The Humidity Budget of the NCEP Reanalysis 1979-2003

This paper will look at the humidity budget of the globe during the last 26 years by looking at the data contained within the NCEP reanalysis I. The zonal mean budget will be the primary focus, in an effort to further our understanding of the dynamical processes which change the relative humidity values on a daily scale. Climatology and trends in the general budget will be displayed via global zonal mean charts and interpretations will be made in the remarks section, accompanied by other author’s thoughts in the background section. In conclusion, we will comment on the ever-changing humidity budget of the earth, the problems with the NCEP dataset in general, and on future work needed in this field.

The humidity budget of the earth is one of the least resolved problems in atmospheric science today. This study deals with the relative humidity budget as seen through the data contained within the NCEP reanalysis I. Relative humidity differs from other moisture parameters in that it is a balance between the amount of moisture in the atmosphere and the atmospheric temperature. Specifically, relative humidity is the ratio of the water vapor pressure, which depends primarily on moisture, and the saturation vapor pressure, which depends on temperature. As such, one can change the relative humidity of an air parcel by changing the amount of moisture (for example, by evaporation and/or condensation) or by changing the capacity of the air to hold moisture (by changing the temperature).

There are many reasons for looking at the global relative humidity budget. Dynamical processes in the atmosphere, such as convection and subsidence can bring about changes in temperature by adiabatic compression/expansion arguments. Depending on these processes and on the relative humidity budget in an area, one may cloud formation or precipitation. The relative humidity budget and the cloud budget are therefore very closely related. These phase changes and dynamical processes in the atmosphere warrant the use of relative humidity as a moisture parameter.

Many authors have noticed that global relative humidity values in models will not change with time. This, in theory, is due to the relationship between water vapor and temperature. If there were an increase in temperature globally, one would expect to see an increase in the amount of water vapor that the air can “hold”. This increase in water vapor will then increase the temperature even further by greenhouse gas arguments, resulting in a similar ratio between water vapor and temperature as before. This ratio is directly proportional to the relative humidity parameter, which also would not change with time. On the other hand, authors who have used data obtained from observations have noticed both increases and decreases in the globally averaged relative humidity values over the last 20 years. This study will look at this debate by comparing relative humidity data from the NCEP reanalysis over a 25-year period with these findings.

Moisture is a difficult parameter to measure in the atmosphere. Older satellites lack the instrumentation to penetrate cloud cover to retrieve vertical profiles of water vapor. Radiosondes often fail to accurately measure the relative humidity at higher levels, due to the vast discrepancy between the high amounts of water vapor low in the atmosphere and the relative sparse amount of water vapor higher up. This makes the calibration of such hygristors tricky in that your response times greatly increase with height, due to the lower number of water molecules (7) at that level. Recent work with microwave water measurements is promising, and more robust observational datasets now exist. Future work in this field will most likely integrate these datasets into working models to predict the time evolution of the relative humidity field in response to global warming.

There are many studies on relative humidity in the current literature. A handful of these studies raise questions that are very closely related to this paper.

Peixoto and Oort (7) found in 1995 that the relative humidity in the Tropics is primarily determined by the moisture values there (q). Conversely, they noticed that the relative humidity in the midlatitudes is determined mostly by the temperature (T). They also notice the increase of relative humidity as one goes from the subtropics to the midlatitudes, and suggest that the mechanism responsible is moisture advection from the subtropics as well as the decrease in temperature as one moves northward. Peixoto and Oort also showed that temperature influences relative humidity over the continents more so than over the ocean and suggest that there is a much larger seasonal difference in relative humidity over the continents than over the ocean. This, of course, depends on which latitude you look at since in the midlatitudes transient and stationary eddies often exchange very different air masses on varying time periods.

Held and Soden (9), in 2000, comment on the difficulty of deciphering the relative humidity trends due to the complexity of the atmospheric circulation. Furthermore, the authors suggest that the few radiosonde measurements over the dry subtropical areas exacerbate the problem but giving the atmospheric community a general lack of good data. The problem, as they see it, is becoming smaller due to the extensive coverage over those areas by satellites which measure water vapor. They also notice that current computer models largely show relative humidity values as indifferent to changes in global temperature, and that most current GCM’s tend to underestimate water vapor concentrations by about 5%. This may be a moot point though, since “much of the vertical transport of heat, momentum, and moisture in the tropics occurs on scales of a few kilometers or less, in turbulent eddies generated by moist convection, scales that are not explicitly resolved in global climate models.” Held and Soden point to the fact that relative humidity does not change in current GCM models involving global warming, which they feel proves the strength of the positive water vapor feedback on temperature.

