A spatiotemporal comprehensive constraint interpolation method for missing values in monthly composite Suomi-NPP VIIRS Cloud Mask images

Figures & data

Figure 1. Study area of Ningbo city.

Figure 2. Technical flowchart.

Figure 3. The total value of light brightness in Ningbo city in each month.

Figure 4. The calculation process figure of dramatic change pixels in June

Figure 5. Calculation process for pixels with dramatic changes in monthly differences.

Figure 6. Local Moran’s I of the $3\times 3$ neighborhood of VIIRS images in July and August 2014.

Table 1. The Global Moran’s I for each month of 2014, 2016, 2019 and 2021.

Month	Global Moran’s I in 2014	Global Moran’s I in 2016	Global Moran’s I in 2019	Global Moran’s I in 2021
January	0.892916	0.882883	0.832225	0.848658
February	0.774834	0.879178	0.849597	0.870334
March	0.873147	0.870784	0.916499	0.858178
April	0.911771	0.852261	0.907604	0.840055
May	0.852819	0.890959	0.876389	0.844779
June	0.906374	0.888372	0.89799	0.852179
July	0.877829	0.849194	0.872625	0.795484
August	0.885892	0.885693	0.889694	0.700308
September	0.90604	0.880858	0.912757	0.875363
October	0.879569	0.821109	0.833101	0.875663
November	0.870138	0.867238	0.895201	0.851655
December	0.877551	0.849684	0.899276	0.87659

Figure 7. Calculation process for locally autocorrelated abnormal pixels in a $3\times 3$ neighborhood.

Figure 8. Image simulated by the temporal interpolation methods (TIM). (a)Original image of June 2016 (b)Bezier3 (c)Dr (d)Exponent (e)Hermite3 (f)Spline3 (g)Lsm (h)Lsm2 (i)Lsm3.

Figure 9. Image simulated by spatiotemporal comprehensive constraint interpolation method (SCIM). (a)Original image of June 2016 (b)Bezier3 (c)Dr (d)Exponent (e)Hermite3 (f)Spline3 (g)Lsm (h)Lsm2 (i)Lsm3.

Figure 10. Comparison of the total brightness values of the original image and simulated images using different methods.

Table 2. Absolute value of the difference between the original image and images simulated with different temporal interpolation methods.

Methods	Absolute value of the difference				Average difference
	0–1	1–5	5–10	> = 10
Bezier3	123454	31840	6380	3406	0.470
Dr	125643	30380	5746	3311	0.554
Exponent	125641	30380	5749	3310	0.554
Hermite3	124345	30934	6145	3656	0.561
Spline3	119972	32615	7911	4582	0.481
Lsm	127421	30429	5118	2112	0.344
Lsm2	128939	29692	4320	2129	0.325
Lsm3	128939	29692	4320	2129	0.325

Table 3. Absolute value of the difference between the original image and the images simulated with the spatiotemporal comprehensive constraint interpolation method.

Methods	Absolute value of the difference				Average difference
	0–1	1–5	5–10	> = 10
Bezier3	123417	32034	6390	3239	0.410
Dr	125608	30455	5729	3288	0.540
Exponent	125632	30436	5729	3283	0.539
Hermite3	124636	30950	6054	3440	0.508
Spline3	122065	32096	7164	3755	0.357
Lsm	127964	30086	4946	2084	0.331
Lsm2	129270	29467	4236	2107	0.311
Lsm3	129270	29467	4236	2107	0.311

Figure 11. Comparison of the number of spatially heterogeneous pixels between the original image and the images simulated using different methods.

Figure 12. Scatter plots of original image and images pixel light brightness simulated by the temporal interpolation methods (TIM) in 2014. (a)Bezier3 (b)Dr (c)Exponent (d)Hermite3 (e)Spline3 (f)Lsm (g)Lsm2 (h)Lsm3.

Figure 13. Scatter plots of original image and images pixel light brightness simulated by the spatiotemporal comprehensive constraint interpolation method (SCIM) in 2014. (a)Bezier3 (b)Dr (c)Exponent (d)Hermite3 (e)Spline3 (f)Lsm (g)Lsm2 (h)Lsm3.

Table 4. The fitting degree (R²) between the simulated image and the original image.

R^2 between the original image and images simulated by different method	Bezier3	Dr	Exponent	Hermite3	Spline3	Lsm	Lsm2	Lsm3
TIM (2014)	0.893	0.928	0.929	0.928	0.876	0.934	0.936	0.936
SCIM (2014)	0.902	0.935	0.935	0.934	0.898	0.948	0.950	0.950
TIM (2016)	0.885	0.896	0.896	0.887	0.845	0.901	0.910	0.910
SCIM (2016)	0.897	0.906	0.906	0.908	0.881	0.907	0.916	0.916
TIM (2019)	0.907	0.915	0.915	0.907	0.889	0.923	0.931	0.931
SCIM (2019)	0.911	0.918	0.918	0.912	0.893	0.934	0.934	0.934
TIM (2021)	0.814	0.818	0.818	0.813	0.763	0.822	0.831	0.831
SCIM (2021)	0.831	0.878	0.878	0.871	0.797	0.875	0.883	0.883

Data availability statement

The data that support the findings of this study are openly available at https://eogdata.mines.edu/products/vnl/.