Is Houston’s Summer Climate Changing?

Climate change is expected due to increasing CO2. Is there evidence that climate change is happening in the Houston, Texas area’s warmest months?

“Warmest months” is defined as the warmest six months (May thru October). Houston’s weather is dominated in this period by warm, moist flow from the Gulf of Mexico. Weather patterns tend to be redundant – warm, humid, with a chance of showers. There is variation, though, with summer sometimes experiencing periods of drought when upper-air anticyclones dominate.

“Houston area” is defined as NOAA’s Texas Upper Coast climate division (light blue, below), which consists of 13 coastal counties roughly centered on Houston.

Texas Climate Divisions

First, let’s look at the maximum temperatures since 1895. The data source is NOAA’s NCDC’s Climate at a Glance plotter . The maximum temperatures show variation over time, with a warm 1930-1950s and cooler 1970s and a rise in recent decades, but the direction is not clearly higher. The recent rise might be natural variation only or natural variation plus a CO2-induced rise. Visual examination alone does not provide a clear answer.

Houston Max Temp. in Warm Months

Might the variation in summer maximum temperatures be related to soil moisture? It’s established that there is a relationship between soil dryness and maximum summer temperature. I believe the relationship is due to the lack of heat removal by vaporization and possibly because the weather regime which produces droughts in Texas also produces sunny skies and warm air aloft, hindering convective heat removal.

Let’s use a measure of drought (Palmer Modified Drought Index, PMDI) and compare its variation with summer maximum temperature. When comapred, an r-squared value of 0.4 is found, a fairly good correlation. Over the last 30 years the r-squared value has been 0.52.

Below is a plot of the standard deviation of Houston PMDI and max temperature which illustrates the relationship:

Houston Max Temp and PDMI

An r-squared value of 0.52 over the last 30 years it’s fair to say that most of the recent rise in maximum temperature can be explained by drier warm-weather conditions.

What about minimum temperatures in the warm months? This could be interesting, as the effects of increasing CO2 might be more noticeable at night, when radiative cooling dominates heat transfer. Here’s a plot of Houston area minimum temperatures during May – Oct:

Houston Min Temp in Warm Months

Well, this indeed looks interesting. It appears directionless until the 1980s, at which time it rises to record values. Also interestingly, it bears a resemblance to the global average annual temperature pattern. It’s not correlated to drought (soil moisture), as the r-squared relationship between min temperature and PMDI (not shown) is essentially zero.

My surmise is that this temperature rise includes effects from increasing CO2. The effect might be direct (slower IR loss at nighttime) and/or indirect (higher moisture in the lower troposphere which increases both dewpoint and late-night low clouds). My impression from living along the US Gulf Coast for decades is that the summertime low temperature is mainly determined by the dew point – near the dew point the cooling from IR radiation generates dew rather than lower temperatures.

How about the diurnal temperature spread (max – min)? Here’s a plot of the spread since 1895:

Houston Temp Spread

The plot shows a general decline over time, which is generally expected with CO2-driven climate change. However, the decline seems to occur in one step about 1960. I have no explanation for that. The step decline could be random, could be an artifact of the temperature record (which includes adjustments) or it could have a physical basis. It’s something to ponder.

It’s also worth noting that temperature spread is fairly correlated with drought. The r-squared for temperature spread versus drought index (1895-2014) is 0.42. A correlation makes sense as drought conditions are generally associated with less atmospheric water vapor and less water vapor means faster radiation of heat (IR) from the ground into outer space.

Finally, what about Houston area rainfall? Here’s a plot of total May-Oct yearly rainfall since 1895:

Houston Precip

The plot shows variation over the years, with perhaps a dry 1930s and wet 1970s and ’80s. The 1950s were also drier than average. The second half of the period is wetter, on average, than the first half. I see no clear trend which would correspond with rising CO2.

So, where does all this leave us? My view:

Maximum temperatures – no CO2-related trend

Minimum temperatures – rising in recent decades, plausibly due to higher CO2, in whole or in part

Max-Min Spread – no clear CO2-related trend

Precipitation – no CO2-related trend

Posted in Uncategorized | Leave a comment

Is Heavy Rain Increasing in the Houston Area?

Very heavy rain occurred in the Houston, Texas urban area on Memorial Day 2015. Several people drowned and property damage was substantial in a small part of the city. Was this a random weather event or was it part of a trend toward more-frequent heavy rain, possibly driven by global warming?

