If I use the function peakdet (http://www.billauer.co.il/peakdet.html) I can find the maximums and minimums of the plot. I can then find the first minimum and the maximum value that follows it. After I make sure that the length of the two vectors match, I find the difference between the subsequent maximums and minimums. Where there is an upward trend, the variability of the differences is much higher than when there is a downward trend. Basically, when there is a upward trend there are a lot more data points between maximums and the minimums behind them. Whereas on downward trending data, there are typically very few (mostly no) data points between maximums and the minimums preceding them.
You can use the data I supplied earlier to test this out, swapping 'upward' and 'downward' as needed.