I was working the forecast desk at Scott Air Force Base when I heard from the duty officer that an incoming C-5 was low on fuel and needed to land. My name is Slade, Master Sergeant Jack Slade, and I'm a forecaster.

The time was 0900Z, and there was precip in the forecast. While light rain wouldn't be a problem, heavy rain or snow would drop conditions below minimums. We had a layer of dry air near the surface, conducive to coolingperhaps enough to allow a change from rain over to snow. And while no precip had fallen yet, the St. Louis radar was painting echoes over central Missouri.

I had to decide which would get here first: the C-5 or the precip? And what kind of precip were we going to get?

I made a quick review of my forecast.

The 0000Z surface analysis suggested that the Scott AFB region was under the influence of an anticyclone.

In fact, we had a sizeable ridge at 500 hPa over the region.

Closer inspection of the lower troposphere revealed easterly flow and dry air with wet bulb temperatures around 0°C at the surface in the rear of a retreating high, a feature that evolved into a high amplitude ridge at 850 hPa.

And we had a good deal of moisture upstream, a pattern reflected at 700 and 500 hPa as well.

Looking at the models, the vertical motion and moisture fields revealed the likelihood of precipitation with good moisture and plenty of lift over Missouri.

I also looked at the 700 hPa frontogenesis. I wanted to see whether we had much potential for banding. There was a region of enhanced values running northwest-to-southeast over eastern Missouri, with Scott AFB near the eastern edge. Things were looking good for some banded precip in my Area of Responsibility.

But to decide what to tell the C-5, I needed a clear distinction between a forecast of rain or snow.

I pulled up the model soundings for the Base and followed their progress from initialization at 0000Z through the current time and out to 1800Z. When I looped the soundings, they looked like an inverted zipper. The temperature and dewpoint traces closed in on the wet bulb trace from aloft and moved downward toward the surface. More importantly, the wet bulb temperature throughout that layer was close to 0°C. Near the end of the loop, the near-surface layer was cooling from a few degrees above freezing down to the 0°C isotherm.

The time was 0935Z, and the flight was due in 30 minutes. The 0900Z obs from Scott showed visibility of 7 miles with a broken ceiling of 10,000 feet. I took one more look at the St. Louis radar, then I picked up the phone to call the duty officer.

What do you think?
Would conditions remain above minimums for another half hour?
Or should Master Sergeant Slade revise his forecast and redirect the C-5?

Mesoscale precipitation bands present a significant challenge to modern operational weather forecasters. Basic meteorological training gives only brief exposure to this problem, and NWP models have difficulties generating accurate forecasts of narrow precipitation bands. This module will explore the synoptic environments and mesoscale processes that lead to banded precipitation.

The term "precipitation band" has been defined as either an instantaneous feature on a radar image or as an elongated region of accumulated precipitation. Novak and his colleagues (2002) defined a single precipitation band to be at least 250 km in length and no more than 100 km in width. Their definition relies on thresholds for radar reflectivity maintained over a minimum time period.

We will tend toward the instantaneous feature definition in this discussion. In either case, significant precipitation bands of either type in the midlatitudes typically occur in conjunction with an extratropical cyclone. In particular, the frontal zones of extratropical cyclones are often closely associated with many precipitation bands. Therefore, it is necessary to briefly review cyclone structures, especially as they relate to the generation and maintenance of precipitation bands.

These dimensions can also apply to the accumulated, storm-total precipitation generated by a single storm. For example, this plot shows a band of snow that fell across southern Missouri and Illinois on 13-14 March 1999. However, such elongated regions of accumulated precipitation do not always result from an organized mesoscale precipitation band.

In a similar fashion, it is important to realize that instantaneous precipitation bands may or may not result in a banded accumulation of precipitation. The pattern of accumulated precipitation depends on the motion of the instantaneous band relative to its orientation. This radar loop shows a squall line moving perpendicular to its length. Note that as the squall line sweeps from west to east, individual cells move north-northeast, as determined by the wind-shear vector.

To learn more about squall lines and their motion, see the module Severe Convection II: Mesoscale Convective Systems (www.meted.ucar.edu/mesoprim/severe2).

The passage of the squall line resulted in a broad band of accumulated precipitation, as shown in this plot of storm-total precipitation.

Conversely, when an instantaneous band moves parallel to its length, narrow bands of accumulated precipitation may result.

The passage of single cells of intense precipitation may also result in narrow bands of accumulated precipitation. The north-northeast bands within the storm-total precipitation plot reflect the passage of individual cells within the squall line.

Before we delve into precipitation banding, lets revisit briefly those components necessary for rain or snow formation, chiefly moistureand lift. While these two ingredients are sufficient to generate rain or snow, instability must also be present if the precipitation is to be convective.

When forecasting precipitation, it is best to begin with an assessment of the moisture available in the atmosphere. The atmosphere may well be rising, but in the absence of moisture for cloud development, all you will have is dry ascent, such as a thermal.

This plot comes from the scenario that started the module. It shows mean relative humidity in the 1000-500 hPa layer. Note the values exceeding 90% across eastern Nebraska and western Iowa. Considerable moisture is assumed to be present when the relative humidity exceeds 80% at a given level. Nucleation begins in the atmosphere at around this value.

With adequate moisture, we need rising motion to generate clouds and precipitation. After all, the atmosphere may be saturated, but in the absence of ascent, all you may have is a cloud or fog.

We can look at the output from an NWP model and assess the vertical motion fields to find ascent and descent. For example, this plot shows vertical velocity, expressed as Omega, at 700 hPa for the scenario case that started this module. Negative values indicate upward motion. Notice the region of ascending flow across central Missouri.

Once we have defined regions of sufficient moisture and lift, we may look more closely at these areas in order to assess the likelihood of mesoscale banding. Precipitation banding results from mechanisms that organize either the lift or the moisture into narrow regions.

Models of midlatitude cyclones provide a good place to start a discussion of banded precipitation. The Norwegian Cyclone Model, for example, exhibits a fairly broad precipitation area associated with the leading warm front and a narrower mesoscale precipitation band associated with the trailing cold front. Obviously, processes at work in cyclones lead to banded precipitation. It follows that being familiar with the Norwegian model of frontal and precipitation structures will help us understand the generation and organization of mesoscale precipitation systems.

A relationship often exists between banded precipitation and convection in both the warm and cold seasons. For example, this radar plot shows bow echoes within a longer squall line that occurred over Arkansas. However, we will not address Mesoscale Convective Systems here as they receive extensive treatment in the Mesoscale Primer module Severe Convection II: Mesoscale Convective Systems.

This section briefly reviews mesoscale precipitation bands in the context of the larger synoptic-scale cyclones in which they are found. We will not attempt an exhaustive treatment of cyclone structures, as you are likely already aware of those topics. Instead, we will focus on how mesoscale bands form within various cyclones.