The US Historical Climatology Network (USHCN) provides convenient online data from climate monitoring stations scattered across the US. Unfortunately, there are no USHCN sites in Houston but there are three climate sites nearby: Liberty, Danevang (love that town’s name) and Brenham, located east, west and northwest of Houston, respectively. Here are the mapped sites, with Houston visible as the lighter-colored region in the middle:




I looked at daily rainfall data for the three sites for the period 1951 thru 2014, a 64-year period. This covers the time when atmospheric CO2 strongly increased.

First, for background, let’s look at annual rainfall at the three sites. Below is a plot of annual rainfall at the three sites (combined) since 1951. Southeast Texas has an east/west gradient in annual rainfall, with eastern areas getting more rain on average. Liberty, roughly 100 miles east of the other sites, receives on average about 30% more rain than do Brenham or Danevang. The three sites (combined) average about 46 inches a year, fairly typical for the eastern US. The plot of annual rainfall below shows considerable interannual variability.
Of interest is that there’s no clear trend in annual rainfall. Totals generally rose after the dry 1950s but have trended downwards since the 1980s. This lack of a trend is confirmed by rain data for the “Upper Texas Coast” from NOAA’s Climate at a Glance website. The data shows variation but no trend for the last 120 years.
OK, what about heavy rain events? For this review I defined a “heavy rain” as 4 inches or more in a day. For Houston, that amount in a day typically causes street flooding and motorist inconvenience but poses no risk to life or property. I summed the heavy rain events (Liberty + Brenham + Davevang) for each day since 1951 and then totaled the events over three year periods. A plot of total heavy rain events over the past 36 months gives an indication whether the frequency of heavy rain is growing, shrinking or is trendless. Here’s the plot since the early 1950s:


Here’s the same data except starting in the mid-1980s:


There’s no clear trend. There might have been an increase as the region emerged from the relatively-dry 1950s but any increase disappeared in recent decades. It’s doubtful that a linear trend would be evident in any case, as the atmosphere tends to oscillate over these time scales rather than trend in a straight line.

The review used 4-inch or greater rainfall days. What if the threshold was 2 inches instead of 4 inches? Here’s the plot”

0716152Here, too, there is no clear trend. Whatever uptrend occurred as the area emerged from the relatively dry 1950s seems to disappear over the last decade.

How about 6+ -inch rainfall days? The plot is below. As expected, there are few such heavy rain events, a fact that complicates any search for trends. A linear trendline slopes upward but there might have been a step upwards about 1980 (the time of the PDO shift to its warm phase) with little change thereafter.

If the 65-year trendline is accepted as real then it means that the 6-inch rains per site have increased by 0.1 – 0.15 per year, or about one additional event per 9 years. The increase largely occurred over 30 years ago.



Now for a side issue: if overall rainfall increases but the distribution (big, medium, small) of rainfall events is unchanged, then it seems reasonable to expect that the count of each sized event would increase. So, what if an adjustment was made to account for the change in total rainfall? Below is a plot of the count of 4-inch rain events in the prior 36 months divided by the total rainfall in the prior 36 months:


And 6-inch rains adjusted for total precipitation:




There are no clear trends in heavy-rain, especially in the last 30 years. The practical effects of any increases, if they exist, are weak. Perhaps, over a longer period, increasing CO2 will drive increasing heavy rain events in the Houston area. There’s no clear evidence of that happening today.

Posted in Uncategorized | Leave a comment

Is Global Precipitation Changing? The reanalysis view.

Precipitation is an important way that Earth “keeps its cool”.  Precipitation processes account for about a quarter (estimated 80 W/m3) of the 340W/m3 our planet must remove to avoid warming.

For perspective, contrast that 80W/m3 with the 4W/m3 impact of doubling CO2 (ignoring CO2’s possible knock-on effects). Precipitation, besides this direct heat transfer role, places water vapor, a greenhouse gas, into the mid- and upper-atmosphere, affecting IR radiation from lower levels. And, condensed water vapor (cloud) plays a role in Earth’s reflection of a portion of incoming sunlight as well as a portion of upwelling IR. A change in global precipitation processes could, in concept, affect Earth’s heat removal effectiveness in several ways. That change, given the various roles of precipitation, could be significant.

How has Earth’s precipitation trended? Below is a plot of global precipitation rate since 1979 (per ERA reanalysis data). As it indicates, Earth saw a mostly downward trend until around 2005, at which time an uptrend occurred. In the last several years that uptrend seems to be reversing. In absolute amount, the ERA data (1979-2010) averaged 2.92 mm/day (1979-2010).  It’s important to note that precipitation estimates, whether by ERA, other reanalyses or TRMM (a sophisticated measurement program), are all estimates – there is no single unquestionable source for historical global rain and snow amounts.