Conversely, Lindzen (8) in 2001 suggests a negative feedback on this system via diminishing cirrus detrainment from cumulus convection with increasing temperature. This imaginative mechanism would produce less “greenhouse heating” clouds (i.e. cirrus) with increasing temperature. In numerous calculations, Lindzen finds that the value of this negative feedback is at least equal to the positive feedback values suggested by other authors. That Lindzen can produce such a mechanism with scientifically sound arguments suggests that our knowledge of the cloud budget’s (in a sense, the relative humidity budget) control on climate is very limited.

Sherwood (6) in 1986 found that humidity values near “convection are determined more by the characteristics of the large scale flow field than by cumulus-scale behavior per se, if the ECMWF is not greatly overdoing the natural variability in these regions.” A mental note should be taken here, as authors in the future find that the ECMWF does greatly overdo the humidity in those regions (500 mb and up in the Tropics), and that cumulus convection is very important there. Sherwood worked with maps at different pressure levels, and basically found that eddies force small varying nodes of high/low relative humidity values in the subtropics via transport of water vapor.

Hall and Manabe (5) in 2000 found atmospheric water vapor anomalies to be correlated primarily to lower tropospheric temperature anomalies which they suggest is the primary reason why relative humidity values (a relationship between moisture and temperature) remain largely unchanged in most global warming model scenarios. Furthermore, they point to this “hand in hand” relationship between water vapor and temperature as key in understanding the future of greenhouse-gas induced precipitation and hydrological cycle intensification. They suggest a more precise understanding of moisture and temperature’s relationship must occur before we can begin to predict the effect global warming will have on global scale variability of precipitation.

To put a few numbers to the notions here, Soden and Schroeder (4) in 2000 show that climate models indicate a 7% increase in precipitable water per 1 degree C increase in temperature. Using the upper end of the modeled warming that current carbon dioxide increases could cause (4.5 C), the authors note this suggest an unbelievable 20% increase in precipitable water during the next half century. They point out, as others have, that sparse observations over the tropics and subtropics limit the accuracy of global or tropical water vapor trends. Importantly, Soden and Schroeder notice that at many latitudes there is a distinct fluctuation in water vapor amounts, which correlates to the El Nino/Southern Oscillation cycle. This will be shown in our analysis section via time series of the relative humidity values at various areas of the globe. Soden and Schroder also notice discrepancies in certain models/datasets during periods when radiosonde instrumentation is changed.

Lindzen (2) in 1990 hinted at the drying effects of cumulonimbus convection, and at climate model’s problems with resolving them. Lindzen notes,” Although in nature, warming increases water vapor near the ground, warming is also associated with more and deeper cumulus convection. This leads to drying of the upper troposphere. This cumulus convection occurs in deep towers of rapidly rising air. The air cools as it rises, and the water vapor in the air condenses and falls out as rain. By the time these clouds top out (at altitudes as great as 16 km), they are relatively drained of water vapor. Of course, these rapidly rising towers of air cannot exist without compensating air subsidence almost everywhere else; this subsidence acts to fill the atmosphere above about 3-5 km with dry air.” This mechanism is crucial in explaining why climate models have a hard time resolving the moisture budget in tropical areas. With little radiosonde data in these areas, and with considerable instrument lag on radiosonde hygristors at those heights, we are at the mercy of technology to provide ever better satellite observation systems.

Finally, Hall and Manabe (1) in 1998 put numbers to the notions again by showing a 3.38 C increase in temperature in a water vapor feedback model vs. only a 1.05 C increase in temperature in a model without water vapor feedback. They show the common theme of carbon dioxide induced global warming, followed by increased water vapor in the atmosphere through the Clausius-Clapeyron equation, and finally followed by increased warming due to water vapor feedback effects.

