Digging a Little Deeper: Spatial Variation of the Growing Season Length

Figure 1: The probability distribution function of growing season length for the Hadex2 gridded dataset.

Figure 1: The probability distribution function of growing season length for the Hadex2 gridded data set.

In last week’s post, I displayed the 110 year trend of the annual Growing Season Length using the HadEX2 gridded data set. This analysis strongly suggested that the length of the growing season has been increasing in the near-shore Chesapeake Bay region. Figure 1 is a probability distribution function of the growing season length from 1960-1985 (black line) compared to 1986-2010 (gray line). The mean length of the growing season in this time range has increased from 259.4 days to 273.1 days (or an average of 13.7 days longer).

Gridded data sets are extremely useful since they incorporate, in our case, hundreds of meteorological observations to provide a regional “snapshot” of temperature and precipitation extremes. However, gridded data sets are coarsely resolved, thus, they are an area-weighted average of each Extreme Climate Index.

Figure 2: Google Earth image of the NCDC-Daily weather stations used in this post. Yellow is Laurel, MD; orange is Princess Anne, MD; Blue is West Point; Green is Richmond Airport. The white stars represent Jug Bay, MD and Taskinas Creek, VA, which are the two CBNERRS reserves with meteorological data.

Figure 2: Google Earth image of the NCDC-Daily weather stations used in this post. Yellow is Laurel, MD; orange is Princess Anne, MD; Blue is West Point; Green is Richmond Airport. The white stars represent Jug Bay, MD and Taskinas Creek, VA, which are the two CBNERRS reserves with meteorological data.

What does this mean for the seven CBNERRS reserves central to this study?

What if we want to investigate if the Growing Season Length is changing at the same rate at the Jug Bay, MD reserve region compared to the Taskinas Creek, VA reserve region?

This is where the individual NCDC-Daily weather station data can be utilized. Below I have displayed the probability distributions functions in the same 25 year chunks for the following NCDC-Daily stations (see Figure 2 for a map of the locations): Laurel, MD (yellow), Princess Anne, MD (orange), West Point (blue), and Richmond Airport (green).

Figure 3: The probability distribution function of growing season length for the Laurel, MD NCDC-Daily station.

Figure 3: The probability distribution function of growing season length for the Laurel, MD NCDC-Daily station.

1) Laurel, MD (USC00185111)

Laurel is located in a suburban area and is geographically close to Jug Bay. Thus, Laurel is likely representative of the temperature extremes that the Jug Bay ecosystems experience. This is one of the longest meteorological records we have, starting in 1894 and still collecting data.

Probability distribution functions show the “histogram-like” shape of the data. Thus, the tallest peak of the curve displays the value observed most often. Note that this “highest density” value is often different from the mean value (represented by the vertical dashed lines).

Laurel, and by extension Jug Bay, has also had an increase in the growing season length from 1960-2010, but this change is even greater than the HadEX2. The mean growing season length from 1960-1985 was 279.5 days compared to 307 days from 1986-2010.

Figure 4: The probability distribution function of growing season length for the West Point, VA NCDC-Daily station.

Figure 4: The probability distribution function of growing season length for the West Point, VA NCDC-Daily station.

2) West Point, VA (USC00449025)

West Point is situated on the York River, which directly feeds into Chesapeake Bay. It is geographically located in the same region as Sweet Hall Marsh and Taskinas Creek (two CBNERRS sites), thus likely represents the climatic trends experienced by the habitats at these reserves.

What is particularly important, analytically speaking, is the shape of the probability distribution functions. From Figure 4, we can see that the mean growing season length has not changed much between 1960-1985 and 1986-2010. However, the variance of the data has changed, indicating that the probability of a certain growing season length has decreased. This means that the year-to-year growing season length is likely to be more variable from 1986-2010 compared to 1960-1985.

This display of data shows why comparing mean changes is not always the best approach to assessing changes in climate!

Figure 5: The probability distribution function of growing season length for the Richmond, VA NCDC-Daily station.

Figure 5: The probability distribution function of growing season length for the Richmond, VA NCDC-Daily station.

3) Richmond, VA (USW00013740)

If you can remember back to a previous post, the Richmond International Airport actually had the best correlation to Taskinas Creek. This station also was similar to the HadEX2 trend, displaying an increase in the mean between the 25-year time ranges. The mean growing season length increased by approximately 8 days.

4) Princess Anne (USC00187330)

Figure 6: The probability distribution function of growing season length for the Princess Anne, MD NCDC-Daily station.

Figure 6: The probability distribution function of growing season length for the Princess Anne, MD NCDC-Daily station.

Princess Anne is located on the Eastern Shore of Maryland in a more rural area between Chesapeake Bay and the Atlantic Ocean. From Figure 6, we can see that the mean and the data spread (variance) of the growing season length at this location has not changed much over the 50 year time frame.

I thought it was important to show this site since it demonstrates how complex climate really can be! While almost all sites and data sets in the Chesapeake Bay near-shore region show an increase in the growing season length, Princess Anne has not seen the same effects. If anything, the mean growing season length has actually decreased by 2 days.

To conclude: 1) The mean is not always the best way to compare changes in time series data; including changes in the variance can be just as important to ecosystems. 2) Growing season length changes do not occur at the same rate everywhere in Chesapeake Bay, but the vast majority of the data sets we are using follow the same trend of an increased growing season length, justifying the use of the HadEX2 data set.

Kari Pohl

About Kari Pohl

I am a post-doctoral researcher at NOAA and the University of Maryland (Center for Environmental Science at Horn Point Laboratory). My work investigates how climate variability and extremes affect the diverse ecosystems in Chesapeake Bay. I received a Ph.D. in oceanography from the University of Rhode Island (2014) and received a B.S. in Environmental Science and a B.A. in Chemistry from Roger Williams University (2009). When I am not busy being a scientist, my hobbies include running, watching (and often yelling at) the Boston Bruins, and taking photos of my cat.
This entry was posted in Atmospheric Temperature and tagged , , . Bookmark the permalink.

2 Responses to Digging a Little Deeper: Spatial Variation of the Growing Season Length

  1. Victoria Coles says:

    Interesting look at the spatial differences in growing season length. I thought maybe this map from the USDA on growing zones might explain some of this pattern. Princess Anne and Laurel are pretty close to each other geographically, but they have slightly different zones on the USDA map. Growing season is about the same length at Princess Anne and Laurel today, but in the past, it was shorter in Laurel… Hmmmm.

    http://planthardiness.ars.usda.gov/PHZMWeb/InteractiveMap.aspx

Leave a Reply

Your email address will not be published. Required fields are marked *