Feature Importance of RECLAIM Ensemble

Because individual models can yield substantially different importance rankings, we adopted an ensemble approach: a weighted average (based on mean model weights used by RECLAIM in the training dataset) for Sedimentation Rate. Below table lists all 88 variables along with their importance, cumulative importance, and rank for ensemble in RECLAIM using mean of dynamnic weights in training data.

Feature importance of all 88 variables from RECLAIM ensemble.

Rank	Variable	Feature Importance (%)	Cumulative Importance (%)
1	log_OBC	14.633882	14.633882
2	log_LCWB	8.004204	22.638086
3	log_SA_mean_clip	7.383239	30.021326
4	log_CA	6.589474	36.610800
5	log_LCM	4.316336	40.927136
6	log_LCG	3.379634	44.306770
7	log_LCSG	2.390016	46.696786
8	log_SA_std	1.985361	48.682147
9	wind_mean	1.840513	50.522661
10	log_rain_per_area	1.765137	52.287797
11	log_LCC	1.607280	53.895078
12	AGE	1.472827	55.367905
13	R_SA_cap	1.340215	56.708120
14	log_O_std	1.290960	57.999079
15	SILT	1.229707	59.228786
16	log_LCS	1.186893	60.415679
17	LON	1.109406	61.525085
18	wind_cv	1.087307	62.612392
19	E_std	1.055644	63.668036
20	log_SOUT	1.035725	64.703761
21	NSSC1_std	1.025786	65.729547
22	log_MAR	0.930290	66.659837
23	log_DCA	0.915694	67.575531
24	NSSC2_max_persis	0.882480	68.458011
25	wind_kurt	0.833630	69.291641
26	wind_skew	0.812826	70.104466
27	log_LCAS	0.805496	70.909963
28	O_cv	0.786492	71.696454
29	AECC	0.781898	72.478352
30	SAND	0.754986	73.233339
31	log_MAO	0.752642	73.985981
32	NSSC2_mean	0.717024	74.703005
33	SLOP	0.682984	75.385989
34	log_LCT	0.667932	76.053921
35	I_cv	0.647218	76.701138
36	NVGF	0.645478	77.346616
37	NSSC1_cv	0.644180	77.990796
38	NSSC1_skew	0.642675	78.633471
39	R_shrub_bare	0.632633	79.266104
40	rel_SA_mean_clip	0.627533	79.893637
41	I_above_90	0.622972	80.516608
42	CURV	0.612935	81.129543
43	log_LCBS	0.600176	81.729719
44	log_ESR	0.568189	82.297908
45	wind_std	0.562443	82.860351
46	log_SA_mean	0.545410	83.405761
47	BULK	0.537301	83.943063
48	R_treaa_bare	0.523224	84.466287
49	R_coarse_sand	0.517022	84.983308
50	log_LCHV	0.509614	85.492923
51	ELEV	0.506835	85.999757
52	log_ROBC	0.504507	86.504264
53	I_max_persis	0.499298	87.003562
54	SA_skew	0.490533	87.494095
55	NSSC1_mean	0.489971	87.984066
56	ASP	0.486329	88.470395
57	NSSC1_kurt	0.470670	88.941065
58	log_HGT	0.461355	89.402420
59	log_GC	0.454284	89.856704
60	log_PAI	0.448311	90.305015
61	SA_cv	0.447305	90.752320
62	DLC	0.445204	91.197524
63	log_TE	0.434373	91.631898
64	E_mean	0.423686	92.055584
65	log_ECLR	0.421004	92.476587
66	#_rain_above_50	0.419211	92.895798
67	log_MAI	0.396943	93.292742
68	SA_above_90	0.386455	93.679196
69	log_SIN	0.370423	94.049620
70	log_RP	0.369811	94.419431
71	NSSC2_above_90	0.367674	94.787105
72	AECI	0.363760	95.150865
73	log_SA_kurt	0.363547	95.514412
74	#_rain_above_10	0.350768	95.865180
75	tmin_mean	0.342336	96.207516
76	COAR	0.339481	96.546997
77	log_FL	0.338535	96.885531
78	log_RT	0.333771	97.219302
79	log_I_std	0.333185	97.552487
80	CLAY	0.330524	97.883011
81	log_RA	0.302114	98.185125
82	tmax_mean	0.285973	98.471097
83	AECS	0.284486	98.755583
84	#_rain_above_100	0.281851	99.037435
85	MRB	0.253738	99.291173
86	HILL	0.246680	99.537852
87	LAT	0.238995	99.776847
88	log_LCSV	0.223153	100.000000