In the early 1900s, a group of Norwegian meteorologists proposed a model of cyclone morphology and evolution that persists to this day. Although highly descriptive and limited by geography and available data, the Norwegian Cyclone Model has provided a valid framework around which synoptic meteorology has grown up and matured in the intervening decades.

As you view the different aspects of cyclone models, keep in mind that although each of the conceptual models is useful in its own way, none of them has a lock on describing all of the forms that an individual cyclone and its associated precipitation structures may take. So don't place too much stock in any single cyclone model to describe precipitation structures. Instead, look for those dynamic mechanisms that generate and organize banded precipitation.

The Norwegian Cyclone Model presents the atmosphere as a juxtaposition of air masses, separated by boundaries where the temperature changes greatly over a short horizontal distance. These boundaries are called fronts. A leading warm front is followed by a faster cold front. As the cold front outpaces the forward motion of the warm front, they collide and form an occluded front.

In the Norwegian model, each front is viewed as a sloping material surface, a hard boundary like an enormous sheet of glass that separates a warmer air mass from a colder one. In this model, motions near fronts are forced upward or downward along the sloping frontal surfaces. In modern views of fronts, surfaces of potential temperature or moist potential temperature act much like the material surfaces that the Norwegians envisioned.

As you know, fronts can generally be classified as cold fronts, warm fronts, occluded fronts, or stationary fronts. Precipitation in cyclones is strongly organized by these frontal structures. Hence, consideration of the structure and processes associated with fronts will help identify where and how banded precipitation occurs in cyclones.

In the case of the warm front, moist warm-sector air flows along and up the broad extent of the warm frontal slope. The slope of the warm frontal zone is relatively gentle, much less steep than that of a cold front. As a result, precipitation is often continuous in character and tends to fall in a wide band north of the surface warm front location.

We know now that mesoscale bands may form within the broad region of precipitation associated with a warm front.

In this example from our scenario case, satellite imagery shows a broad band of clouds and precipitation north of the warm front. The warm front, associated with a surface cyclone over the southern Rockies, extends from southeastern Kansas across southern Missouri at 2000 UTC on 10 April 1997.

At this time, a band of moderate-to-heavy precipitation is located north of the warm front from east-central Kansas across northern Missouri and into central Illinois.

The mesoscale band in this case resulted from the instability in the air mass flowing up and over the warm front. We can see the weak stability shown by the near-vertical contours of equivalent potential temperature in this vertical cross section of the warm front. The result was elevated convection above the warm front, leading to a relatively narrow band of precipitation.

Cold fronts typically have a steeper slope than warm fronts. Consequently, motions along such a frontal zone are often much stronger, both upward and downward. The outcome is a narrow band of convective precipitation that is showery in character. All things being equal, this is where we would usually think to look first for thunderstorm development within an extratropical cyclone.

Following the introduction of the Norwegian model, observational studies of cyclones and frontal zones determined that there were two primary types of cold frontal zones: the anafront and katafront. The katafront is one where the component of flow perpendicular to the frontal zone is faster than the frontal speed and is downward toward the surface of the earth. Remember that the flow in the frontal zone is largely parallel to the front, with only a small perpendicular component.

The downward flow leads to compression and warming and subsequent evaporation of any moisture or cloud droplets. Upward motion will be forced largely by boundary layer convergence near the surface front. This results in a long, narrow band of precipitation oriented along and parallel to the surface front. The mesoscale precipitation band in the original Norwegian Cyclone Model results from a katafront.

Anafront structure occurs when a front advances faster than the flow in which it is embedded. How does that happen? In the case of the anafront, the concept of the material surface no longer applies. The front advances due to wave propagation, and the flow responds accordingly. Warm air flows up the face of the frontal surface, which is typically not as steep as in a katafront. This upward motion is not as vigorous and occurs over a broader area, generating an expanded precipitation shield. A cold anafront is akin to a warm front in reverse.

From an operational perspective, the sensible weather associated with an anafront differs markedly from that associated with a katafront. An anafront generally produces heavier rainfall, but convection ahead of a katafront may produce heavy banded precipitation. Typically, passage of an anafront is marked by a sharp drop in temperature at the surface, but not humidity, while passage of a katafront is marked by a sharp drop in humidity, but not temperature. Examining a sounding taken through the front is one way of distinguishing an anafront from a katafront.

In the Norwegian model, occluded fronts represent the collision of a cold front with a warm front. This collision occurs because warm fronts often move more slowly than the cold frontal zones that follow them. The cold front then catches up to the warm front ahead, and the occluded front is formed. This process lifts air in the warm sector away from the surface and forces the warm, moist air aloft. The outcome is often cloudiness, low visibilities, and precipitation that may range from drizzle to a thunderstorm.

It would seem clear that occluded fronts are often the focus of inclement weather, yet that is not always the case. Sometimes the passage of the surface occluded front signals the onset of precipitation and low visibilities, but other times occluded frontal passage coincides with the end of poor conditions. These differences depend largely upon whether you're dealing with a warm or cold occluded front. Although there was no real distinction between warm or cold occluded frontal types in the original Norwegian Cyclone Model, the theory left the door open to such modification, and we examine those differences here.

Warm occluded fronts occur when the air mass behind the trailing cold front is warmer than the air mass ahead of the leading warm front. Warm sector air is still occluded upward, but the three-dimensional structure of the warm occlusion creates a cold front aloft and ahead of the surface occluded front. Given this structure, the occluded front at the surface tends to lie behind the thickness ridge on a sea level pressure and 1000-500 hPa thickness chart. The cloudiness and precipitation that occur will typically do so ahead of the surface occluded front so that improving weather occurs after the surface occluded front passes by.

Cold occluded fronts occur when the air mass behind the trailing cold front is colder than the air mass ahead of the leading warm front. Warm sector air is again occluded upward, but the three-dimensional structure of the cold occlusion creates a warm front aloft (what had been the surface warm front) and behind the surface occluded front. In this situation, the occluded front tends to lie near or to the warm side of the thickness ridge on surface and 1000-500 hPa thickness charts. The cloudiness and precipitation that occur will typically do so behind the surface occluded front so that deteriorating weather occurs after the surface occluded front passes by.

An airstream or conveyor belt model of cyclone structure and evolution has emerged in the last few decades. Unlike the earlier air mass approach of the Norwegian Cyclone Model, the conveyor belt view provides for relatively narrow ribbons of air flowing along sloping isentropic surfaces. These airstreams are defined in a system-relative-flow sense and thus represent air trajectories through the moving and evolving cyclone. Consequently, the airstreams, as they are lifted along isentropic surfaces, produce clouds and precipitation structures during various stages of midlatitude cyclone evolution. As with the Norwegian model, these airstreams also focus precipitation, but usually on the larger meso-alpha scale.

