what statistics should I use for demonstrating two data points I want to plot a time-series of emissions of a pollutant. For each year, I only have two data, one min value and one max value for the emissions.
I only demonstrated the min, max and mean value in the time-series figure, but my boss said it was not enough. Should I use mean plus standard errors or Confidence Intervals or SD or others? and why it is better than just demonstrate min-mean-max?
 A: If you have JUST the min and max, there is literally nothing you can do per year except plot those values. Even the mean calculated from min and max will be meaingless and misleading if any data driven bussiness decisions are to be made. What you could (and I'm not saying should) do is regressional analysis (trendline). This would show and give a rough prediction whether those values are going to increase or decrease in the future.
If you could obtain more data than just min and max, please consider doing that. You can remind your boss that incomplete data leads to bad decisions and financial loss.
A: The problem with minimum and maximum, is that they may not represent well the distribution of the variable.
Consider for example a sample of size 25 of variable x that looks like this (when we present the observations from small to big):
20,25,28,29,29,32,32,35,37,37,38,40,41,41,44,45,46,46,46,48,51,53,71,74,212
Then the mean is 48, the minimum is 20, and the maximum is 212.
If someone would tell you these 3 numbers, you would never guess that 212 is just a single observation that is irregular, you would think that the distribution is something like this:

While in fact the distribution is not like that, there's just one very non-representing observation (212) that caused the maximum to be as it is. 
We call such observations/values that are very different from the others, and don't represent the data, "outliers".
When you have many observation (e.g. hundreds), then there is usually a high probability that the maximum and/or minimum will be outliers that don't represent the data.
One approach to this problem is to draw the mean with a confidence interval for the mean, which can be calculated by taking the mean and adding/subtracting (for the lower and upper bound of the interval, respectively) the standard deviation multiplied by 1.96 (for 95% confidence level). E.g. if the mean is 10 and standard deviation is 2, then the confidence interval is [10-2*1.96 , 10+2*1.96].
Another approach, which is even more robust (is not affected easily by outliers) is to draw the median, and the interquartile range. You can draw the mean as well.
In R, the boxplot function creates a box plot such that the box represents the interquartile range, a line inside the box represents the median, and observations outside the interquartile range are drawn as points.
A: If you have the maximums for several observations, those maximums actually follow a GEV distribution. 
