Fig 1.
Timing diagram of soil moisture in Yanqing area.
Fig 2.
Soil moisture autocorrelation plot.
Fig 3.
The Taylor plot of weather data.
Table 1.
Correlation between various features and soil moisture.
Fig 4.
The general deep regression network structure.
Table 2.
Comparison of training results of different model structures.
Fig 5.
The DNNR model training results(ten times).
Table 3.
Single sample t-student test.
Fig 6.
The comparison of daily soil moisture prediction in Yanqing (DNNR).
Fig 7.
The comparison of daily soil moisture prediction in Shunyi and Daxing (DNNR).
(A)Comparison of daily soil moisture prediction in Shunyi.(B)Comparison of daily soil moisture prediction in Daxing.
Table 4.
Multi-region prediction error analysis of DNNR model.
Fig 8.
The comparison for average soil moisture predictions and real values in three regions(DNNR).
Table 5.
The comparison of multi-model prediction evaluation measures.
Fig 9.
The comparison of daily soil moisture predictions in Yanqing, Shunyi and Daxing(MLP).
(A) Comparison of daily soil moisture prediction in Yanqing.(B) Comparison of daily soil moisture prediction in Shunyi. (C) Comparison of daily soil moisture prediction in Daxing.(D) The comparison for average soil moisture predictions and real values in three regions.
Fig 10.
The fitting between the predictions and real values in Yanqing, Shunyi and Daxing.
(A) DNNR model in Yanqing area.(B) MLP model in Yanqing area. (C) DNNR model in Shunyi area.(D) MLP model in Shunyi area. (E) DNNR model in Daxing area.(F) MLP model in Daxing area.
Fig 11.
The distribution of prediction residuals in Yanqing, Shunyi and Daxing.
(A) DNNR model in Yanqing area.(B) MLP model in Yanqing area. (C) DNNR model in Shunyi areae.(D) MLP model in Shunyi area. (E) DNNR model in Daxing area.(F) MLP model in Daxing area.
Table 6.
Multi-region prediction error analysis results of MLP model.