappears to behave randomly but is, infact, governed by rules. It is also highlysensitive to initial conditions, so thatseemingly insignificant, arbitrary inputscan have profound effects that lead quicklyto unpredictable consequences. In thecase of hurricanes, small changes in suchfeatures as the ocean’s temperature, thelocation of the large-scale wind currents(which drive the storms’ movements), oreven the shape of the rain clouds spinningaround the eye can strongly influence ahurricane’s potential path and power.The atmosphere’s great sensitivity totiny influences—and the rapid compoundingof small errors in weather-forecastingmodels—is what makes longrangeforecasting (more than five days inadvance) so difficult. But this sensitivityalso made me wonder whether slight,purposely applied inputs to a hurricanemight generate powerful effects thatcould influence the storms, whether bysteering them away from population centersor by reducing their wind speeds.I was not able to pursue those ideasback then, but in the past decade computersimulation and remote-sensingtechnologies have advanced enough to renewmy interest in large-scale weathercontrol. With funding support from theNASA Institute for Advanced Concepts,my co-workers and I at Atmospheric andEnvironmental Research (AER), an R&Dconsulting firm, are employing detailedcomputer models of hurricanes to try toidentify the kinds of actions that mighteventually be attempted in the real world.In particular, we use weather-forecastingtechnology to simulate the behavior ofpast hurricanes and then test the effectsof various interventions by observingchanges in the modeled storms.Modeling ChaosEVEN TODAY’S BEST weather predictioncomputer models leave much to bedesired when it comes to forecasting, butwith effort they can be useful for modelingthese storms. The models depend onnumerical methods that simulate astorm’s complex development process bycomputing the estimated atmosphericconditions in brief, successive time steps.Numerical weather prediction calculationsare based on the assumption thatwithin the atmosphere there can be nocreation or destruction of mass, energy,momentum and moisture. In a fluid syswww.sciam.comSCIENTIFIC AMERICAN 71COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.
tem like a hurricane, these conservedquantities are carried along with thestorm’s flow. Near the boundaries ormargins of the system, however, thingsget more complicated. At the sea surface,for example, our simulations account forthe atmosphere gaining or losing the fourbasic conserved quantities.Modelers define the atmosphericstate as a complete specification of themeasurable physical variables, includingpressure, temperature, relative humidity,and wind speed and direction. Thesequantities correspond to the conservedphysical properties on which the computersimulations are based. In mostweather models these observable variablesare defined on a three-dimensionalgrid representation of the atmosphere, soone can plot a map of each property foreach elevation. Modelers call the collectionof values of all the variables at all thegrid points the model state.To generate a forecast, a numericalweather prediction model repeatedly advancesthe model state from one instantthrough a small time step (a few secondsto a few minutes depending on the scalesof motion resolved by the model). It calculatesthe effects during each time stepof winds carrying along the various atmosphericproperties and of the processesof evaporation, rainfall, surfacefriction, infrared cooling and solar heatingthat occur in the area of interest.Unfortunately, meteorological forecastsare imperfect. In the first place, thebeginning model state is always incompleteand inexact. Initial states for hurricanesare particularly difficult to defineTHE AUTHORbecause direct observations are few anddifficult to make. Yet we do know fromsatellite cloud images that hurricaneshave complex and detailed structures. Althoughthese cloud images are potentiallyuseful, we need to know much more.Second, even with a perfect initial state,computer models of severe tropicalstorms are themselves prone to error.The atmosphere, for example, is modeledonly at a grid of points. Features smallerthan the grid length, the distance betweentwo neighboring grid points, cannot behandled correctly. Without very high resolution,a hurricane’s structure near theeye wall—its most important feature—issmoothed out and the details are unclear.In addition, the models, just like the atmospherethey simulate, behave in achaotic fashion, and inaccuracies fromboth these error sources grow rapidly asthe forecast computations proceed.Despite its limitations, this technologyis still valuable for our purposes. Wehave modified for our experiments a highlyeffective forecast initialization systemcalled four-dimensional variational dataassimilation (4DVAR). The fourth dimensionto which the name refers is time.Researchers at the European Center forMedium-Range Weather Forecasts, oneof the world’s premier meteorologicalcenters, use this sophisticated techniqueto predict the weather every day. To makebest use of all the observations collectedby satellites, ships, buoys and airbornesensors before the forecast begins,4DVAR combines these measurementswith an educated first guess of the initialatmospheric state—a process called dataROSS N. HOFFMAN is a principal scientist and vice president for research and developmentat Atmospheric and Environmental Research (AER) in Lexington, Mass. His primary areasof interest cover objective analysis and data assimilation methods, atmospheric dynamics,climate theory and atmospheric radiation. He has been a member of several NASA scienceteams