L et’s take a look at the trends in the tropics and extratropics. Below is a plot of trends in the tropics (20N-20S) as well as the poleward areas north and south. The three areas are of about the same size.  What stands out is the large role of tropical rain. My analogy is that the tropics are like the troposphere’s “furnace”, providing considerable heat for the more poleward regions, with significant heat exhausting from the tropics to the midlatitudes via the middle and upper troposphere.


What about trends? Below is a plot of the precipitation anomalies in the three regions (annual cycles removed), based on ERA reanalysis data. The most significant feature (based on visual appearance) is the ramp-up that began about 2006. and apparently peaked four or five years later. That increase, roughly 3 or 4% in global precipitation, equates to an extra 2 or 3 W/m2 of latent heat moving from the surface into the troposphere. (I realize that my last statement involves many assumptions.) That increased energy movement is of a magnitude similar to the likely to-date effect of the CO2 added to the atmosphere to-date. No causation is implied – rather, mine is just an observation.


Finally, let’s take a look at data from TRMM, a high-quality Japan/US program which studies tropical rainfall. Below is TRMM’s map of global rainfall changes over the last 35 years. It shows an increase in the tropics (West Pacific and Indian Oceans) and some decrease in the central tropical Pacific (which I think of as the El Nino region).


Here’s a TRMM plot  of tropical ocean precipitation anomalies (red) since 1996. It seems to show an increase 2007-2011 followed by a leveling. At a crude level that is similar to the ERA pattern but certainly differs in detail. 


Below, simply for reference, is a global map of the precipitation observed by TRMM, with the three latitude bands marked:


In summary, there is evidence that tropical rainfall has increased over the last decade, especially in the very warm Indian and West Pacific Oceans, the mid- and upper-troposphere’s “furnace”. Characterization of the mid-2000 shift to increasing precipitation is worth pursuing.

Posted in Uncategorized | Leave a comment

Is Amazon Rainfall Changing?

This is a look at ERA rainfall data for the Amazon region (as I define on the photo below). The area is a proxy for the Amazon Basin, meaning that it doesn’t precisely conform to the basin but it does encompass most of the Basin and not a lot of non-Basin area.


Ecologists and others are concerned about the effects of deforestation in the Amazon region. So, just for exploration, I’ll plot both rainfall and deforestation.

Let’s start with basics. Here are plots of monthly rainfall since 1979 and of rainfall by month of the year:


Rainfall in the Amazon region has trended upward over the last 35 years. The trend has been irregular but substantial – average rainfall today is about 20-25% higher than 35 years ago, based on this ERA data. Wow.

Using the region’s average rainfall and size, and assuming that 78 W/m3 of energy leaves Earth via precipitation processes (per IPCC), about 1.4 W/m3 of global heat removal is associated with the Amazon region rainfall. A 25% growth in rainfall means that 0.35W/m3 has been added to global heat removal (everything else being equal, which might or might not be true).

The next plot shows rainfall by month. It indicates substantial intra-annual variation (wet season and dry season), possibly due to the annual travels of the ITCZ (intertropical convergence zone) across the region.

0131153Let’s see if the 35-year trend varies by wet season and dry season. Below is a plot of the trend for the average of the wettest three months and of the driest three months. It shows that both seasons have gotten wetter , with the wet season’s growth happening a little faster than the dry season’s.


That variation over the year deserves a little closer look – is the rainfall increase only a wet-season phenomenon? Below is a plot of the slope of the linear trendlines for each month of the year. There is a clear seasonality to the rates of change, with the wet-season showing the greatest rates of change. That makes me wonder if the ITCZ (to which I attribute the wet season, to be explored) is more active nowadays than in the past. If so, why?


Now, a look at deforestation.  Below, in red, are estimates of the decline in Brasilian Amazon forestation as a percent of 1960s forest cover. The plot also shows rainfall trends in the Amazon region. All that can be said from this is that forest coverage declined while rainfall trended upwards: correlation of trends is not necessarily causation.

There’s a considerable amount of peer-reviewed literature on the possible impacts of deforestation and climate change, much of it based on modeling. One paper speculates that reduced forest cover means reduced transpiration and thus less water vapor. Others consider the possibility that there might be an impact on ground-level wind or albedo. I believe that we’re far from knowing the answer.

If there actually is a relationship (less forest cover leads, say, to a wetter wet season) then the latent heat released by that greater rain might be large enough to alter regional weather patterns elsewhere on the globe.


