Our research interests focus on the nexus of water and numerous impact areas such as farming, stormwater <-> infrastructure, socio economics, as well as aspects of the WASH (WAter, Sanitation, Hygiene) realm, such as potable water supply systems. We share these interests with a number of local partners and collaborators such as DINEPA (the Haitian water organization), DATIP (an organization to aid (on technical aspects) citizens of the Palm region), and also the Haitian Reforestation Partnership which runs the COmprehensive DEvelopment Project, CODEP, seeking to reforest the hills and mountains above Leogane.
All aspects of water are geographically confined to the surface watersheds and sub-surface groundwater layers as these are the sole suppliers of water to the population that lives in the these watersheds, i.e. typically there is no cross-watershed water transfer in Haiti. Consequently, estimating how much water is available and where it occurs and how long it stays is fundamental to assessing water impacts on the local population.
What impact did de-forestation had (compared to a pristine environment) and how effective is re-forestation on improving the water budget of the Leogane plain?
What impact will projected climate change have on the water budget of the Leogane plain and to what extent will it adversely affect a) farming operations and b) frequency of stormwater/flooding events?
A meaningful assessment and path forward to address the above objectives requires the use of numerical tools to compute state variables (such as water volume) over large spatio-temporal scales, i.e. 150km^2 of watershed area and time frames stretching from monthly to seasonal to yearly to dozen of years. Both objectives have what we call "what-if" scenarios embedded which by definition can only be looked at through simulations. In our case we simulate the "pristine state", the "current state", and then "degrees of reforestation states" that ultimately could lead back to the pristine state (it will not of course, but it could reach a new "good" state). The second modeling trajectory is the "prediction chain" that would seek to assess what the states are in 10 years from now, 20 years, 30 years and so on, with the aim of assessing the year 2100.
In order to support this "what-if" modelling strategy the models will need forcing and initial data or boundary conditions that set those initial states and then time series data to drive the models to produce time series data of the state variables. While forecast/prediction data will necessarily have to be produced by other, especially atmospheric, large scale numerical models (such as GSSHA and WRF) we are also collecting a set of hydroclimatological data (since 2012) such as relative humidity, air temperature, wind, solar radiation, and precipitation that will help calibrate our models. We also collect groundwater level (shallow layer) readings at various locations throughout the plain.
We are using a number of tools to assess different aspects of our watersheds.
1) A combination of US Corps of Engineers HEC-1/HEC-HMS (Rainfall-Runoff) and HEC-RAS (River Routing). These models are used for short term (extreme) event assessment. These are Desktop applications.
2) We also use GSSHA, a finite difference model, that has
a high spatial resolution to better map out the landscape.
We run GSSHA in competitive mode to the HEC pair
described above to compare and validate.
3) SWAT (Soil Water Assessment Tool) to conduct long term simulations that incorporate detailed landcover and landuse (LC/LU) changes. This is a Desktop application.
We have developed numerical representations of the three watersheds that dominate the alluvial fan of the Leogane plain. We have different models using the above code stacks for each of the three watersheds so we can test/run short term (such as Hurricane Sandy) and long term (IPCC informed predictions on long term climate change) scenarios. For example, we sought to re-create a pristine (pre colonial times) environment and compare hydrologic state of "bach then" with states of more recent times, all the way to today. To this end we have use LandSat based LC/LU coverages that reach back to the late 70' and thus provide about a 40 year gap in which we can track the LC changes; most notably de-forestation. It also allows us to examine what if scenarios, for example to assess the effects of re-forestation on the hydrologic cycle, in our case that of CODEP. We can also use the numerical models to examine what might happen decades down the road; it has been well accepted that climate change will have a significant impact on the Caribbean islands and the many subsistence farming operations. For this work we also need to find ways to downscale global circulation models (CGM) climate predictions computed on large grid cells to the local weather patterns in the Leogane area. We had the good fortune to receive dynamically downscaled data from meso- and regional scale climate models for a small selection of years (2x2km grid) while for the other years we used statistical procedures.