The model consists of a warm conveyor belt, a cold conveyor belt, and a dry, descending airstream. The warm conveyor belt usually transports warm air poleward and is the warm, moist air mass that is lifted over the warm front. The cold conveyor runs parallel to the warm front on its cold side. The dry airstream supplies drier, descending air behind the surface cold front, which may then ascend around the cyclone center. Although the warm conveyor belt is perhaps the most useful for describing and diagnosing where precipitation will form, the cold and dry airstreams do play a role in the organization of precipitation.

As its name implies, the warm conveyor belt is a warm airstream, often flowing poleward along most of its length. Originating in the warm sector of midlatitude cyclones, the warm conveyor belt slopes upward as it flows poleward, along and ahead of the cyclones cold front. It is then lifted as it flows up and over the warm front.

The forward-sloping warm conveyor belt is similar in many ways to a cold katafront, where the flow perpendicular to the cold front moves forward more rapidly than the frontal zone. Note that the three-dimensional flow of the warm conveyor belt is mostly parallel to the cold front with just a slight component perpendicular to the front to yield the katafront structure.

A forward-sloping warm conveyor tends to confine clouds and precipitation along and ahead of the surface cold front. This effect may be enhanced if dry air also descends above the cold frontal surface and overruns the underlying warm conveyor belt. The resulting instability leads to precipitation, possibly banded, in the warm sector.

This type of system commonly occurs downstream of a diffluent trough axis in the flow aloft.

A pronounced case of the forward-sloping warm conveyor belt results in the split-front structure. Strong advection of low-humidity air down the katafront surface leads to generation of potential instability. Consequently, a shelf of cloudiness forms along and ahead of the surface cold front. Heavy convective rain may fall from this deeper shelf of cloudiness. Subsequent passage of the surface front may be associated with a dry cold frontal passage or just a narrow line of precipitation near the front.

The rearward-sloping warm conveyor belt is similar to a cold anafront. Most of the warm conveyor flow is parallel to the front with a slight component up and over the advancing cold air. Thus, flow ascends the isentropic surfaces of the frontal zone. The rearward-sloping conveyor structure leads to heavy convective precipitation at the time of frontal passage, followed by lighter, more stratiform precipitation behind the surface front.

This type of system commonly occurs downstream of a confluent trough axis in the flow aloft.

In well-developed mature cyclones, a portion of the warm conveyor belt will turn cyclonically and be pulled into the circulation. This creates a trough of warm air aloft (a trowal) that is often relatively narrow and focused.

The trowal is a sloping canyon formed by ascending warm air with relatively cooler air in place on both sides. The trowal originates at the apex of the warm sector in the pre-occlusion phase of a cyclone. It gradually rises above the surface occluded front. Often, the outcome is a region of precipitation well separated from the warm sector in the wrap-around portion of a cyclone. When the trowal is present in a cyclone, it contributes to the cloudiness and enhances the precipitation associated with the comma-head signature of a cyclone.

This plot shows the results of a 5-year cold-season climatology in the northeast U.S. The lines in the plot show the axes of snowbands relative to the center of their respective cyclones. We can see that the majority of the bands fall in the northwest quadrant of the plot, 300-500 kilometers from the cyclone center, and are oriented roughly northeast-southwest. This plot confirms that heavy precipitation bands frequently form in the occluded, wrap-around portion of a cyclone, far from the warm and cold fronts of the Norwegian Cyclone Model.

Not all banded precipitation occurs within a midlatitude cyclone. For example, in the warm season over the central U.S., lines of thunderstorms may form in the absence of an organized cyclonic circulation. Likewise, bands of convective snowfall may occur downstream of the Great Lakes even though the responsible cold northerlies are due to a surface low hundreds of kilometers away. Sea-breeze and land-breeze circulations near the coast are two more examples of mesoscale banded precipitation that do not require a cyclone. Non-cyclogenic processes are dealt with in other modules.

In this module, we focus on those processes responsible for banding within extratropical cyclones. The previous section addressed banded precipitation associated with isentropic ascent up frontal surfaces. However, banded precipitation also occurs due to other processes at work in extratropical cyclones. This section explores some of those processes.

Please see the References section for an annotated list of related COMET modules.

Regions of deformation should be monitored for banded precipitation. Deformation is the process by which an airstream tends to stretch or alter the distribution of a tracer embedded in the flow, such as moisture. Under certain circumstances, deformation leads to frontogenesis, which can focus precipitation into a narrow band.

Deformation can arise in various ways. This figure illustrates one way, whereby two conflicting airstreams are directed toward one another in the middle or upper levels of the atmosphere. At a great distance, neither flow is disturbed, but as each flow approaches the other, they split and fan out in a diffluent pattern. A col point is generated in the center of the deformation field, at the intersection of the axis of contraction with the axis of dilatation.

Consider how the clouds are spread out along the axis of dilatation if the southeasterly flow is moist and the northwesterly flow is dry, as shown here. Over time, the moisture gradient increases along the axis of contraction and spreads along the axis of dilatation. Deformation can and does occur in the conveyor belts discussed previously.

For example, the splitting warm conveyor belt in a trowal leads to deformation along an axis perpendicular to the trowal, as shown in this figure. Another deformation zone occurs with confluence between the dry and warm conveyor belts along the back edge of the cold frontal clouds, leading to contraction of the moisture gradient.

Deformation can also affect temperature fields, leading to frontogenesis. Frontogenesis is a measure of the change with time in the magnitude of a temperature gradient. When a temperature gradient strengthens with time, frontogenesis values are positive, and we say that frontogenesis is occurring. This plot shows a case where deformation leads to frontogenesis. Note that the isotherms are oriented perpendicular to the axis of contraction.

When a temperature gradient weakens with time, frontogenesis values are negative, and we say that frontolysis is occurring. Frontolysis may occur when isotherms are oriented perpendicular to the axis of dilatation.

When frontogenesis or frontolysis alters the thermal gradient, there is a direct impact on the pressure or height distribution in the region of frontogenesis. This can be seen in the change in thicknesses across the region of frontogenesis. Because of the atmosphere's strong tendency to maintain geostrophic balance, this change in the thermal structure forces ascent and descent to occur in regions of frontogenesis. Ascent occurs on the warm side, which adiabatically cools the rising flow. Simultaneously, descent on the cool side adiabatically warms the flow there. If the air is moist, these induced vertical motions support cloud formation and precipitation along the axis of dilatation. Consequently, deformation and frontogenesis are very important in organizing precipitation into mesoscale bands.