Posted in Uncategorized | Leave a comment

Tropical Precipitation and ENSO

Let’s take a look at how tropical precipitation (as reported by ERA) behaved in two larger ENSO events. One event happened during 1996-2000 and the other 2007-2011. Both periods included a large El Nino event and a significant La Nina in short order. Here’s an ENSO time series, with the two periods circled:


I’ll break the Indo-Pacific into two regions and call them “A” and “B”. “A” covers part of the Indian and Western Pacific Oceans. This could be considered the core of the Indo-Pacific Warm Pool.

“B” is in the central and eastern Pacific and experiences sea surface temperature changes associated with ENSO changes. B is not far from the downwelling leg of the Pacific Walker circulation where large amounts of air radiatively cool and sink. Its behavior during an ENSO event might be more complicated than area A’s.



First, let’s look at how the two areas behaved in 2007-2011. Here’s a plot of area B in 2007-2011. I have added the Oceanic Nino Index (ONI) which is a measure of the strength of ENSO events.

B’s precipitation tended to rise as the Pacific moved into El Nino conditions (indicated as positive ONI values) and declined as the area switched back to La Nina. The calculated correlation (r2) between rainfall and ONI over the five years was about 0.5, which is respectable. The area behaved in a reasonable way, as warmer sea surface temperatures means more evaporation and water vapor (the “fuel” for thunderstorms) and perhaps a shift to a less-stable atmosphere. Nothing surprising to me.


How did area A act? Here’s a plot of area A rainfall and ONI:

0127151Not much of a relationship. At least in this period, precipitation in A didn’t respond to  ENSO activity.

(A comparison of rainfall and area A SST (not shown) gave an R2 of .02, which is poor. That surprises me a bit as I expect higher SST in a rain-prone region to lead to higher rainfall and vice-versa.)


This period is interesting. The super-El Nino in 1997-8 and subsequent strong La Nina are in a period when there appears to have been a step-up in global temperatures. That step-up has been followed by an extended time when global temperatures, on average, have not risen.

First, region B, the area where ENSO behavior is most pronounced. As the plot below indicates, there is a strong correlation between rainfall and ONI. This is consistent with the behavior of the 2007-2011 period and with the general description of ENSO events. So, the reanalysis data in this overwhelmingly-unmeasured region fits my preconceived notions on how the global atmosphere works.


Now, a look at region A (the western Pacific). Unlike 2007-2011, there is actually a bit of a correlation with ENSO, with a noticeable depression during the super – El Nino of 1997-1998 and some elevation in the adjacent La Ninas:


That’s interesting and raises a question for further exploration – did the relationship between ENSO and region A rainfall change at the time of the hypothesized “climate shift” at the end of the 20’th century? Inquiring minds want to know.

Overall, my confidence in the reanalysis rainfall data in remote regions is OK, at least for the moment. Next I’ll look at broader trends in the tropics and see if they fit my existing layman’s understanding.

Posted in Uncategorized | Leave a comment

Tropical Precipitation

The website KMNI Climate Explorer is an excellent source of climate data. The site offers access to a wide variety of data in user-friendly ways.

Among Climate Explorer’s many offerings are daily precipitation estimates.  The word “estimate” is important as estimates are all we have for remote, sparsely-measured regions such as tropical oceans. One estimate method is known as “reanalysis”. Reanalysis has limitations in accuracy so its data must be viewed with the proverbial “grain of salt”. Nevertheless, it can be helpful.

The daily reanalysis data I use in this post is known as ERA-interim, offered by the highly-respected ECMWF . My focus is on precipitation in the tropics. The area I chose (15N to 15S) covers about 25% of Earth’s surface. Here’s the rainfall plot covering the last 25 years (1990 through 2014):


The plot shows is a regional average of about 5.1 mm/day (about 80 inches of rain a year). That accounts for roughly 50% of global precipitation, which is a significant amount.

Importantly, tropical precipitation is a process that transports energy (heat) from the surface into the mid and upper troposphere, energy that is important to maintaining the global circulation and to warming latitudes away from the tropics.

Interestingly, precipitation processes are thought to remove about 80 W/m3 of energy from Earth. Since tropical precipitation (15N-15S) accounts for about half of global precipitation we can reasonably guess that tropical precipitation is involved in roughly 40W/m3 of global heat removal. That’s an order of magnitude greater than the expected warming impact of doubling CO2 (4W/m3, assuming no feedback). It is conceivable that relatively small changes in tropical precipitation could affect global heat transport and/or the warming rate, at least in the short term.

