Mike's PhD project home page

Model results

Introduction

A mesoscale atmospheric circulation was used to simulate the effect of land use change over the island of Puerto Rico. The weather of 5 September 1997 is simulated, because this was a largely cloudless day. In one run, it was assumed that the entire island is covered by forest, and the other run by pasture. Here follow the results.

Forest run

Panel (a) shows that the surface is about 34 oC over land, but it is about 4 degrees colder over sea. This temperature difference causes a sea breeze circulation to develop, as shows panel (c) (The wind direction is onshore on both the north and south coast in the lower 500 m). The sea breeze on the north coast is opposed to the large scale south-easterly winds. Therefore a sea breeze front forms, with related updrafts, as seen in panel (c). The updraft transports moisture to higher levels. On top of the updraft a thin cloud is formed (panel (d)).

Watch a movie of the cloud cover in the forest run (1,015kb).

Figure 1. Cross section of (a) potential temperature, (b) vertical wind speed (contours every 0.25 m/s, starting at +/- 0.125 m/s.), (c) lateral wind speed (to north is positive) and (d) vapor mixing ratio and cloud mixing ratio (contours at 0.01, 0.1, 0.2,... g/kg). Model run with forest vegetation over the entire island, at 15.00h local time on September 5, 1997.

Pasture run

In a second simulation, the forest vegetation is replaced by pasture. Figure 2 shows what happens to the circulation when Puerto Rico is deforested.

The land surface is only 30 oC, 4 degrees cooler than over forest, and it is about 2 g/kg moister. Because the temperature difference between the land surface is not as large, the sea breeze circulation is weaker in the pasture run. This can be seen in the onshore wind speeds on both coastlines. The related updrafts are consequently also weaker. The weaker circulation brings less moisture to higher levels.

Watch a movie of the cloud cover in the pasture run (804kb).

Figure 2. Same model results, except the forest vegetation was replaced by pasture.

Difference

Figure 3. Cross sections of differences between forest and pasture model runs of (a) potential temperature, (b) condensate mixing ratio (contours every 0.05 g/kg, starting at 0.025 g/kg, (c) lateral wind speed and (d) vapor mixing ratio.

Field campaign

What are the properties of forest and pasture, that causes the differences between both runs?

By selecting a vegetation type in the model one sets specific values for the albedo, the surface resistance to transpiration and the surface roughness.

This directly influences the net radiation at the surface and its partitioning into sensible and latent heat flux, the surface temperature and humidity, the momentum flux and wind speed.

When the project was started, these variables were all unknown. Therefore an extensive measurement campaign was set up, that lasted from May 1997 until May 1998.

Two main field sites were selected, one in virgin forested wetland, which were common in the coastal plains in Puerto Rico, and the other on a pasture. Both sites were located on the property of the U.S. Naval Security Group Activity, Sabana Seca (NSGASS), some 20 km west of San Juan, within a km from the northern coastline. The two sites were separated by no more than 1 km.

The table on the right lists the measurements that were performed. The data were sampled every 30 seconds by Campbell 21X dataloggers and 30 minute averages were stored.

Two other measurement sites were located in the Caribbean National Forest, also referred to as El Yunque or the Luquillo Experimental Forest. One station was in a palm forest at 900 m elevation. The fourth station was (more or less) portable and has been moved around quite a bit.

Observations

Intro

• annual cycle of albedo
• net radiation (pasture vs. forest)
• energy balance (pasture and forest)
• annual cycle of net radiation and evapo-transpiration rate (pasture)
• annual cycle of net radiation and evapo-transpiration rate (forest)

The albedo of the pasture is typically 0.20 and of the forest 0.13. The forest absorbs about 7 percent more shortwave radiation.

The months May thru August 1997 were at the end of a rather dry period. In September the trees in the forest defoliated, flowered and grew fresh leaves. The effect on albedo is clearly visible.

The difference in net radiation comes down to the difference in absorbed shortwave radiation, because the net longwave radiation is small in the tropics. The forest receives about 8 percent net radiation.

At the pasture, about evapo-transpiration (LE) consumes about 80 percent of the net radiation (Rn), leaving relatively little energy for sensible heat (H). The soil heat flux (G) is small, because the dense vegetation acts as an isolator.

The evapo-transpiration rate is not as large at the forest. As a result of this and of the larger net radiation, the sensible heat flux is larger. The soil heat flux is quite large. The soil is drenched with water and the foliage is not too dense.

These results are quite uncommon. Usually a grassland has a lower transpiration rate than a forest. In this specific case, both sites have sufficient soil moisture not to limit transpiration. In fact, at the forest site, the transpiration rate may even be limited because of stress due to high salinity or low soil oxygen.
(link to simulated energy balance)

The careful observer may notice that the net radiation has to maximums, one in April and one in August. The sun is in the zenith in May and passes to the North. It moves back to the South in August.

The evapo-transpiration rate is 3 to 5 mm/day at the pasture.

At the forest, the evapo-transpiration rate is smaller, but more constant throughout the year. This adds to the assumption that the transpiration rate is limited by physical stress to the trees.

Vegetation parameterization

Intro

The original vegetation parameterization in the model did not appear to function very well for the forest and pasture vegetation. Especially the simulated surface energy balance did not match the observations.

