I was trying to plot the global (spatial) distribution of a certain parameter (let's say temperature or precipitation). What I found that plotting a time averaged figure would not be effective in my case. I want to understand how a certain parameter is changing globally/ regionally on a temporal scale. Can any one suggest me any idea or plot type for such analysis?
Dear Mrinmoy
Hovmoller plot can be used to see the variation of Met parameters along latitudes or longitudes and there is an R package to plot this. I am sharing the R package link
https://cran.r-project.org/web/packages/rasterVis/rasterVis.pdf
Dear Rohith
Hovmoller is a good option. Thank you for the suggestion. But, this also provides an average of a particular array of pixels in a line. Instead of averaging, can we have some other option? Still, I will try once with your suggestion. Thanks again.
Dr. Towe,
I want to plot the global change of a climate parameter in a temporal scale without temporal or spatial averaging. Although, I have a plan to use GIS shapefiles to divide the global area into small parts or zones (like oceans and continents) and make the average of the concerned parameter for each part. But, it seems, zonal averaging will not demonstrate a rigid understanding of temporal changes in the global domain.
Having done some GIS analysis, and if I understand your question and goals about mapping in this case, you will always face an issue with averaging. For example, say you have a temperature measure. That measure covers some amount of space. Depending on where you make measurements, there is always going to be spatial variation in the temperature measure which could have a lot of parameters depending on the size of the space. For example, I take a temperature measure at Point X. How far away (X + n) does that measure reach? Are there other climate factors within the space parameter (X + n) that could affect the temperature? In other words, the temperature measure for a space is already being defined as an average for the space, even if you have multiple measures within that space. If there is one measure, you are assuming it has some space scope, and it is the average for that space. So, this can't be avoided at some level of conceputalization or analysis.
So even if you are using small space (ie. as in a hotspot analysis), your measure is still representing an average for that space, UNLESS you can argue that the temperature is the true measure of the actual temperature in that given space.
Now, across a given space (X) over a period of time, you could assess changes in temperature. I'm not really sure of the method here as applied to climate change, but NASA has a visual example, using radar images ....https://earthobservatory.nasa.gov/GlobalMaps/view.php?d1=TRMM_3B43M&d2=MOD13A2_M_NDVI
How you would model that statistically -- I'm not that smart!
I often use the free disease surveillance software SaTSCan (www.satscan.org). It allows several different modeling options, including a test for space-time interaction. Get the mean value(s) for each time point, and create a plot that you can use to better understand the results from the space-time analysis. It can happen that what appears to be a space-time interaction in fact is caused by spatial changes. This can be looked into via a nonparametric permutation test in SaTScan. If the permutation test is significant, then the space-time interaction is not driven by purely spatial causes.
I hope that this information in helpful to what you have asked about here.
Dr. Towe,
The data I will be using are basically multi model simulated data for hundreds of years (annually averaged data). Here, we have a list of parameters which are simulated using coupled earth system models in the global domain. Now, I wanted to study a particular region for a particular season, say the temperature distribution in boreal summer in India. Usually, the studies report a decadal average or a range of years of simulated values for that particular season for that particular region. As Mr. Rohith mentioned about Hovemoller plot (https://www.youtube.com/watch?v=ox65V6WYgeg), I wanted to visualize a climate parameter which changes in three dimensions (lat, long and time).
You may try to plot spatial distribution of trend of the selected parameter. Temporal trend is an expression of temporal change. you need to calculate first the temporal trend of the parameter per grid, and then display spatially in a single map.
Something like this:
http://www.mdpi.com/2225-1154/5/1/18
Is this kind of map what you mean?
You may try to use box plots. In the box you put your spatial data and you create a box for each year. You can do this for the ensemble average and check is there is a temporal trend. You can use a similar plot with each model's spatial average and assess the uncertainty associated to the climate change projection
- Badges
- Science topic
- Similar topics
- Geoscience
- Atmospheric Sciences