and was a member of the National Research Council Committee on the Statusand Future Directions in U.S. Weather Modification Research and Operations. Hoffman wouldlike to thank the NASA Institute for Advanced Concepts for supporting his work as well as hisAER colleagues, particularly John Henderson, for their efforts in this research.assimilation. This first guess is usually asix-hour forecast valid at the time of theoriginal observations. Note that 4DVARaccounts for each observation just whenit was taken rather than grouping themacross a time interval of several hours.The result of merging the observationaldata and the first guess is then used to initiatethe subsequent six-hour forecast.In theory, data assimilation producesThe altered version of Hurricane Iniki veered off, so thatKauai escaped the storm’s most damaging winds.an optimal approximation of the weatherin which the fit of the model’s representationto the observations is balancedagainst its fit to the first guess. Althoughthe statistical theory for this problem isclear, the assumptions and informationneeded for its proper application are onlyapproximate. As a result, practical dataassimilation is part art and part science.Specifically, 4DVAR finds the atmosphericstate that satisfies the model equationsand that is also close to both the firstguess and the real-world observations. Itaccomplishes this difficult task by backadjustingthe original model state at thestart of the six-hour interval according tothe difference between observations andmodel simulation made during that period.In particular, 4DVAR employs thesedifferences to calculate the model’s sensitivity—howsmall changes in each of theparameters would affect the degree towhich the simulation fit the observations.This computation, using the so-called adjointmodel, runs backward in time overthe six-hour interval. An optimizationprogram then chooses the best adjustmentsto make to the original model stateto achieve a simulation that most closelymatches the progress of the actual hurricaneduring the six-hour period.Because this adjustment is made usingan approximation of the model equations,the entire process—the simulation,the comparisons, the adjoint model andthe optimization—must be repeatedagain and again to fine-tune the results.When the process is complete, the conditionsof the simulation at the end of the72 SCIENTIFIC AMERICAN OCTOBER 2004COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.
six-hour period then supplies the firstguess for the next six-hour interval.After simulating a hurricane that occurredin the past, we can then changeone or more of its characteristics at anygiven time and examine the effects ofthese perturbations. It turns out thatmost such alterations simply die out.Only interventions with special characteristics—aparticular pattern or structurethat induces self-reinforcement—will developsufficiently to have a major effecton a storm. To get an idea of what thismeans, think of a pair of tuning forks,one vibrating, the other stationary. If theforks are tuned to different frequencies,the second fork does not move, despitebeing struck repeatedly by sound wavesemitted by the first. But if the devicesshare the same frequency, the secondfork will respond in a resonant mannerand vibrate sympathetically. In an analogousfashion, our challenge is to findjust the right stimuli—changes to the hurricane—thatwill yield a robust responsethat leads to the desired results.Calming the TempestTO EXPLORE WHETHER the sensitivityof the atmospheric system could beCONTROL OF SIMULATED HURRICANESResearchers are using computer models to simulate twodestructive 1992 hurricanes, Iniki and Andrew. The colorsrepresent wind-velocity categories, whereas blackcontour lines indicate gales of 56 miles per hour, generallythe lowest wind speed that produces damage.In the simulations of Iniki (right), the original track ofthe eye (black dotted line) takes the storm’s high windsonto the Hawaiian island of Kauai. But when several of themodel’s initial conditions, including its temperature andhumidity at various points, were altered slightly, thesimulated storm track (red dotted line) veered to the west ofKauai, passing over a target location some 60 miles away. Itthen continued northward, moving farther west of the island.The maps of the seas off Florida and the Bahamasbelow depict simulations of Andrew in its unaltered state(left) and in an artificially perturbed (right) form. Althoughdamaging winds remain in the controlled case, maximumvelocities have been reduced significantly, thus calminga Category 3 hurricane to a much milder Category 1 state.------------------Targetlocation-x-KauaiHawaiianActual storm trackModified storm track-Islands56 mph--Originalhurricane-----------------------------Wind speed(miles per hour)Category 4hurricane- 130Category 3hurricane-110Category 2hurricane-95Category 1hurricane-----73----Florida------Florida56 mph 56 mph--Tropicalstorm----LUCY READING----CubaBahamas------------CubaBahamas---------38Tropicaldepression-27exploited to modify atmospheric phenomenaas powerful as hurricanes, ourresearch group at AER conducted computersimulation experiments for twohurricanes that occurred in 1992. WhenHurricane Iniki passed over the Hawaiianisland of Kauai in September of that year,several people died, property damage wasenormous and entire forests were leveled.Hurricane Andrew, which struck Floridajust south of Miami the month before, leftthe region devastated.Surprisingly, given the imperfectionsof existing forecasting technologies, ourfirst simulation experiment was an imwww.sciam.comSCIENTIFIC AMERICAN 73COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.
