Contribution of citizen science towards cryptic species census: “many eyes” define wintering range of the Scaly-sided Merganser in mainland China

Table 3 Measures of the predictive performance of the random forest species distribution models

Year	AUC^a	OAA^b	Kappa^c	Performance
2014	0.91 (0.84)	0.94 (0.92)	0.63 (0.52)	Substantial
2015	0.78 (0.71)	0.81 (0.76)	0.33 (0.28)	Fair
2016	0.91 (0.87)	0.84 (0.77)	0.56 (0.47)	Moderate

Values in parentheses are based on testing data, which are independent of the training data
^aArea under the receiver operating characteristic (AUC). Models with AUC values > 0.75 are suitable for conservation planning (Pearce and Ferrier 2000)
^bOAA: Overall accuracy
^cKappa coefficient: 0–0.20, slight; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, substantial; and 0.81–1, almost perfect, according to Landis and Koch (1977)

ISSN: 2053-7166