Table 2. Sampling targets for forest percentage cover.
Using our results on upscaled forest species richness, it is possible to estimate the percentage ppred% of the forest that must be sampled to achieve an estimation error of approximately 5% with a certainty of 95%. We derived these values by creating the predicted forest at the whole-forest scale (we generated Spred numbers according to an NB with parameters r and ξ) and sampled it at increasingly larger scales until the desired accuracy in the estimation of the global species richness was reached (see section S7 for more details). The last column indicates how much extra sampling is needed (if the number is greater than 1) to reach 5% precision.
| Forest | ppred% | ppred/p* |
| Amazonia | 0.0003 | 1.875 |
| Barro Colorado | 3 | 1 |
| Bukit Barisan | 0.05 | 18 |
| Bwindi | 5 | 386 |
| Caxiuana | 0.01 | 0.55 |
| Cocha Cashu | 0.003 | 8.57 |
| Korup | 0.02 | 1.06 |
| Manaus | 0.02 | 0.17 |
| Nouabalé-Ndoki | 0.015 | 10.5 |
| Pasoh forest reserve | 0.5 | 1.4 |
| Ranomafana | 0.1 | 6.84 |
| Udzungwa | 1.5 | 497 |
| Volcan Barva | 0.02 | 0.25 |
| Yanachaga | 1 | 269 |
| Yasuni | 0.3 | 0.49 |