mediate success. To alter the path of Iniki,we first chose where we wanted thestorm to end up after six hours—about60 miles west of the expected track. Thenwe used this target to create artificial observationsand fed these into 4DVAR.We set the computer to calculate thesmallest change to the initial set of thehurricane’s key defining properties thatwould yield a track leading to the targetlocation. In this early experiment we permittedany kind of possible artificial alterationto the storm system to takeplace.The most significant modificationsproved to be in the starting temperaturesand winds. Typical temperature adjustmentsacross the grid were mere tenths ofa degree, but the most notable change—an increase of nearly two degrees Celsius—occurredin the lowest model layerwest of the storm center. The calculationsyielded wind-speed alterations of two orthree miles per hour. In a few locations,Computer simulations of hurricanes indicate that changes inprecipitation, evaporation and air temperature could alter astorm’s path or weaken its winds. Intervention might takevarious forms: Carefully targeted aerial cloud seeding withsilver iodide or other precipitation-inducing materials mightserve to rob a hurricane’s violent eye wall—the key feature ofa severe tropical storm—of the water it needs to grow andintensify (left). Biodegradable oil could be distributed acrossthe sea surface in the path of a hurricaneto limit evaporation—the source of athough, the velocities changed by as muchas 20 mph because of minor redirectionsof the winds near the storm’s center.Although the original and altered versionsof Hurricane Iniki looked nearlyidentical in structure, the changes in thekey variables were large enough that thelatter veered off to the west for the firstsix hours of the simulation and then traveleddue north, so that Kauai escaped thestorm’s most damaging winds. The relativelysmall, artificial alterations to thestorm’s initial conditions had propagatedthrough the complex set of nonlinearequations that simulated the storm to resultin the desired relocation after sixhours. This run gave us confidence thatwe were on the right path to determiningthe changes needed to modify real hurricanes.For the subsequent hurricane simulationtrials, our team used higher gridresolutions to model the hurricane andset 4DVAR to the goal of minimizingproperty damage.HURRICANE INTERVENTIONIn one experiment using the modifiedcode, we calculated the temperature incrementsneeded to limit the surface winddamage caused by Hurricane Andrew asit hit the Florida coast. Our goal was tokeep the initial temperature perturbationto a minimum (to make it as easy to accomplishas possible in real life) and tocurtail the most destructive winds over thelast two hours of the first six-hour interval.In this trial, 4DVAR determined thatthe best way to limit wind damage wouldbe to make the greatest modificationsto the beginning temperature near thestorm’s eye. Here the simulation producedchanges as large as two or three degrees Cat a few locations. Smaller temperature alterations(less than 0.5 degree C) extendedout 500 to 600 miles from the eye.These perturbations feature a wavelikepattern of alternating rings of heating andcooling centered on the hurricane. Althoughonly temperature had beenchanged at the start, all key variables werestorm’s energy (center). Future earth-orbiting solarpower stations, which could employ large mirrors tofocus the sun’s rays and panels of photovoltaic cells toharvest the energy for transfer down to the earth,might be used to beam microwaves tuned to beabsorbed by water vapor molecules in the storm or inits surroundings (right). The microwaves would causethe water molecules to vibrate and heat thesurrounding air, thus causing the hurricane to weakenor move in a desired direction.OrbitalpowerstationPrecipitationinducingmaterialsCloud-seeding aircraftHurricaneHeatedwatervaporHurricane pathReduced evaporationBiodegradable oil slickDAVID FIERSTEIN74 SCIENTIFIC AMERICAN OCTOBER 2004COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.