A common theme in all these papers is the lack of adequate observations in the subtropical and tropical areas of the globe. The exception to this rule is Held and Holden’s paper (9), where the authors suggest that the current observation network allows for “no surprises” in the water vapor values. This is very debatable, as it seems quite important to measure the small-scale fluctuations of water vapor surroundings cumulonimbus towers in the ITCZ. Until we have a firm observational network in place there, we can also speculate on the true distribution of relative humidity values at those locations. However, looking at the data in the NCEP reanalysis can at least give us a first order understanding of the mechanisms involved in the distribution of heat and moisture, and this understanding can show us how values of relative humidity may change in time.

First we will compare the data between the ECMWF and NCEP Reanalysis products. The mass weighted humidity box plot comparison vividly shows big discrepancies in the two datasets, especially higher in the troposphere. The results here speak for themselves, with the high level ECMWF boxes in both seasons being more wet by a factor of three. Many authors have noted the considerable wetness of the ECMWF at these heights, even when the values are compared with satellite observations (3). The NCEP data set has been shown to be roughly 5% drier than observations in this region. In reality, neither the ECMWF or the NCEP moisture data should be trusted above 300-500mb, so I limited my study accordingly.

The legend (above) shows the reader which diagnostic terms are used in calculating the humidity budget. Terms shown are various heat and moisture divergence effects on the atmospheric humidity budget. In the following plots, for example, H_DVTZM would indicate the meridional divergence of heat's effects on the atmospheric humidity (% RH/day). Terms such as DVTZM or DVQZM would simply indicate simply the meridional divergence of heat (K/day) or moisture (g/kg/day) by the zonal mean flow without calculating the effects on humidity. These terms are multiplied by "humidity factors", which calculate their effects on RH values and thus would give H_DVTZM and H_DVQZM. The humidity factor for temperature is negative, indicating that an increase (decrease) in temperature gives a decrease (increase) in relative humidity. The humidity factor for moisture terms is positive, indicating increases in moisture lead to increases in RH values, etc.

The ZM terms indicate changes resulting from zonal mean flow where as the EDDY terms indicate changes resulting from the combination of transient and stationary eddies.

The H_DTDIAB term is the summation of the H_DTCOND (change in RH due to latent heating) and the H_DTRADN (changes in RH due to radiational heating/cooling) terms. The H_DQCOND term indicates the changes in RH due to water vapor condensing into droplets (H_DQCOND down) or water droplets evaporating into water vapor (H_DQCOND up). If general evaporation is happening within a given box plot, then H_DQCOND and H_DTCOND would both go up in value since you have cooling and water vapor release adding to RH values. The opposite is true for condensation processes.

For the heat divergence terms, increases in the convergence of heat (DTDIVG, etc. up) give decreases in RH values (H_DTDIVG down). Conversely, for moisture divergence terms, increases in moisture convergence (DQDIVG, etc.) give increases in RH values (H_DQDIVG up).

***It should be noted that POSITIVE values of the divergence terms without the H factors calculated indicate a CONVERGENCE of the given parameter. In other words, DTDIVG positive indicates convergence of heat which when multiplied by the H factor for temperature gives a negative number for RH effects (as it should).*** This is a sign convention I have used to make the budget easy to decipher.

Finally, terms such as DTDIVG and DQDIVG indicate the total changes in T and Q due to divergence/convergence by the zonal mean and eddies. DTDIVG = DTDVZM + DTDV_EDDY and DQDIVG = DQDVZM + DQDV_EDDY. The zonal mean and eddy divergence terms are further decomposed into their meridional and vertical directions (see Legend and Methods for further clarification).

The following graphs shows the important heating terms for the various boxes on the globe (as indicated on the graph).