Evaporation and melting-induced circulations can act to focus precipitation into mesoscale bands. When a dry layer exists near the surface, frozen precipitation, especially snow, falling into it will melt and/or evaporate. This process has several effects on the dry near-surface atmosphere:

• It cools down to near or below freezing
• It moistens, allowing later snowflakes to fall farther before evaporation
• It becomes dominated by a mesoscale downdraft, induced by evaporative cooling aloft

A thermodynamic diagram, such as a Stuve or skew-T chart, is ideal to illustrate these changes. The sort of sounding evolution depicted here is the kind that forecasters should look for in real-time balloon flights as well as model output.

Notice the progression in this animation. As snow starts to fall, it first melts, then evaporates before reaching the surface. As a result, the atmosphere cools and moistens from the top downward. Once the atmosphere is completely saturated, no further cooling can occur by evaporation. However, this does not rule out cooling through melting. Even after the atmosphere becomes saturated, additional cooling can still occur while the temperature is still above freezing. Like evaporating raindrops or sublimating snowflakes, melting can rob the air of heat energy. In order to melt ice, more heat must be taken from the surrounding air. Notice in the animation that melting-induced cooling continues after saturation is achieved so that by the end of the animation the temperature profile is nearly isothermal at 0°C.

When these processes occur in a synoptic environment with continuous precipitation, the result is a continuous downdraft.

In addition to generating a downdraft, cooling the air on the cold side of a front increases the temperature gradient. The resulting frontogenesis strengthens vertical motions along the front. Thus melting and evaporation lead to a mesoscale convergence zone at the surface. Consequently, warm air is forced up and over the cold air, creating a mesoscale version of the synoptic cold front, which results in banded precipitation.

Banded precipitation of this type often occurs when low-level temperatures are close to freezing. Thus, such events frequently begin as rainfall and change over to snow. Furthermore, surface frontal zones often coincide with the spatial transition from rain to snow. When this occurs, the heaviest precipitation falls in a band oriented along and parallel to the frontal zone, in the vicinity of the 0°C surface wet bulb isotherm.

The scenario that started this module examined a case of banded precipitation that struck Scott Air Force Base and the St. Louis region on 10 April 1997. Light precipitation was in the forecast. This is not surprising, given that model progs indicated deep moisture and weak ascent associated with an approaching trough. The challenge of the forecast was twofold: (1) to predict when the rain would change over to snow and (2) to predict if and where the precipitation would be banded. Evaporative and melt-induced cooling played a large role in that forecast challenge. The scenario focused on the changeover from rain to snow. Now let's look at the question of banding.

This loop of soundings from St. Louis, Missouri accompanies the surface plots we just examined.

The loop of forecast soundings looks like an inverted zipper. Over time the temperature and dewpoint traces close in on the wet bulb trace from aloft, moving downward toward the surface. More importantly, the wet bulb temperature throughout that layer hovers near 0°C. Near the end of the loop, the near-surface layer cools from a few degrees above freezing down to the 0°C isotherm. In short, evaporation initially moistens and cools the near-surface layer to near freezing. Once the layer is saturated, melting-induced cooling maintains the cold temperatures at very near 0°C from about the 700 hPa level down to the surface.

So what happened in this event? From 1100 UTC through 1500 UTC, the time period of significant snow in the area, a persistent band of precipitation oriented WNW-ESE extended from east-central Missouri through the southern tip of Illinois, with individual reflectivity elements moving east-southeastward along the band. From 1300 through 1430 UTC the strongest reflectivity occurred in a narrow region on the southern edge of the band, where the rain/snow boundary was likely to be located, just south of St. Louis.

Look at the 0400-2000 UTC loop of observations centered on the area of interest. Note that mixed precipitation begins to move into the region from the west by 0900 UTC or so, and there is evidence of mixed precipitation through 1300 UTC over central MO. Temperatures in the Scott vicinity were 3 to 5°C at 0900 UTC, but wet bulbs were below 0°C at most locations. Snow showers began in the immediate vicinity by 1000 UTC, with temperatures above freezing. By 1100 UTC the surface layer was saturated, with light snow at most locations. Light to moderate snow continued through 1500 UTC. No stations at the surface to the south reported rain during this period, but surface temperatures south of St. Louis and Scott were 1 to 2°C with high relative humidities. This suggests that there was rain in the area, but that AWOS stations did not report it. By 1700 UTC, gradual warming on the west side of St. Louis had occurred, with light rain and drizzle reported at two stations. This west-to-east trend continues through 2000 UTC, with temperatures at that point well above freezing.

To summarize, in this event model forecasts predicted a broad and light precipitation field. Instead, a narrow band of precipitation occurred over parts of Missouri and Illinois. Cooling induced by melting and evaporation apparently forced local circulations that focused the precipitation in a narrow band. This band was located along the transition from rain to snow, roughly coinciding with the southern edge of the 0°C surface wet bulb temperature.

Frontal merger is the joining of frontal zones that are sloped in roughly the same direction. This contrasts with the development of an occluded front, which is formed by the merger of frontal zones that slope in opposite directions. This conceptual animation shows an example where an advancing warm front climbs up a stationary front. The additional lift generates a narrow band of precipitation.

In March of 1992 Neiman and his colleagues observed frontal merger during an extensively documented field experiment. In this case, a warm front associated with a midlatitude cyclone rode up over a stationary front. Subsequently, an arctic air mass dropped south out of Canada and bulldozed its way first under the stationary front, and then under the cyclone. As the arctic air mass advanced southward and encountered fronts, it essentially pushed under the warmer air masses. Consequently, air masses and fronts to the south decoupled from the surface and rode up over the cold dense arctic air. The additional lift produced heavy banded precipitation.

As with occlusion, the outcome of frontal merger is the displacement of relatively warm air upward and the creation of clouds and precipitation. To determine the occurrence of frontal merger, closely examine cross sections, such as the one shown here. Regions of frontal merger are most likely to experience mesoscale bands of precipitation.

Frontal merger does not conveniently fit with any cyclone model. However, cyclones may play a role in the formation of the fronts involved. On a synoptic scale, this process will most likely occur when an arctic outbreak from high latitudes encounters either a stationary front or a warm front at lower latitudes. Thus, the interaction of synoptic systems can enhance vertical motion in a linear pattern. In a moist environment, this can lead to mesoscale precipitation bands.

Precipitation bands frequently occur in the absence of a frontal structure. When these bands appear in a stable environment, in the absence of upright convection or other known forcing mechanisms, we may be dealing with slantwise convection.

Slantwise convection may account for bands that appear in the midst of a broad precipitation shield or for nearly all of the precipitation in a single event. The great importance of slantwise convection lies in the fact that it produces significant, sometimes unanticipated, banded precipitation.

