The Buffalo Bayou River watershed, that covers Houston City, Texas.


This study is to estimate the inundation area during Harvey Hurricane 2017 in Houston, Texas, from Aug 27 to Sept 3, 2017 (Wikipedia). This hurricane is one of most costliest U.S. Atlantic hurricanes, that cost 125 billion USD. The similar costly hurricane is Katrina 2015.

Infrared satellite loop of Harvey making its third U.S. landfall in Louisiana on August 30, source: Wikipedia

news3 Flooding in Port Arthur, Texas on August 31


I select the Buffalo Bayou River (BBR) watershed that encloses Houston city as the modeling boundary.

Watershed The Buffalo Bayou River (BBR) watershed. Area = 2713 $km^2$. BBR watershed covers Houston City in Texas.


Meteorological data is from NLDAS-2 re-analysis data in the whole year of 2017.

precp The daily precipitation in the BBR watershed. Maximum precipitation is ~ 40 $cm/day$.


The rSHUD in R realizes the domain discretization, quasi-automatically. Figure and information about the modeling domain are shown below:

Table: Information of Model domain

Number of elements8571
Number of river reaches392
Number of river order3
Number of river segments2057
Number of bank elements1607
Maximum elevation61.02$m$
Minimum elevation0.11$m$
Watershed area2713.53$km^2$
Average element area0.32$km^2$
Total river length696.70$m$
Shortest river reach336.98$m$
Average river slope0.00097$m/m$


Hydrologic Unstructured Domains for Buffalo Bayou River (BBR) watershed. Color is elemvationf from 0 to 80 meters above sea level. Total number of elements is 8571; average area of the elements is 0.32 $km^2$— equivalence of $560m \times 560m$ resolution.


Below is the simulated inundation area, compared to MODIS data. vs Comparison of the SHUD modeling results and MODIS analysis from USGS.

There is agreement, more or less, between the SHUD model and MODIS analysis. The analysis of MODIS data suggested the northwestern rural area was flooded, but the model simulation barely captured the inundation in those rural areas. In contrast, the flooding in urban areas was less in MODIS result than SHUD simulation.

The first question is: what was the real condition in the urbanized areas? Were they flooded or not? So the next step is to collect more local flooded data to validate the inundation area against model results and MODIS estimation.


中国科学院西北生态环境资源研究院副研究员,SHUD模型作者; 专注于数值方法的分布式水文模型、水文大数据的机器学习、气候/人类活动对水循环的影响和集成模型耦合研究