North midlatitude box: Clearly the heat budget here is dominated by the zonal mean flow adding heat in the meridional direction and removing heat in the vertical direction. However, one must look at the totals of the zonal mean and eddy heat transport terms to get a more vivid picture (DTDVZM and DTDV_EDDY respectively). Looking at those terms, it is clear that the eddies are removing heat from the system while the zonal mean flow is adding heat. The net result is indicated by the total divergence term (DTDIVG), which is forced to balance (sum to zero) with the the work term (DTWORK) plus the diabatic heating terms (DTDIAB=DTCOND+DTRADN). It appears that the small addition of heat by the divergence terms and heat release caused by condensation effects is compensated for by heat removed by radiation escaping to space.
South midlatitude box: This box has similarities to the NMID box in general budget terms, however there is more heat added by convergence (mostly by the zonal mean). The meridional eddies seems to be slightly less active for in the south midlatitudes, which seems plausible since there are far fewer stationary eddies at work there. Also, the radiation term seems a bit stronger which may indicate increased cloud cover (cloud top radiation?) at the heights we are examining.

North subtropical box: This box shows more activity happening with the zonal mean than with the mid latitude boxes. This makes sense, as we know that the descending branch of the Hadley cell is located in this general area. The net heat divergence is negative, with heat being removed primarily by the zonal mean, with a significant amount being removed by the eddies as well. Heat is being added to the system by work, indicating adiabatic descent.
South subtropical box: This box is almost identical to the NSUB box, except perhaps a little more active in general. This may be due to the assymetry of the Hadley cell about the equator acting to cause differences between the NSUB and SSUB boxes (which ARE symmetric about the equator, see graph).

Intertropical Convergence Zone (ITCZ) box: The ITCZ box indicates that heat is being added from the meridional zonal mean component and by condensation, while the heat is being removed by adiabatic ascent and by the vertical component of the zonal mean. This matches pretty well with the general concept of the ascending branch of the Hadley Cell.

NMID box: For the northern midlatitudes, moisture is converging via eddies and being removed by condensing into water droplets. Both vertical and meridional transport by eddies are important here.
SMID box: For the southern midlatitudes, the situation is the same as for the NMID box although a bit more active overall.

NSUB box: For the northern subtropical region, moisture is being added to the system by evaporation and then removed by eddies acting in the meridional and vertical directions. This means that this area is a south of moisture, mostly for the midlatitudes.
SSUB box: This box is interesting when compared to the NSUB box. It appears that the divergence of moisture is split between eddy and zonal mean effects. Together, the eddies and zonal mean remove a great deal of moisture which is compensating for by a strong amount of moisture added by evaporation. One theory would be that there are far fewer stationary eddies in the SH, thus the zonal mean takes on part of the job. However, the box indicates that the eddies are just as active as the eddies in the NSUB box, but that the zonal mean is acting in addition to them. Again, the difference here could simply be related to the position of the NSUB and SSUB boxes in relation to the Hadley Cell.

ITCZ: This box indicates that moisture is being lost by condensation and by divergence caused by meridional and vertical eddies while moisture is being added by convergence by the zonal mean in both the vertical and meridional directions. I find the moisture divergence by the eddies a little hard to swallow, however it may be possible considering the extent in the vertical of this box (up to 300 mb). In the future, I intend to plot up the full vertical contour maps (not in this paper) of the situation, to see exactly where this may be happening (up high or down low in the box).

This calculation involves the divergence of humidity itself, with humidity values taken directly from the NCEP dataset. This varies from the calculated values farther in this paper, which will consist of the terms above (the T and Q budget) multiplied by factors which show their effects on humidity

NMID and SMID boxes: Humidity seems to be diverging mostly due to eddies in the vertical direction while humidity is converging by meridional eddies.

NSUB and SSUB boxes: Humidity seems to be diverging by vertical and meridional eddies. The eddies are relatively more active for the NSUB box than for the SSUB box when compared to the zonal mean (more eddies there generally from NH topography and land/sea contrasts).

ITCZ box: Humidity is diverging by vertical and meridional eddies, and slightly converging do to meridional zonal mean flow. Humidity is not diverging as much as in the other boxes.

It should be noted that the above graphs calculate the contribution of the heat and temperature diagnostic terms to the relative humidity. In doing so, the diagnostic code multiplies the "humidity factor for moisture" (known as HFACQ; essentially 1/(saturation specific humidity) which scales with temperature squared.) times the diagnostic moisture terms at each time step in the calculation (4x daily over 26 years). The same calculation applies for the humidity changes due to temperature budget terms, where the T budget terms are multiplied by HFACT (the humidity factor for temperature) at each timestep. HFACT is essentially relative humidity times heat capacity of water divided by the gas constant for water vapor times temperature squared (or (H*L)/(R*T^2)). See the methods section for the derivation.

