Test the accuracy of predicted 2D-data I am trying to find a good way to test predicted time/space data with the actual data. I sketched my data in the picture below. How would I go on to test the accuracy of the predicted data points to the actual measured data points?
I started by calculating the euclidean distance of each predicted point to their corresponding measured point and calculated the (N)RMSD.  
Is this the way to go? What other options are there?

 A: Since time and distance happen on completely different scales, you are almost literally comparing apples and oranges. Your accuracy measure will change, for instance, if you measure time in seconds or in minutes. So you will definitely need to account for this. You can either use weights in your Euclidean distance (with weights changing for different units of measure), or you could standardize the "component" distances, e.g., by expressing them as multiples of the in-sample standard deviation.
In addition, there are of course many, many different notions of "distance" that may make more sense in your specific use case, e.g., the different p-norms. (See above on combining these.)
You can look at measures of "regular" forecast accuracy. Many of these can be adapted to your 2-d case, specifically all that involve terms of the form $\hat{y}-y$, that is differences between forecasts and actuals. You would replace all these terms with appropriate distances $d(\hat{y},y)$ as per the previous paragraph and then summarize in some way, e.g. using (weighted) sums. 
You may want to look at Vector Autoregression (VAR) and skim some literatur to see what kind of accuracy measures are used there.
Finally, your specific example picture shows forecasts that are biased upward in both dimensions: every single forecast is too high for both time and distance. This suggests that it might make sense to also look at one-dimensional marginal errors and check for bias there, e.g., using simple Mean Errors (ME). (Higher dimensional error measures won't be able to capture that.)