The precipitation bands associated with slantwise convection can be from 100 to more than 500 km long and from 5 to 40 km wide. This radar image shows banded precipitation near Salt Lake City, Utah on 4 March 1999. The bands go right across the mountain ranges, which can be seen as areas where the echo was removed due to ground clutter. Most of the accumulation from this storm came from these of bands.

Slantwise convection results from the release of conditional symmetric instability (CSI). CSI is not easy to visualize. Basically, the idea is that an air parcel may be stable with respect to vertical displacement (convective stability), and it may be stable with respect to horizontal displacement (inertial stability), but it can be unstable when it ascends along a slantwise path. This is why CSI is sometimes called slantwise convection.

But what does it mean to be stable? First let's look at vertical stability. When a dry parcel of air is displaced vertically, it cools at the adiabatic lapse rate of about 10°C/km [animate]. If the surrounding air is cooler yet, that parcel becomes warmer and less dense than its surroundings and will rise convectively. This environment is absolutely unstable, which is quite uncommon. When a saturated parcel of air is displaced vertically, it cools more slowly, at the moist adiabatic lapse rate of about 6.5°C/km. If it still is warmer than its environment, it will continue to rise convectively. If the environment is stable with respect to dry ascent, but not moist ascent, then it is conditionally unstable. If the environment is stable with respect to moist ascent, then it is absolutely stable.

Horizontal stability arises from inertial stability. Flow with positive inertial stability follows isobars or height contours. CSI requires weak inertial stability, so that flow can deviate horizontally from following heights or isobars as it ascends. Let's see how that occurs.

Inertial stability is determined by the absolute vorticity. When absolute vorticity is negative, the atmosphere is inertially unstable. Absolute vorticity is the sum of the earth's vorticity and the relative vorticity. The earth's vorticity is a cyclonic vorticity caused by the earth's rotation, as shown here.

A strong, anticyclonic relative vorticity may counteract the earth's cyclonic vorticity, resulting in a weak absolute vorticity. Weak absolute vorticity leads to weak inertial stability.

Anticyclonic vorticity may arise in two ways. The first is when flow bends in an anticyclonic fashion, as shown here.

A second factor contributing to anticyclonic vorticity may be horizontal wind shear. Typically, this is found on the equator side of a jet, as illustrated here.

CSI occurs under conditions of weak or neutral inertial and convective stability. If we position ourselves on the equator side of a jet, then that flow generates an anticyclonic relative vorticity. This is the area where we may find weak inertial stabilitythe area most prone to CSI.

To review, a parcel may be stable with respect to vertical displacement (convective stability) and the parcel may be stable with respect to horizontal displacement (inertial stability), but the parcel can be unstable when it ascends a slantwise path.

One way to assess CSI is with a vertical cross section drawn perpendicular to the mean wind shear. On this cross section we plot equivalent potential temperature, known as theta-e, and geostrophic momentum. Under conditions of weak or neutral inertial and convective stability, when we lift an air parcel, it follows a surface of constant momentum. If that path takes the parcel to a lower equivalent potential temperature, then the advected air parcel becomes warmer than its environment. The parcel then continues to rise convectively along the slanted path, and we get slantwise convection. Thus, in this cross section, CSI occurs where the contours of equivalent potential temperature become steeper than the momentum contours. Slantwise convection would result in bands oriented parallel to the flow and perpendicular to the cross section.

There are several caveats here, but the most important one is that if the environment is unstable with respect to vertical convection, then that process will dominate. In a cross section, vertical convection would occur anyplace where theta-e decreases with height. This includes locations where the contours of equivalent potential temperature are overturned past vertical.

Also note that some initial ascent is required to release CSI. Without that initial trigger, CSI is not released and slantwise convection does not occur.

Under what conditions will contours of equivalent potential temperature become steeper than the contours of momentum? This conceptual vertical cross section depicts a relatively uniform thermal environment with weak conditional stability and flow into the page. The atmospheric lapse rate would be slightly less than the moist adiabatic lapse rate. On a cross section, contours of equivalent potential temperature would be horizontal and spaced far apart. Under these conditions, rising air could never move from higher to lower equivalent potential temperature.

In order to achieve CSI, we first need to tilt the lines of equivalent potential temperature, so that the atmosphere is cooler to the left. Thus, the first thing we need with CSI is a horizontal temperature gradient.

Under conditions of weak conditional stability, there is only a small vertical gradient in equivalent potential temperature. Thus we require only a relatively small horizontal temperature gradient to tilt the contours steeply over a substantial depth. With a more stable environment, the same horizontal temperature gradient would result in very little tilt.

In our cross section depicting a uniform atmospheric environment, momentum contours would be essentially vertical. Remember, reduced to its most simplistic terms, momentum is just mass times the velocity. On this cross section, momentum increases from left to right as an artifact of how we compute geostrophic momentum, but what we really care about is the orientation of the contours.

Now let's add strong wind speed shear into the line of the cross section so that momentum increases as we ascend. This tilts the momentum surfaces over to the left, causing momentum to increase with height in any given vertical profile.

We see that contours of equivalent potential temperature can become steeper than momentum surfaces if we satisfy two conditions:

1. A horizontal gradient in equivalent potential temperature
2. A strong vertical gradient in geostrophic momentum

Thus, CSI may form in an environment with

1. Weak conditional stability
2. A horizontal temperature gradient
3. Strong unidirectional wind shear

Before CSI contributes to banded precipitation, we need to add two crucial ingredients: moisture and lift. For moisture, we require a relatively saturated atmosphere with relative humidity over 80%. For lift, consult model products such as vertical velocity or omega that indicate rising motions required to initiate the release of CSI. Because CSI requires horizontal temperature gradients, it frequently occurs in association with frontogenesis, which may provide the initial ascent.

Generally, the conditions for CSI are met in areas associated with a shallow front. These can be either ahead of a warm front or behind an anafront. These regions are usually characterized by a convectively stable environment and conditions that are near saturation. Another typical feature is the presence of a jet just north of the area. This provides the anticyclonic relative vorticity that is typical of regions of low inertial stability. The jet also contributes to the strong vertical shear found in CSI events.

In summary, a few simple rules will help you be aware of the potential for CSI and mesoscale banding associated with it.

1. CSI occurs in regions of thermal gradients, which may or may not be frontal zones.
2. CSI occurs in moist environments, near saturation.
3. CSI occurs under conditions of strong vertical wind speed shear, commonly unidirectional.
4. CSI occurs under conditions of weak stability.
5. CSI occurs under conditions of weak inertial stability, indicated by small absolute vorticity.
6. CSI typically occurs in the middle levels of the atmosphere, frequently above orographic influences.

See the In Depth section for a detailed look at one method for forecasting CSI and slantwise convection.