At first thought, one would think that the RH budget should be balanced since the heat and moisture budgets were balanced (in regards to eddy plus zonal mean divergence = total divergence). This is not the case for the above calculation, as apparently there is a strong covariance between some of the moisture or temperature terms with their respective humidity factors. One way around this confusion is to take the calculated heat and moisture budget terms and multiply them by their respective TIME AVERAGED humidity factors. In this way, the balance is maintained and allows for an easier deduction of the important terms. There is, of course, loss of information by using this method but information is only good so long as one can adequately explain the system using it. For simplicity, I'll use the time averaged method to deduce the system, however I leave the above graph for comparison. Since relative humidity is a highly non-linear parameter, I see the above graph as far too simplistic to deduce the covariances spoken of. The graph does show that the system is highly complicated and difficult to balance. It is interesting that the lack of eddies in the ITCZ allows the above graph to balance in that zone, and one must wonder whether the eddies are solely responsible for the high covariances.

As a first order approximation with boxplots, the time averaged method works quite well.

NMID box: It certainly appears from the graphs above that moisture divergence by eddies and heat divergence by the zonal mean are important in the humidity budget over the north midlatitudes. Also, relative humidity goes down due to water vapor condensing into water droplets, hence the general cloudiness witnessed in this region. The heat release from this process also lends in lowering the RH, while radiational cooling tends to increase humidity. These plots indicate that both moisture AND temperature play crucial roles in relative humidity over the midlatitudes at the heights indicated.

SMID box: This region shows a striking resemblance to the NMID box, although one could argue that the area is a bit more active dynamically than the NMID area (longer bars).

NSUB box: Since the scaling on the plots has been set to the maximum and minimum values within ALL of the plots, one must look closely at what is happening in this region. First off, vertical convergence of heat is causing the RH values to drop while meridional divergence of heat is causing the RH values to go up. Heat introduced by adiabatic compression is also happening here since we are within the descending branch of the Hadley Cell. One interpretation of this process would be that heat gained from latent heat release from the ITCZ is being moved in from the equator over the top of the box and forced downward (H_DWTZM goes down). This same air is being compressed and warmed in its desent (H_DTWORK goes down) and then is removed by tradewinds at the bottom of the box (H_DVTZM goes up). H_DTCOND is going up, indicating that this is an area of high evaporation, which we know to be the case from looking at the moisture budget previously. . As far as moisture, it seems the vertical eddy components are important in adding moisture while the meridional eddy components are important in removing moisture. The zonal mean components seem to only play a part in heat divergence where as the eddies are important in moisture divergence.

SSUB box: This box is very similar to the NSUB box but perhaps a bit more active. The work term in particular is a bit stronger than for the NSUB box, and one must wonder if this box catches more of the descending branch of the Hadley Cell than the NSUB box does (as the latitudes seem to indicate).

ITCZ box: The ITCZ shows meridional convergence of heat by the zonal mean important in lowering RH values, with vertical divergence of heat by the zonal mean important in raising RH values. Furthermore, adiabatic expansion of rising air is acting to increase RH values. The zonal mean is very important in the RH budget with respect to heat and moisture in general. Both vertical eddy and zonal mean convergence of moisture are important in raising RH values, while just meridional eddy divergence of moisture is important in lowering RH values. Also important is that water vapor is condensing to form water droplets, thus lowering the RH values. This scenario works very well with our current understanding of the general circulation of the ITCZ.

Overall differences: The first thing to notice from the above graphs are the correlations between moisture and temperature's effects on the relative humidity. The subtropical boxes have higher specific humidity values than the midlatitude boxes do, but due to the higher temperature values in the subtropical boxes, their RH values are much lower than the midlatitude box RH values. The ITCZ box contains the most moisture but the midlatitude boxes contain the highest RH values due to having the lowest temperatures.
The temperature trends are fairly constant for all boxes except for at the midlatitudes. The temperature rises a bit at the midlatitudes over the course of 26 years, which is causing the RH values to drop accordingly. On the other hand, the ITCZ and subtropical boxes moisture values fall over the 26 years while the temperatures remain quite constant, causing the RH values to drop. The net result is that all boxes RH values drop over the 26 years when taking a mass weighted calculation over their respective latitudes and heights (at least for the NCEP data presented here).

