Читать книгу The Digital Agricultural Revolution - Группа авторов - Страница 59

The parameters selected as input variables for neural network were derived for each season separately for all years. The APAR derived from Landsat 8 image on 14 October 2015 is shown in Figure 2.4. The spatial distribution of APAR values varied from 0.4369 to 0.7741. The higher APAR was observed at cropped area and was the lowest at water bodies. Similarly, the crop water stress index was derived and was presented in Figure 2.4 (a). The CWSI on the day of derived is shown in Figure 2.4 (b), which is a proxy for water stress. The CWSI varied from 0-1. The CWSI matched with irrigation canals. The CWSI is high in the areas where there is a low supply of irrigation water, especially in the lower parts of the area. Mandal wise-extracted mean crop yield parameters are presented in Table 2.2.

The selected yield parameter maps were derived from satellite images and were given as input to FFBP NN model. Time series NDVI, T_s, CSWI, and APAR maps were retrieved from remotely sensed images from transplanting to the harvesting stage of paddy crop. The values for the ground truth points of all parameters derived by remote sensing were extracted. Spatial distribution of sample points is shown in Figure 2.5. The attributes of abovementioned derived thematic maps are extracted for all the sample points and exported to Excel as .csv file (Figure 2.6) for preparation of input files to neural network structure.

Figure 2.4 (a, b). APAR and CSWI maps of KCD on October 14, 2015 (original figure).

Figure 2.5 Synoptic view of spatial distribution of sample points of crop collected (original figure).

Table 2.2 Sample normalized input data of FFBPNN yield estimation model of paddy crop in kharif during 2015.

S. no.	Mandal name	NDVI	Ts	APAR	Water index	Normalized average yield
1	Vijayawada rural	0.542	0.831	0.775	0.354	0.674
2	Kankipadu	0.548	0.148	0.778	0.448	0.855
3	Challapalle	0.539	0.792	0.773	0.180	0.544
4	Pamarru	0.528	0.577	0.768	0.224	0.533
5	Vuyyuru	0.531	0.330	0.770	0.325	0.744
6	Movva	0.687	0.363	0.844	0.723	0.725
7	Thotlavalluru	0.687	0.439	0.844	0.686	0.450
8	Avanigada	0.563	0.454	0.785	0.359	0.000
9	Pamidimukkala	0.575	0.687	0.791	0.302	0.900
10	Guduru	0.459	0.767	0.735	0.000	0.915
11	Penamaluru	0.478	0.706	0.744	0.063	0.525
12	Koduru	0.748	0.482	0.874	0.821	0.900
13	Pamarru	0.771	0.390	0.885	0.929	0.619
14	Machilipatnam	0.533	0.452	0.771	0.284	0.921
15	Pedana	0.491	0.203	0.750	0.266	0.957
16	Mopidevi	0.496	0.790	0.752	0.080	0.544
17	Nagayalanka	0.511	0.521	0.760	0.203	0.750

Table 2.2 represents the maximum and minimum values of the normalized data. The average values of NDVI ranged from 0.459 to 0.687 for different mandals. Similarly, APAR and CWSI ranged from 0.750 to 0.874 and 0 to 0.929, respectively. Surface temperature and average yield were normalized. The normalized surface temperature was in the range of 0.148 for Kankipadu to 0.831 at Vijayawada rural. The highest normalized yield is at 0.957 at Pedana. The point wise normalized parameter values were used as input to the NN model.

Figure 2.6 Exporting the attribute values of the points to excel (original figure).