FLOOD FORECASTING METHOD BASIS AS FLOOD MONITORING PROGRESS OF MEKONG RIVER

In this paper, used MODIS satellite image (MOD13) for monitoring the progress of flood in Mekong River Basin and testing the flood forecasting method for the Mekong Delta in flooding stage of 2015. The results showed that (80% reliability): MODIS image can be used to monitor the progress of flood in large areas of the Mekong River Basin. There was a close relationship between enhanced vegetation index EVI, land surface water index LSWI with growth status of plants and surface water of the flood. Risk flood maps during the flood season of the study area were established as the basis for developing the flood forecasting method applied to the Mekong Delta. With accuracy about 91%, this flood forecast method opened a new direction for researching about environmental disasters using the resource of satellite image at low cost. Therefore, it should use these images for monitoring the process, forecasting flood capability and other related fields in combination with other types of vegetation indices.


INTRODUCTION
In recent years, due to the effects of climate change, Mekong Delta (MRD) has experienced several major floods, most of which are caused by the floods from Mekong upstream (LV,et al., 2014;K., et al., 2010).Therefore, the meteorological and hydrologic prediction agencies shall take practical measures to making forecasts of flood possibility, so that people can prevent or minimize the damages to people and properties.The remote sensing image technology is a tool used to track, monitor and support the flood prediction.Currently, there are many researches for its application of the flood prediction model in the Mekong Delta (ZHONG, et al., 2011;ZHEN, et al., 2011;XIE, et al., 2016).However, monitoring methods and this model require the data and complex computations and are constrained by many factors such as flow cycle, flow direction and upstream floods.Through using multispectral-multi-time MODIS remote sensing images provided by NASA in combination the DEM (Digital Elevation Model), the progress of the Mekong River Basin Flood can be monitored as the basis for flood prediction at Mekong Delta.The research was conducted with the aim of: (1) Evaluating the possibility of using the multi-time MODIS images to monitor the progress of Mekong River Basin floods; (2) Proposing the solutions to forecasting the Mekong Delta flood progress in the flooding period in 2015.

Multi-time MODIS/ TERRA image data
Use the MODIS/TERRA SURFACE REFLECTANCE 8-DAY L3 GLOBAL 500 M SIN GRID V005 (MOD13) satellite images (MOD13) with a temporal resolution of 8 days and spatial resolution of 500 m (PENG, et al., 2004;ZHAO, et al., 2012).The images have been collected from the data system of the NASA Earth observation (EOS, 2013) in the period of 2013 to the first half of 2015; Use the Band 1, 2, 3 and 6 (red, near infrared and infrared short wave etc) to calculate the EVI, LSWI and DVEL indexes as the basis for the analysis and evaluation of the results.

SPOT Imagery Data
SPOT 5 images with a resolution of 80 x 80m and 3 Red -Green -Blue bands are collected from Singapore Remote Sensing Center (CRIPSs) to be used for comparing and inspecting the results interpreted from MODIS.

DEM digital elevation map of the research area
DEM Digital Elevation Model of Mekong River Basin collected from the ERSDAC Center of Earth Data Analysis in Japan (http://www.gdem.aster.ersdac.or.jp/) to be used to evaluate the flow direction of the river systems on the topographic forms of research areas.The model is calculated on the basis of elevation, slope and flow direction indexes to define the direction of river flows.

Flood mapping methods
The method of Sakamoto et al. is proposed to assess the flood.If the EVI value is greater than 0.2, it will be considered as opaque pixels covered by clouds and will be removed from the image (Thenkabail, et al., 2005;Xiao, et al, 2006).According to Xiao, et al. (2005Xiao, et al. ( , 2006)), if the EVI value is greater than 0.3, the object is classified as a pixel without flood.If the EVI value is ≤ 0.05 and less than or equal to 0, the pixel will be defined as relevant water pixels.Sakamoto et al, 2007) Then, it is necessary to classify into flooding pixel, combined pixel or long-term submerged objects.If the water-related pixel has EVI <0.1, it is considered as a flooding pixel.If the EVI value is greater than 0.1 but less than 0.3, the water-related pixel is identified as a combined pixel.The constantly flooded area is constantly being separated from the flood and mixed pixel.The pixel involves water flooding time> 120 days will be classified as long-term submerged object.With relatively high coefficient R2, it demonstrates that these two values are proportional to each other, when the water elevation gets higher, the flooded area will increase and vice versa.This is particularly significant in assessing the flooding depth by the flooded area of the study area at a certain time.

Start date, end date and prolonged time of floods
The spatial distribution of start dates is changed in each year.The start date of floods in 2013 was earlier than floods in 2014 (Figure 6).Results from comparing the two types of images of An Giang province showed a high level of compatibility (80%).Dark blue area determined to be flooded had a great level of compatibility between the two images.The % rate of deviation (only in SPOT image or MODIS image) was negligible.Apart from the same area to be present in two different types of images, the different area only at MODIS images was relatively higher than the flood area interpreted only in SPOT image.This partly explained for the differences in resolution and how the different types of images caused the different results.

Results of flood prediction for Mekong Delta
The implementation process was divided into 11 stages from 4th July 2015 to 30th September 2015, corresponding to 11 results from the actual verification.Correlated results between the predicted flooded area and actual flooded area interpreted from MODIS images shows a relatively high correlation coefficient (R2 = 0.91).The figure 8 shows that the degree of deviation in the forecasts is negligible and completely acceptable.This proves that the application of MODIS images in flood prediction in combination with other data is relatively accurate at flood events.

CONCLUSIONS
Based on the results above, the possibility of MODIS remote sensing images (MOD13) in monitoring the progress of floods in Mekong River Basin is pretty good.There is a relationship between characteristics of pixels and the spatial and temporal distribution of floods in Mekong River Basin.The high precision of analysis results indicates a possibility of widespread application of MOD13 images in monitoring, observing the flood progress as well as forecasting the floods in a long and contiguous time for a region as wide as a nation or territories.With the precision of approximately 91%, the forecasting methods can be applied to predict the likelihood of floods in Mekong Delta region and the temporal limit for prediction is one week.
While the MODIS images are not costly as today, together with the gradual improvement of forecasting methods, this can be seen as an effective solution for monitoring and forecasting the flood and disasters.
Fig.1 Methods for monitoring the progress and predicting floods for Mekong Delta by MODIS (Source:Sakamoto et al, 2007) Then, it is necessary to classify into flooding pixel, combined pixel or long-term submerged objects.If the water-related pixel has EVI <0.1, it is considered as a flooding pixel.If the EVI value is greater than 0.1 but less than 0.3, the water-related pixel is identified as a combined pixel.The constantly flooded area is constantly being separated from the flood and mixed pixel.The pixel involves water flooding time> 120 days will be classified as long-term submerged object.

HydrologicFig. 2
Fig.2 Methods of predicting the possibility of flooding in the Mekong Delta 1 Average increase (decrease) coefficient k = DT flooding risk of week 1/ Increased (Decreased)DT of week 2 compared to week 1 = 0.02 is selected as a value with the highest frequency of appearance in the Fig.6 Start date, end date, prolonged time of floods

Fig. 8 Forecast
Fig.8 Correlation and deviation between the predicted flooded area and actual flooded area