NMID and SMID boxes: More information can be gained from looking at percent deviation values of each parameters in the boxes. The above two plots show the percent deviation of the temperature, moisture, and RH values from their 26 year averages. One can see in the above trends that moisture and temperature deviations tend to be in phase (approximately) at midlatitudes. As found by many authors, the temperature shifts seems to dominate the RH budget here. For example, when the temperatures rise the moisture rises but the increasing temperature causes the RH budget to drop off more so than the increased moisture causes the RH values to drop. Peixoto and Oort showed that temperature is about 20 times more important in changing RH values than moisture is, which can be seen in these plots. The "20 times as important" part of that idea should be met with a little suspicion since the percent deviation values of temperature are very small with respect to those of moisture. Even though temperature can be 20 times as strong in dictating RH values, it rarely does so since it does not fluctuate nearly as much as the moisture values do. Regardless, in midlatitudes, temperature surely dominates the RH budget on the whole. In fact, it seems that temperature seems to guide both the moisture AND RH values here.
It is interesting, however, that the moisture and RH trends for the southern midlatitudes become more in phase in the 21st century while the temperature remains quite constant.

NSUB and SSUB boxes: The temperature swings for the subtropical boxes are bit more muted than for the midlatitudes boxes. The fairly constant temperature values allow the moisture swings to exert more of an effect on the RH budget. One can easily see that the moisture and RH shifts are more in phase for these boxes than for the midlatitude ones. In the subtropics, when temperature shows a marked deviation than the RH budget shifts accordingly. In the absence of a temperature swing, than moisture dominates the RH budget and the two parameters shift linearly (as they should). It is not an increase in moisture deviations that cause the subtropic RH values to respond to moisture more than the midlatitude RH values do, but it is the relative constant values of temperature that allow this to be the case.
Precipitous drops in RH values can be seen in the 21st century in both the NSUB and SSUB boxes. For the NSUB box, this 5 year drop is mostly due to a rise in temperatures where for the SSUB box it is due to a rise of temperatures and a drop in moisture.

ITCZ box: The ITCZ shows temperature deviations that vary very closely with past El Nino events. When these temperature deviations occur, they show a profound effect on the RH budget (for example, 1998 was a strong El Nino year). Temperature and moisture seems to take turns dominating the RH budget at the ITCZ, with the temperature and moisture changes almost in phase with one another.
The increased temperatures of the 21st century cause the RH values to diminish, and the combination of warm temperatures and low moisture in 2004 cause the RH values to drop to 3.5% lower than their 26 year average.

NMID box: The above graph shows the deviation of various diagnostic terms important in changing the RH values in the northern midlatitudes. The graph is expressed as an anomaly, which is the average value of a term for a given year subtracted from its 26 year average.
The highlights for the northern midlatitudes involve the effects on RH by heat divergence/convergence by the zonal mean (H_DTDVZM), heat gain/loss by adiabatic work (H_DTWORK), moisture divergence/convergence by eddies (H_DQDV_EDDY), and moisture gains/losses by evaporation/condensation (H_DQCOND).
One interpretation of what is happening in regards to the heat budgets effects on the RH values is as follows. When heat is diverging due to the zonal mean (RH up), than the compensating air flow comes from above and is heated by adiabatic compression (RH down). Conversely, when the zonal mean is causing a convergence of hot air in the area (RH down), than that air tends to lift and cool adiabatically, causing the RH values to rise. An interesting tidbit from the temperature graph is the extreme correlation between RH changes due to eddy heat divergence (H_DTDV_EDDY) and RH effects caused by heating/cooling from condensation/evaporation (H_DTCOND). This suggests a strong correlation between eddies and storms, which we know is the case in the midlatitudes.
Although the moisture terms do not change the RH values quite as much as the temperature terms do, it is still worth mentioning some of them. Here we see that the moisture based eddy term (H_DQDV_EDDY) is out of phase with the moisture condensation term (H_DQCOND), which points to the eddies being highly correlated with storms/condensation/clouds. One of these cycles can be explained by eddies causing divergence of moisture laden air (H_DQDV_EDDY down) which is then replaced by initially saturated descending air which evaporates water droplets (H_DQCOND up), cooling the air initially (H_DTCOND up) and then warms adiabatically once under saturated values (H_DTWORK down).
To take an example of what may be happening in the whole system (moisture and temperature combined), let's look at 1989 on the above plots. Here we have the zonal mean moving heat out causing the RH values to rise while the eddies are moving moisture out causing the RH values to drop. At the same time, part of the box experiences evaporation of water droplets causing temperatures to lower and air-born water vapor to rise, both of which increase RH values. Descending air in part of the box is warming adiabatically and causing RH values to lower.
The limitations of the mass weighted box techniques are clear here, since we cannot separate out which parts of the boxes things are happening in. The boxes are very useful, however, in picking out which global areas are experiencing significant RH changes.
As for 26 year trends within the NMID box, no terms seem to be on the rise or decline overall but obvious fluctuations can be seen over the course of the 26 years. We generally tend to see 5 to 7 year cycles (most likely in response to El Nino shifts) in the relative importance of each of the 5 terms described above. My future research will extract possible ENSO trends (El Nino/Southern Oscillation) from the 26 year trend, in order to see what other atmospheric processes may be changing the RH budget.