During the early morning hours of 4 March 1999 a small, but unexpected snow event took place along the Wasatch Front and adjacent mountains. Snowfall totals of 1-2 inches were common along the valleys and benches with up to 6 inches in the mountain valleys and ski areas.

On average two to three CSI events occur in northern Utah per winter. CSI-induced bands will nearly always concentrate precipitation and produce intensities and amounts much greater than what the general synoptic pattern would suggest is possible.

This short case exercise will examine the March 1999 event, which was reported by Dunn in 1999.

From a synoptic viewpoint, at 1200 UTC on 4 March, the 500 hPa Eta 12 hour forecast showed generally zonal, low-amplitude flow across the West with a few weak embedded waves. In the northern Utah region, the height gradient at 500 hPa increased slightly to the north and the omega field showed an area of weak upward motion covering most of Utah.

Precipitation bands occur in various widths. They range from the meso-alpha rain shield ahead of a classic warm front, hundreds of kilometers wide, to the meso-gamma line of convective rain shafts of a classic supercell thunderstorm, 1 to 2 km wide. Yet, our present numerical modeling prowess allows an accurate portrayal only of the former, meso-alpha band.

The reasons for this failure are varied. Under the best of conditions NWP models may struggle to produce an accurate precipitation forecast. In fact, NWP models typically show the least skill at quantitative precipitation forecasts. When confronted with processes that operate on small spatial and time scales, models may fail to forecast any precipitation at all. This section will examine some of the reasons that limit the ability of NWP models to forecast mesoscale banded precipitation.

The following discussion assumes that you already have some familiarity with mesoscale NWP models. For a brief review of mesoscale models, see the COMET module How Mesoscale Models Work.

The primary reason that our present NWP models fail to predict meso-gamma precipitation bands is that the bands are so small that the models simply cannot clearly produce them.

Consider a model that uses a grid spacing of 5 km. This means that the model looks at the atmosphere every 5 km. In numerical modeling, 5 grid points (4 grid spaces) are generally accepted as a minimum to resolve a given wave although more are always helpful. Thus, the smallest wave that our model will be able to simulate is 20 km long. This is the models minimum grid resolution. Our 5 km model will not produce waves smaller than this value.

The lesson here is that you cannot expect to see a 15 km wide band of precipitation in a model forecast produced with an operational 15 km grid spacing. The most narrow you could reasonably expect would be 60 km.

Although convection is addressed in many other sources, there is no other atmospheric phenomenon that is of such small horizontal dimension and yet spans the depth of the troposphere while providing NWP models and modelers with the challenges of:

• Rapid cloud development
• Precipitation generation (including hail)
• Incoming shortwave and upwelling longwave radiation
• Violent upward and downward motions
• Turbulence generated by strong, small-scale vertical motions.

Furthermore, mesoscale banded precipitation frequently results from convective precipitation embedded within a larger precipitation field. For these reasons, if a model can adequately depict convection, we can be reasonably certain that it will adequately depict mesoscale banded precipitation related to midlatitude cyclones.

A hydrostatic atmosphere implies that vertical motions are insignificant. Air parcels move great horizontal distances without rising or sinking. However, the convective environment is one of powerful updrafts and downdrafts; we term such an environment non-hydrostatic. Thus, non-hydrostatic NWP models allow strong vertical motions. Current models with grid spacings of 10 km or less treat the atmosphere non-hydrostatically. In fact, most mesoscale NWP models are now non-hydrostatic, including COAMPS and the AFWA MM5.

This graphic depicts some of the process that must be accounted for over the life cycle of a single convective cell. High-resolution models that treat the atmosphere non-hydrostatically will ideally be able to simulate these processes. To look at clouds in this painstaking detail is to treat them explicitly in a model. However, a model with wide grid spacing cannot treat clouds with such care since the model grid is too widely spaced to see each individual thunderstorm. Yet we cant afford to ignore the convection altogether. So what do we do? The answer is cumulus parameterization.

Clearly, dealing with convective clouds explicitly is superior to cumulus parameterization. However, an explicit approach to cloud modeling takes far more computational time than most cumulus parameterization schemes. In short, cumulus parameterization is expedient, but cannot resolve individual convective clouds, only the aggregate changes in temperature and moisture that they provide. As a result, parameterization is better suited to models with wider grid spacing (> 25 km). Explicit representation of convection can be very detailed, but such insight only comes at the expense of a great deal of computing power and time. Thus, explicit cloud treatment is more appropriate for models of a finer grid spacing (< 10 km).

The representation of precipitation processes is of particular importance to our larger discussion of mesoscale banded precipitation, as well as our specific example of convection.

NWP models treat precipitation with either diagnostic or prognostic moisture physics. With diagnostic moisture physics, the model checks each grid point from the top down to determine if the atmosphere is supersaturated (relative humidity > 100%). If it is, the excess moisture is carried to the next lowest level in the model. This moisture is accumulated until it reaches the ground.

With prognostic moisture physics, the model has separate equations that account for rain production, ice crystal production, and even hail generation. These equations examine variables such as temperature and wind velocity in addition to excess moisture at different levels for each grid point.

The more elaborate prognostic moisture physics will generally produce a better forecast, but requires more computational work than diagnostic moisture physics. And remember, regardless of the moisture physics, NWP models still struggle to produce accurate quantitative precipitation forecasts.

Whether modeling with a cumulus parameterization scheme or an explicit treatment for convection, it is important to recognize that such small-scale features may well have an impact on meteorological features of larger scales. If a thunderstorm occurs on a grid column in your model, then that column has become warmer and more moist through a much deeper layer than it had been previously. While the changes in moisture are important, the changes in temperature are crucial to having a realistic simulation. Having a large area of thunderstorm activity means a broad area of deep convective heating. This heating can alter flow in the upper troposphere and may impact the jet stream.

Thus, a mesoscale model that accurately depicts convection (or other mesoscale phenomena) may also produce a better synoptic forecast.

There is an almost bewildering array of models available to forecasters, both at home and abroad. Each model has its strengths and weaknesses. Depending on the model resolution and physics, some may be better suited than others for a particular event that might include bands of heavy precipitation.

So, good forecasting requires that you:

1. Have a good situational awareness of the synoptic situation,
2. Have a good awareness of the climatology of your particular area of responsibility,
3. Examine the model analysis to see how well the current model run initialized, and
4. Are familiar with model characteristics.

Model characteristics may be compared by going to the Operational Models Matrix: Characteristics of Operational NWP Models.

On 10 December 2003, banded heavy snow fell over portions of Missouri in the vicinity of Whiteman Air Force Base. While snow totals there were only about 1 inch, there were locations less than 100 km away from Whiteman with accumulations as high as 5 inches, as shown in this figure.

