1 Uncertainty in Extrapolations of Predictive Land Change Models R Gil Pontius Jr Joe Spencer Prepared for presentation at the Open Meeting of the Global Environmental Change Research Community, Montreal, Canada, October, 2003.
2 Major Points Validation is a waste of time, –unless you use the validation statistic to express the level of certainty of predictions of the unknown. The prediction’s accuracy approaches random as the prediction’s time interval grows. –We estimate how fast the accuracy approaches random.
3 Worcester Massachusetts and nine surrounding towns
4 Strategy of Three Runs
5 Percent Built versus Geology 1971
6 Percent Built versus Geology 1985
7 Percent Built versus Geology 1999
8 Percent Built versus Slope 1971
9 Percent Built versus Slope 1985
10 Percent Built versus Slope 1999
11 Real Built 1971
12 Simulated Built
13 Real Built
14 Observed Accuracy Run 1: % 6%
15 Estimated Accuracy
16 Strategy of Three Runs
17 Expected Accuracy
18 Expected Accuracy Run 2: % 6%
19 Real Built 1985
20 Simulated Built
21 Real Built
22 Accuracy Run 2: 1999 Observed 1% Quantity Error 9% Location Error Expected 4% Quantity Error 6% Location Error
23 Expected Accuracy Run 3: % 9%
24 Simulated Built
25 Expected Accuracy
26 Expected Accuracy
27 Major Points Validation is a waste of time, –unless you use the validation statistic to express the level of certainty of predictions of the unknown. The prediction’s accuracy approaches random as the prediction’s time interval grows. –We estimate how fast the accuracy approaches random.
28 Method is based on: Pontius Statistical methods to partition effects of quantity and location during comparison of categorical maps at multiple resolutions. Photogrammetric Engineering & Remote Sensing 68(10). pp PDF file is available at or National Science Foundation funded this via: Center for Integrated Study of the Human Dimensions of Global Change Human Environment Regional Observatory (HERO) Plugs & Acknowledgements