SMID box: The moisture budget here resembles that of the NMID box so I will look at the heat budget trends which appears a bit different. The first thing you see from the graph on the left above is that the fluctuations of H_DTDVZM and H_DTWORK tend to correlate well with the same fluctuations from the NMID box. This indicates that what ever it is forcing these terms is a global signal (tending again to point to ENSO). Another interesting thing from this graph is that the yellow curve (H_DTDV_EDDY; the effects on RH by the divergence of heat by eddies) is consistently rising throughout the 26 year period. This means that eddies in the southern latitudes have become more efficient at removing heat and increasing humidity in the last 26 years, either by advecting cool air into the region or by advecting warm air out of the region. Again, it seems from the right most graph that eddies tends to be correlated with condensed water (i.e. clouds).

NSUB box: One thing to notice from this plot is that the eddies are less active in this region than for the midlatitudes. Changes in RH due to latent heating and due to zonal mean divergence seem to be in phase, suggesting perhaps that heat removed/inserted into the box by the zonal mean is followed by heat gained/lost due to evaporating/condensating water vapor from air near saturation. Both effects act in concert to either raise or lower RH values, however they are removing or adding heat to the system in different ways.
The real evidence that this is indeed the subtropics lies in the moisture budget terms. Conversely to the midlatitudes where eddies effects on condensation/clouds is important, for the subtropics the zonal mean divergence of moisture (H_DQDVZM) and the condensation term (H_DQCOND) are closely out of phase. This suggests that eddies are not closely correlated with clouds in this region. This region is within the descending branch of the Hadley Cell, and as such we know that subsiding air via the zonal mean is very important here and that there are few eddies in general. Perhaps moist air diverging from this region via the zonal mean is replaced by descending air which evaporated water droplets as it falls into the 500 mb zone. Since there are very few high clouds in this region, this phenomena may by explained by evaporation from low marine stratus as well. Again, the box method is not perfect, but it can gave a great deal of global information quickly. A keen eye may be able to see how H_DTWORK and H_DQDVZM tend to go up a bit over the course of the 26 years. One interpretation of this is that the Hadley cell is weakening, resulting in less subsidence (RH up) and less moisture being removed by the zonal mean (RH up).

SSUB box: The southern subtropical zone shows similarities to the NSUB zone, however we see a pronounced 26 year trend in the H_DTDVZM term (change in RH due to the divergence of heat by the zonal mean). It is not obvious to the reader, but this trend would jump right out of the NSUB plot if overlaid with the same scale. Therefore, the SSUB zone has experienced a significant shift in the dynamics in the last 26 years which most likely has to do with changes in the Hadley Circulation. The SSUB box most likely catches more of the descending branch of the Hadley Cell due to the latitudes over which I calculated the boxes. Since the ITCZ tends to be a little more north than the equator in a yearly average, the SSUB box (10-35S) is most likely a better example of the desnding branch of the Hadley Circulation than the NSUB box (10-35N) since the NSUB box probably contains some information from ascending branch of the Hadely Cell during the Northern Hemisphere summer.
One can see that even though the eddies aren't very active in the temperature budget, that they do play a role in the moisture budget in this region. Further analysis of what levels these eddies are acting at removing moisture is a future goal with this work. Most likely these eddies are acting at low levels, removing moisture gained from evaporation higher in the box.