Here, using monthly rather than daily data, is a plot of the tropical precipitation anomalies since 1979.  The change over time is interesting and worth later exploration. The decline in tropical rainfall seemed to end about the time of the 2000 “climate shift” and switch to a rise. The period of decline roughly corresponds to the period of rising global temperature and the period of rainfall increase to the “hiatus”. Relationship? I have no idea.


In future posts I’ll explore some of the wiggles and shifts in tropical precipitation. Experience outside of climate science has taught me that much can be learned by understanding the finer details and characteristics of data – what might be thought of as noise can actually be meaningful. My major goals are to improve my limited understanding of how the global atmosphere works and to see if there are patterns which could have predictive value.

Posted in Uncategorized | Leave a comment

Intraseasonal Oscillation “X” (“IOX”)

Here’s a puzzle. The puzzle involves satellite-derived tropospheric temperature estimates. These temperature estimates are created from AQUA satellite data. UAH offers the daily data at their website. AQUA’s channel 5 data is thought to reasonably represent the bulk of the troposphere.

The puzzle is an intraseasonal oscillation in the data which I’ve been unable to identify – Intraseasonal Oscillation “X” or “IOX” for short. My hope is that someone can identify IOX.

Here’s the story. The channel 5 data varies over time, mainly due to annual cyclicity and ENSO-related effects. It is possible to (mostly) remove the annual cycle and ENSO effects, as explained here , leaving the higher-frequency variability.

Below is a channel 5 time series example (2005-2006) in which the annual cycle and ENSO effects have been removed, leaving the higher-frequency variability. These values (the daily anomalies with ENSO removed) I term as “adjusted daily anomalies” or “ADA” in this post.

This time series’ appearance is interesting in several ways. One, the tropospheric temperature changes are often rapid for several days – sharp rises and equally sharp falls. Two, the rapid changes sometimes come to an abrupt stop, pause for several days then rapidly move in the opposite direction. Three, despite the rapid changes, the temperature anomalies generally stay within a band of+0.2K to -0.2K. These appearances warrant exploration, but not in this post.

What would this ADA time series look like if the very high-frequency variability was also  (mostly) removed? To explore that, here is a plot of 15-day moving averages of ADA (“adjusted daily averages”) for 2005-2006:

This smoothed channel 5 data appears to be oscillatory. Typically, about every 25 to 40 days, the channel 5 data warms and then cools by around 0.1 – 0.2 K.

Is this oscillatory appearance confined to the example 2005-2006? No. A similar plot covering eight years (2003-2010) is here . The 2005-2006 data is representative of the full period.

Regarding the period of the oscillation, a plot of days-between-peaks for 2003-2010 is here . Most of the days-between-peaks are between 20 and 45 days, a bit shorter than the 40 to 50 day oscillation period generally associated with the Madden-Julian Oscillation (MJO).

Perhaps this apparent oscillation is not real. Perhaps it is a data artifact or the consequence of gathering oxygen radiance data by satellite. Perhaps the simple approach is fundamentally flawed and I am seeing ghosts.

Or, maybe it is real. If IOX is associated with a real, known tropospheric phenomena, which one?

The physical nature of such an oscillation, if it is real, is  intriguing. Some of the questions which cross my mind are here, as well as a list of possible oscillations.

Here’s a possible clue about IOX’s identity (or maybe it adds to the mystery): suppose that the 2003-2010 smoothed ADA time series is broken into its annual sections, so that the oscillation’s behavior over the course of a calendar year can be seen. Here is the plot of that, covering 2003 thru 2010. The plot looks disordered, like spaghetti.

Suppose, though, that the oscillations for each calendar date are averaged. This average is an indication of the relative amplitude of the oscillations for each calendar date. Here is that plot:

Figure 5

This plot seems to show that ADA amplitude varies in a rather orderly way over the course of a year. An amplitude peak is realized about every 21 days, with a (usually) distinct trough in between. This is unexpected and mystifying to me.

Also, the black (moving average) line indicates that amplitude tends to peak in midsummer and midwinter and go through a minima in spring and again in fall. This is not so surprising – global temperature/energy differences are probably less during an equinox than in a solstice.

What in the world (er, atmosphere) is all this about? If it is real then how does this oscillation physically correlate with the calendar? Why 21 or so days between amplitude peaks (a plot of the count of days between maxima is here )? If the oscillations are due to changes in tropical thunderstorm activity, as in MJO, then what is this 21 (or so) calendar-based amplitude oscillation about?

The Excel spreadsheet containing the data, calculations and plots is here

So, what is X?

Posted in Uncategorized | 1 Comment