Table 1 Confusion matrices for comparing observed and modeled injection behavior counts.

HRRR-Smoke against HRRR-Smoke PBL		Non-injection forecasted	Injection forecasted	Total
	Non-injection observed	TN = 2	FP = 24	26
	Injection observed	FN = 0	TP = 17	17
	Total	2	41	43
HRRR-Smoke against DIAL-HSRL PBL		Non-injection forecasted	Injection forecasted	Total
	Non-injection observed	TN = 9	FP = 17	26
	Injection observed	FN = 3	TP = 14	17
	Total	12	31	43
WRF-Chem against WRF-Chem PBL		Non-injection forecasted	Injection forecasted	Total
	Non-injection observed	TN = 9	FP = 24	33
	Injection observed	FN = 1	TP = 17	18
	Total	10	41	51
WRF-Chem against DIAL-HSRL PBL		Non-injection forecasted	Injection forecasted	Total
	Non-injection observed	TN = 20	FP = 13	33
	Injection observed	FN = 2	TP = 16	18
	Total	21	30	51

Counts of injection behavior are shown for HRRR-Smoke using the HRRR-Smoke PBL height to evaluate injection, HRRR-Smoke using the DIAL-HSRL PBL height to evaluate injection, WRF-Chem using the WRF-Chem PBL height to evaluate injection, and WRF-Chem when the DIAL-HSRL PBL height is used to evaluate injection. Columns are labeled with forecasted plume behavior and rows are labeled with observed plume behavior. TN stands for true negative, where non-injection was forecasted and observed. Similarly, TP stands for true positive. FN stands for false negative, where injection was not forecast but did occur. FP stands for false positive, where injection was forecast but did not occur. HRRR-Smoke has fewer total cases than WRF-Chem due to certain fires missing in the model (see “Methods”, Sampling Method).

Quick links

Search