Table 4. Percentages of geographical areas (départements for the dataset France, and countries for the global dataset) where the model DLMs has a lower RMSEP than the model DLM0.
| % of cases where DLMs is more accurate than DLM0 over | ||||
| Dataset | Time lag (Year)a | all areasb | areas with low yield increase ratesb | areas with high yield increase ratesb |
| France | 1 | 21.7% | 13% | 30.4% |
| France | 10 | 32.6% | 15.2% | 50% |
| Global | 1 | 28.4% | 7.8% | 49% |
| Global | 10 | 29.4% | 15.7% | 43.1% |
a
RMSEP was computed for predictions one-year ahead and ten-year ahead.
b
Percentages were computed over all geographical areas, over areas with low estimated increase rates (i.e., départements/countries showing an increase rate lower than the median of the increase rates estimated with DLMs in 2010), and over areas with high estimated increase rates (i.e., départements/countries showing an increase rate higher than the median of the increase rates estimated with DLMs in 2010).