As a forecaster early on 10 December 2003, your task would have been to determine if banded heavy snowfall would occur, and if so, whether or not those bands would affect Whiteman.

Model output from both the COAMPS and the NOGAPS models are examined for this exercise. As you know, the COAMPS is a higher-resolution, non-hydrostatic model (grid points every 27 km) while the NOGAPS has a coarser-resolution, hydrostatic architecture (grid points every 1.0° x 1.0°, about 100 km). While the COAMPS should permit us to see finer-scale features, it also deals with convection directly. However, it is crucial to note that the precipitation fields shown here are 6 hour accumulations. As such, there may be smaller bands that occur during a given integration that are masked by the bands very motion. At the end of 6 hours the smaller band may have propagated some distance, and its impact during any shorter period during those 6 hours may be lost or at least diminished. The meso-alpha structure of a precipitation field is retained, but smaller features are not revealed.

At 0000 UTC 10 December 2003, as analyzed by the NOGAPS model, a vigorous cutoff low was tracking across Oklahoma and moving east-northeastward toward southern Missouri.

The COAMPS model analysis at this time depicted a slightly stronger, sharper trough centered over eastern Oklahoma. Compared to the NOGAPS analysis, the cutoff center is slightly deeper and located a bit farther to the southeast. (Note that the contour interval is halved in the COAMPS analysis.)

The storm system aloft was clearly defined in the infrared satellite imagery at that time, with a comma-shaped signature centered over northeastern Oklahoma and widespread deep cloudiness out ahead of the system.

This deep trough aloft had caused a strong surface low pressure system to form out ahead of the 500 hPa trough, centered over central Missouri.

The surface trough was becoming somewhat elongated at this time, with cyclonic flow extending all the way up past the Great Lakes and southwestward into Texas. Also note the extended fetch of northerly flow from the northern plains to eastern Oklahoma. The associated cold front extended from the surface low south-southeastward through Arkansas into eastern Louisiana.

Aside from moist conditions over the cold frontal zone, high relative humidity values at 700 hPa were confined to regions north and northeast of the upper low and also well out in front of the cold front over the southeastern U.S..

The COAMPS analysis indicated a well-defined dry intrusion aloft, the dry conveyor belt, over extreme southeast Oklahoma and northeastern Texas. This dry conveyor was moving northward into western Arkansas toward the Missouri border.

At 6 and 12 hours of simulation in the NOGAPS model, an area of moderate upward motion, was forecast to track across central and northeastern Missouri. The rising motions formed in response to the approach of the strong upper trough.

At 0600 UTC 10 December, NOGAPS was predicting regions of significant frontogenesis at 850 hPa over northeastern Missouri and near the pre-existing surface cold front over Mississippi.

The coarse NOGAPS solution reveals one large, meso-alpha band of precipitation extending from the Illinois/Iowa border down to south-central Missouri. Over 0.75 inch (19 mm) of liquid precipitation (or its equivalent) was predicted by NOGAPS in 6 hours over a broad area. A smaller area of weaker precipitation is predicted along the Oklahoma-Kansas border; a banded structure there is not obvious.

The 12 hour forecast of 700 hPa frontogenesis by COAMPS revealed significant positive values associated with the cold front over western Tennessee into northern Alabama and also over central and north-central Missouri. It appears that this forcing may have played a role in the development of model precipitation over central Missouri.

The finer COAMPS solution depicts several bands for the same period. The first originates in largely the same region as shown by NOGAPS, near northeast Missouri, but angling more southwestward toward southwest Missouri. A second band is shown over extreme northeastern Arkansas, while a third area with more of a banded shape is shown over eastern Kansas. In all, COAMPS forecasts three bands, as opposed to the single large band forecast by NOGAPS.

Closer examination of the NOGAPS and COAMPS predictions in this case reveal some similarities to a case discussed earlier in this module, the well-developed trowal of 10 November 1998. Precipitation bands developed west and northwest of the upper cutoff low in both cases.

Shown here are the 12 hour, 700 hPa wind and theta-e predictions at 1200 UTC on 10 December 2003.

In both cases, warm air aloft is being wrapped in a counterclockwise direction back around the upper low, as part of a relatively moderate warm conveyor belt or trowal situation, as discussed previously. Many previous studies have shown that this airflow continues to ascend on the backside of the storm system, thereby contributing to development of a precipitation band. In both models, significant precipitation was predicted to develop over central and northeastern Missouri, which coincides with the ascending, wraparound motion of the warm air aloft.

This composite radar image was observed at 0800 UTC on 10 December. Note the precipitation band in central Missouri, northwest of the cutoff low. More precipitation fell over southwestern Missouri than either model predicted. NOGAPS seemed to have a better handle on the southern portion of the precipitating region although this models secondary precipitation max was located well to the east of the observed max.

When one takes a look at the COAMPS analyzed fields at 700 hPa for 1200 UTC 10 December, some errors in the 12 hour forecast for this time are apparent. The wraparound moisture in the analysis extends into southwestern Missouri, farther south and west than in the forecast. Moreover, the winds wrapping around the storm aloft are about 10 to 15 knots stronger than those in the forecast, indicating that COAMPS underpredicted the strength and extent of the wraparound of the warm conveyor belt. This was likely the primary cause of COAMPSs severe underprediction of precipitation amounts over southwestern Missouri. The ascending warm conveyor apparently ended up slightly stronger and displaced compared to the NOGAPS and COAMPS predictions.

What do we make of all of this? It would appear that elements of the truth are found in both models solutions. Indeed, a dominant band of precipitation is present in the radar observations, as suggested by the NOGAPS solutions. Concurrently, additional bands were forming, a solution hinted at in the COAMPS values.

In summary, neither model handled the actual weather perfectly. As with most numerical models, the final solutions are based in part on initial conditions (the observations) that do not resolve every detail. NOGAPS has a coarser grid spacing than COAMPS and uses special parameterization schemes for convection and other processes. Moreover, such schemes are often tuned to work within a particular model architecture. As a consequence, in this case the placement and accumulations of the major band are handled better by NOGAPS. By contrast, the non-hydrostatic, cumulus-resolving COAMPS solution portends the presence of bands, although they are not well placed or indicate the correct intensity. Yet, when we consider that so many precipitation bands are convective, the need for non-hydrostatic models and their continued development becomes clear. In the end, the messages from both models were worthwhile and clear: Precipitation bands near St. Louis were likely, but accumulations would not be significant.

In most parts of the world, Doppler radar data are not available. However, satellites can frequently detect banded precipitation. This section provides an overview of the different satellites and sensors used to detect and estimate precipitation. We will examine Hurricane Jeanne, as it came ashore in the southeast U.S. to provide a basis of comparison.