ITCZ box: The ITCZ shows that the zonal mean dominates the RH situation, especially for the temperature budget. The eddy activity in the temperature budget is just about a flat line. Basically, the heat budget says that as more hot air (most of the air surrounding the ITCZ at low levels is quite warm) introduced by the zonal mean in the bottom of the box is forced upwards, water is condensed out (releasing latent heat) but is also cooled adiabatically. This should ring a bell to the reader, as it is the age old view of what happens in the ITCZ.
The moisture budget essentially shows that when most moisture is allowed to move in to the area (H_DQDVZM up), then there is more condensation in the system (H_DQCOND down). This strong anti-phase relationship between these two terms can only exist since the temperatures at the ITCZ are very constant due to lack of significant mixing eddies.

Conclusions

The relative humidity budget of the globe is very complicated dynamically. By decomposing the budget into its intricate parts, one can gain better insight into the relative distribution of humidity. We’ve seen that eddies and zonal mean flow play a major role in redistributing heat and moisture at many latitudes. This redistribution then leads to changes in relative humidity regionally.

In the area of the tropics and the subtropics, we’ve been able to deduce that the zonal mean flow is mostly responsible in changing the humidity values, however eddies also play a role. I believe further analysis in the tropics is necessary before we can begin to accurately assess the changes in humidity that we have seen. In the ITCZ, the dynamics of cumulus convection and cirrus detrainment are not fully understood by the scientific community. More radiosonde observations in the tropics would be a good step towards resolving this issue. The ultimate goal is to obtain observations of water vapor with the same time and space resolution as those of clouds (7). Comparisons of both satellite and radiosonde data should be made, in an effort to reduce the bias tendencies caused by their respective instrumentation. Today’s water vapor observing satellites can penetrate most any cloud cover, but the thick cumulus of the ITCZ still pose problems. The more observations we can have of the tropical areas, the better we will understand the dynamics causing changes in moisture and temperature and how those changes are evolving in time.

The humidity budget results presented in this paper are no doubt hindered by discrepancies in the NCEP data, yet the trends presented show a few real regional changes in relative humidity over the past 25 years. Whether this can be attributed to global warming for certain is presently unclear. There is no doubt that many areas of the globe experience quite significant decadal fluctuations in relative humidity. If one can place a firm grasp on the relationship between global warming and the humidity budget, then one has a greater understanding of the relationship of global warming to the global cloud budget. The next logical step from there is understanding of the change in global precipitation that this may cause. Ultimately, learning how the earth’s hydrological cycle will change due to global warming is possible but it will take a better initial understanding of the relationship between atmospheric water vapor and tropospheric temperature. Also, the roles of aerosols and cloud microphysics can be introduced into the evaluation when the general dynamics are a bit more understood.

Further work should include a repetition of the work contained within this paper, but with a diagnostic set of values obtained from the ECMWF data set in order to find any discrepancies in the trends based on the NCEP and ECMWF data. Optimally, the data collected should be compared with archived radiosonde and satellite data before analysis to eliminate any major flaws in the data set. Assuming the ECMWF data presents 40 years of reasonable accurate water vapor information, one may be able to gain greater insight into the relationship of global dynamics and humidity over a longer timescale. A concentration on one seasonal period is suggested, as humidity budget calculations present a considerable amount of data. A one-season, 40-year set of data may be an easier platform on which to show global warming effects on the humidity budget solely, without bringing significant dynamic noise from global seasonal changes.

My future work in the field will include a repetition of the above work using potential temperature in order to more accurately diagnose the relevant diabatic heating terms in the system. Furthermore, I intend to use hydrogen isotopes to clearly separate areas of evaporation from those of condensation. These additional tools may aid in constructing a working model of the dynamics governing global relative humidity and the global hydrologic cycle.