The authors regret that there were some errors in their original paper and apologise for any inconvenience caused. The corrections can be found below:

1. In the Graphical Abstract, ‘Other objectives', the following sentence should be deleted: ‘Prediction capabilities of extreme water levels’.

2. In the Introduction Section, paragraph 5: ‘Gated Recurrent Unit (MLP)’ is incorrect, the correct presentation is ‘Gated Recurrent Unit (GRU)’; ‘TL-MLP ’ should be presented as ‘TL-MLP (Multi Layer Perceptron)’.

3. The caption for Figure 2 should include a citation, as below:

Figure 2 | Characteristics of the rainfall and water level data used for the study. (a–c) subplots show the rainfall variation of Aralaganwila, Angamedilla, and Polonnaruwa_Agri, respectively. (d) exhibits the water level of Manampitiya. (e, f) boxplots illustrate the annual and monthly water level variation of Manampitiya, respectively (Madhushanka et al. 2024).

4. In Section 2.3. Utilized models, line 4 the sentence should read ‘Unlike RNNs, LSTM incorporates an additional cell state or cell memory (ct) where information can be stored, along with gates (represented by dashed rectangles in Fig. 3) that regulate the flow of information within the LSTM cell.’

5. The following sentence above Equation 3 should be deleted: ‘Subsequently, the current input (xt) and the last hidden state (ht-1) are combined to calculate a potential update vector for the cell state, using the following equations'. It should be replaced by ‘Additionally, the second gate, denoted by a green rectangle, known as the input gate or compute gate, determines how much of the new candidate values (from the current input and previous hidden state) should be written to the cell state (ct). In this gate, the current input (xt) and the last hidden state (ht-1) are combined to calculate a potential update vector for the cell state, using the following equations.’

6. The following sentence under Equation 5 should be deleted: ‘Additionally, the second gate, denoted by a green rectangle, known as the input gate or compute gate, determines the extent to which the information from ct is utilized for updating the cell state in the current time step.’

7. The following sentence under Equation 8 should read: ‘The original transformer consists of an encoder–decoder architecture as depicted in Figure 4.’

8. In Table 2 there is a correction to one of the elements as follows:

dmodel (Dimension of the input vectors)

9. The corrected version of Figure 5 can be found below:

10. The corrected version of Equation 12 can be found below:
(8)

11. In Section 2.6. Experimental setup, paragraph 2, lines 4 and 5, the sentences should read:

i. Case 1 – Past water levels as the only input

ii. Case 2 – Past water levels and rainfall data as the inputs

12. In Section 3. RESULTS AND DISCUSSION the second paragraph should read:

According to the results (blue and green bars) in Figure 7, Case 3 has the highest error and case 2 denotes the lowest error while the error of Case 1 is close to Case 2, for both LSTM and transformer algorithms. For the LSTM, RMSE performance was improved by 47% in Case 1 compared to Case 3 while 12% from Case 1 to Case 2. For the transformer, they were 39 and 9%, respectively, denoting a similar behavior. This high improvement from Case 3 to Case 1 and comparatively small improvement from Case 1 to Case 2 indicate a higher impact of the past water level data on the output, among all the input features.

13. In the References section the following entry should read as below:

Madhushanka, G. W. T. I., Jayasinghe, M. T. R. & Rajapakse, R. A. (2024). Multiple-Day-Ahead Flood Prediction in the South Asian Tropical Zone Using Deep Learning, Journal of hydrologic engineering,30 (1). https://ascelibrary.org/doi/10.1061/JHYEFF.HEENG-6296.

This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).