Infrared (IR) emissions are most useful for estimating precipitation from convective clouds because higher, and thus colder, cloud tops correlate with higher precipitation rates. In this method of estimating precipitation, IR temperatures are averaged over various areas and times. Those averages are then compared with precipitation measurements to arrive at an operational temperature-precipitation correlation. One example of this is the GOES Precipitation Index, or GPI. It uses 235 kelvins as the IR temperature with the best correlation to average precipitation for areas spanning 50-250 km over 3-24 hours.

These images compare a GOES IR image with a composite radar image for Hurricane Jeanne as it moved over Georgia. One can see that the IR image fails to provide the same resolution as the radar image. When averaged over longer periods of time, the resolution drops even further. Thus, the GOES precipitation index provides us with a general overview of precipitation, but fails to depict most mesoscale banding.

Other disadvantages of IR-only techniques are the lack of detection of rain from warm clouds, the inability to differentiate between convective precipitation and cold cirrus, and unreliable results in midlatitude regions during the colder seasons. There is also a tendency to underestimate precipitation early in the lifecycle of convective systems and to overestimate precipitation in the dissipating stages.

The primary advantage of IR-based techniques is the high frequency of images: 30 min or less for geostationary satellites. Recent satellite precipitation estimates have combined low-earth orbiting microwave measurements, which have high resolution, but low frequency, with the more frequently available geostationary IR. At the end of this section, we will examine the blended precipitation product developed by the Naval Research Laboratory in Monterey.

The Defense Meteorological Satellite Program (DMSP) Special Sensor Microwave/Imager (SSM/I) brought fundamental change to the identification of precipitation from satellites because it directly detects precipitation in clouds. These images show the SSM/I-derived rain rate along with the composite radar reflectivity for Hurricane Jeanne over Georgia. Note that the major precipitation band on the northwest side of the storm is preserved in the satellite image, though its fine structure is lost.

The SSM/I measures the microwave scattering and emission signatures of hydrometeors. The brightness temperatures at 37 and 85 GHz are used to infer the quantity of liquid water and ice in a column, which correlates well to surface precipitation. The 85 GHz channel is strongly scattered by ice and is used to define rain intensity.

Both scattering and emission-based methods are used to identify rainfall. Emission techniques work well over the ocean where the background emissivity from the ocean surface is low and uniform. Over land, the background emissivity is higher and more variable, making it difficult to distinguish raindrops. Therefore, rain-retrievals over land rely upon a measure of scattering.

The NOAA Advanced Microwave Sounding Unit (AMSU) includes channels at 89 and 150 GHz that are well-suited for detecting frozen hydrometeors. We see a reasonable correlation between radar and satellite measurements in this comparison of the AMSU rain rate product and the radar base reflectivity.

The basis of the AMSU precipitation product is the scattering of microwaves by frozen precipitation. The equations used to determine the AMSU-derived precipitation allow for more accurate retrievals over land. This contrasts with the earlier SSM/I emission algorithms, which are more suitable for rainfall detection in marine environments.

TRMM is the first satellite dedicated to measurement of rainfall over the global tropics. It travels an orbit which is inclined to the equator so that it covers an area between 35°N and 35°S. Each orbit takes about an hour and a half to complete. Consequently, all times of day and night are bp_d although any given locality may only be imaged once or twice a day.

The TRMM Precipitation Radar is the first satellite-based precipitation radar. It has a 4 km resolution, a 220 km swath width, and is able to detect rain rates as low as 0.7 mm/hr. In this comparison of the TRMM Precipitation Radar and radar base reflectivity, we see a strong correlation between the two images with detailed resolution in the satellite image. We can also see the narrow the 220-km satellite swath delineated by the red lines.

The TRMM Microwave Imager, or TMI, measures the intensity of radiation at 10, 19, 22, 37, and 85 GHz. The four higher frequencies are similar to the SSM/I. The 10 GHz channel was designed to provide a more linear response for the high rainfall rates common in tropical rainfall. In this comparison, we can see how well the TMI rain rate product agreed with the composite radar reflectivity as Hurricane Jeanne came ashore in Florida.

By virtue of being in a lower orbit, the TMI provides better resolution than the SSM/I or the AMSU. However, this also results in a narrow swath across the surface. Unfortunately, the TMI is due to be terminated soon, but will eventually be replaced by even better instruments.

With all of these satellites presently in orbit, it was only a matter of time before all of the satellite-derived precipitation estimates were combined into one product. That time has come. The Naval Research Laboratory provides what we refer to as blended precipitation products, which are available online. Many other groups are also working on similar products. This image shows a 3 hour precipitation estimate derived from the NRL blended IR-microwave product. Compare that to the 3 hour composite radar plot for the same time period. Note the distinct similarity in both the location of precipitation as well as the local maxima.

With blended precipitation products, precipitation estimates derived from geostationary IR imagery are statistically tuned to agree with more accurate estimates from microwave sensors. The geostationary imagery is high frequency with vast areal coverage from 60°N to 60°S, but is not very accurate. In contrast, the microwave measurements are low frequency, but display high accuracy and high resolution. Microwave sensors include the TMI, SSM/I, and AMSU satellite instruments discussed previously.

The way blending works is that every 15° latitude/longitude box has its own calibration between IR temperature and precipitation. This calibration is continually updated based on the most recent microwave-derived precipitation estimates.

In addition to improved calibration, some of the other problems with IR-based precipitation estimates are also mitigated. Special tests are applied to detect and mitigate the effects of high cirrus clouds, to distinguish building versus dying storms, and to account for terrain enhancement.

While still not as accurate as surface-based radar, blended precipitation products provide a tremendous improvement in precipitation estimates in regions without radar coverage, including the world's oceans.

A new generation of satellite sensors is due to be launched in the coming years. These include two of special interest for forecasting precipitation: the NPOESS CMIS and the GPM.

The National Polar Orbiting Earth Sensing Satellite Conical Microwave Imager Sounder, or NPOESS CMIS, builds on a number of passive microwave sensors from the last 25 years. CMIS represents a new era in storm monitoring, supplying data to forecasters in a matter of minutes after overpass, far faster than current polar-orbiting satellite systems. As shown here, CMIS will have a large, 2 meter antenna providing finer spatial resolution than previous microwave sensors. Like TMI, it will provide very high-resolution images, but with a wider 1700 km swath. Precipitation measurements will have an accuracy of 2 mm per hour.

The Global Precipitation Measurement or GPM is an international program that will operate a constellation of precipitation satellites. A series of satellites will carry radar and other microwave sensors to create global coverage of precipitation at 3 hour intervals. GPM technology is based on TRMM instrumentation. Although the GPM is primarily a research program, like the TRMM, it is anticipated that some GPM products will be useful for operational forecasting.