soon affected. In the case of the originalsimulated hurricane, damaging winds(greater than about 56 mph) coveredpopulated areas in South Florida by theend of six hours, but in the altered modelrun, they did not do so.As a test of the robustness of these results,we applied the same perturbation toa more sophisticated, higher-resolutionversion of the model. We obtained verysimilar results, which show that our experimentsare reasonably insensitive toour particular choice of model configuration.After six hours, however, damagingwinds reappeared in the altered simulation,so additional interventions wouldhave been required to keep South Floridasafe. Indeed, it looks as if a series ofplanned disturbances would be required tocontrol a hurricane for any length of time.Who Can Stop the Rain?IF IT IS TRUE, as our results suggest,that small changes in the temperature inand around a hurricane can shift its pathin a predictable direction or slow itswinds, the question becomes, How cansuch perturbations be achieved? No one,of course, can alter the temperaturethroughout something as large as a hurricaneinstantaneously. It might be possible,however, to heat the air around ahurricane and thus adjust the temperatureover time.Our team plans to conduct experimentsin which we will calculate the precisepattern and strength of atmosphericheating needed to moderate hurricane intensityor alter its track. Undoubtedly, theenergy required to do so would be huge,but an array of earth-orbiting solar powerstations could eventually be used tosupply sufficient energy. These powergeneratingsatellites might use giant mirrorsto focus sunlight on solar cells andthen beam the collected energy down tomicrowave receivers on the ground. Currentdesigns for space solar power stationswould radiate microwaves at frequenciesthat pass through the atmospherewithout heating it, so as to notwaste energy. For weather control, however,tuning the microwave downlink tofrequencies better absorbed by water vaporcould heat different levels in the atmosphereas desired. Because raindropsstrongly absorb microwaves, parts of thehurricane inside and beneath rain cloudswould be shielded and so could not beheated in this way.In our previous experiments, 4DVARdetermined large temperature changesjust where microwave heating could notwork, so we ran an experiment in whichwe forced the temperature in the center ofthe hurricane to remain constant duringour calculation of the optimal perturbations.The final results resembled those ofthe original, but to compensate for makingno initial temperature changes in thestorm center the remaining temperaturechanges had to be larger. Notably, temperaturechanges developed rapidly nearthe storm center during the simulation.Another potential method to modifysevere tropical storms would be to directlylimit the availability of energy bycoating the ocean surface with a thin filmof a biodegradable oil that slows evaporation.Hurricanes might also be influencedby introducing gradual modificationsdays in advance of their approachand thousands of miles away from theireventual targets. By altering air pressure,these efforts might stimulate changes inthe large-scale wind patterns at the jetstreamlevel, which can have major effectson a hurricane’s intensity and track.Further, it is possible that relatively minoralterations to our normal activities—such as directing aircraft flight plans toprecisely position contrails and thus increasecloud cover or varying crop irrigationpractices to enhance or decreaseevaporation—might generate the appropriatestarting alterations.What if Control Works?IF METEOROLOGICAL control doesturn out to work at some point in the future,it would raise serious politicalproblems. What if intervention causes ahurricane to damage another country’sterritory? And, although the use ofweather modification as a weapon wasSmall changes can strongly influencea hurricane’s potential path and power.MORE TO EXPLOREbanned by a United Nations Conventionin the late 1970s, some countries mightbe tempted.Before those kinds of concerns arise,however, our methods would need to beproved on atmospheric phenomena otherthan hurricanes. In fact, we believe ourtechniques should first be tried out in aneffort to enhance rainfall. This approachcould then serve as a test bed for our conceptsin a relatively small region thatcould be instrumented densely with sensors.For such reduced size scales, perturbationscould be introduced from aircraftor from the ground. If our understandingof cloud physics, computer simulation ofclouds and data assimilation techniquesadvance as quickly as we hope, thesemodest trials could be instituted in perhaps10 to 20 years. With success there,larger-scale weather control using spacebasedheating may become a reasonablegoal that nations around the globe couldagree to pursue.The Rise and Fall of Weather Modification: Changes in American Attitudes toward Technology,Nature, and Society. Chunglin Kwa in Changing the Atmosphere: Expert Knowledge andEnvironmental Governance. Edited by Clark A. Miller and Paul N. Edwards. MIT Press, 2001.Controlling the Global Weather. Ross N. Hoffman in Bulletin of the American Meteorological Society,Vol. 83, No. 2, pages 241–248; February 2002. Available at http://ams.allenpress.comCritical Issues in Weather Modification Research. Michael Garstang et al., National ResearchCouncil of the National Academies of Sciences. National Academies Press, Washington, D.C., 2003.Available at www.nap.edu/books/0309090539/htmlNOAA’s Hurricane Research Division: www.aoml.noaa.gov/hrd/tcfaqRoss N. Hoffman’s technical presentations on weather modification can be found atwww.niac.usra.edu/studies/www.sciam.com SCIENTIFIC AMERICAN 75COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.