With the available observations, a new parameterization could be developed for the exchange rate of momentum, heat and moisture between the soil, vegetation and the atmosphere. The flux rates are determined using a resistance model. The heat and momentum fluxes from the vegetation are modulated by the aerodynamical resistance. An additional surface resistance regulates the transpiration rate.

In the original model the fluxes also depend on the Leaf Area Index (LAI), but this dependency was removed for the moment, because no independent data are available of flux rates and LAI.

Aerodynamical resistance

The original formulation of the aerodynamical resistance is:

This parameterization was changed to:

because this is a more direct formulation.

Surface resistance

Comparison of the parameterized surface resistance to observations.

Heat capacity

The heat and water storage capacity are very large by default. They are changed to the values listed in the table on the right.

(*) These values are larger than they should be, to increase computational stability. The correct values are 1.2e3 J/m2/K and 1.2 kg/m2.

Longwave radiative exchange between soil and vegetation

The model allows free radiative exchange between the soil and the vegetation. This resulted in a large soil heat flux at the pasture. In reality, the radiative exchange in a tall grass pasture is a stepwise process, which is less effective. To simulate this stepwise process, the following parameterization is introduced:

BATS table

The original model uses the BATS classification of vegetation types. Since most vegetation types do not occur in Puerto Rico it was decided to keep only two vegetation types, pasture and forest, besides ocean. Work is planned to create a new palm forest vegetation type.

Selecting forest or pasture sets the variables listed in the table on the right.

Validation

Surface energy balance

The energy balance in the simulations matches the observations quite well. In the forest run, the sensible heat could be a little larger relative to the latent heat flux. Also, the latent heat flux does not decrease at sundown, but this is, too a lesser extend, also observed. Otherwise, there are no concerns.
(link to observed energy balance).

Surface temperature, water vapor mixing ratio, wind speed and momentum flux

The observations are in solid lines, the model results dashed.

In the forest run, the surface temperature is a couple of degrees high. This is a matter of concern, since it strongly influences the sea breeze circulation. Some more work on the surface energy balance may give better results.

The wind speed is somewhat low in the pasture run and high in the forest run.

The momentum flux is high in the forest run.

Longwave radiation

The downwelling longwave radiation is in blue and the emitted radiation from the surface in red. The observations are in solid lines, the model results dashed.

The Chen and Cotton radiation scheme is used here. There is some concern about the emitted longwave radiation, which is some 30 W/m2 low in the pasture run, while the surface temperature is correct, as seen above. In the forest run, the emitted longwave radiation matches the observations, even though the surface temperature is a couple of degrees high.

The Harrington radiation scheme has also been tested, but produced worse results. However, a bug-fixed version is still to be tested.

Why is this interesting?

'cos it's cool, mon!

Puerto Rico's fresh water supply is insufficient. Every so many years, the citizens suffer from a severe water rationing. The impression is there, that this happened more frequently in recent decades.

While many things, like the management of the water reservoirs, the increase of the population, are related to this problem, the deforestation may also be a factor.

Puerto Rico was covered by forests when Columbus discovered the island in the 15th century. With agricultural development and urbanization, large parts of the island's coastal plains were deforested.

From a micrometeorological point of view, a forested and deforested land differ most importantly in albedo and net radiation, evapo-transpiration rate and consequently the Bowen ratio (the ratio of the sensible and latent heat fluxes), as well as the surface roughness. As a result the surface temperature, humidity and wind speed depend on the vegetation type.

The surface temperature, humidity and roughness determine the strength of the mixing of the air in the vertical. Since the vertical mixing is closely related to cloud formation, changes in the land use can directly effect cloud condensation.

Puerto Rico is the smallest of the larger Antilles. It is 180x60 km. The Cordillera Central is a mountain range that crosses the island from east to west. The Caribbean National Forest is located in the Northeast.

Model setup

RAMS

The mesoscale atmospheric circulation model used is RAMS (Regional Atmospheric Modeling System) version 4.29. It is developed at the Colorado State University.

Grids setup

Three grids are used. The coarse grid domain is 800x600km. It has Puerto Rico in the center and about 300 km of ocean on all sides. Its grid spacing is 20 km.

The next two grids are nested in the coarsest grid. Grid 2 spans a 128x168km North-South transsection of Puerto Rico with a 4 km grid spacing. Grid 3 has a 1 km grid spacing and a 26x86km domain.

All grids have 30 levels in the vertical. The grid spacing is 50 meters at the surface and it is stretched to a maximum of 1250 m. The model top is at 20 km.

The intention is to make the finest grid as fine and its domain as large as possible, in order to resolve cumulus convection, with the boundaries far away. The prescribed settings are the about the limit on what is practical for test runs.

Locations of grid 1, 2 and 3 relative to Puerto Rico.
Elevation contour lines are drawn every 50 m.

Initial and boundary conditions

Variable initialization is applied for all grids. The NCEP reanalysis data are used. These consist of 6 hourly 2.5x2.5 degree resolution fields of potential temperature, relative humidity, u- and v-component wind speed and geopotential height During the model run, the boundaries of the coarsest grid are nudged towards the NCEP reanalysis data.

The NCEP reanalysis data are freely available at the anonymous ftp site ftp://archive.cdc.noaa.gov/Datasets/archive3/ncep.reanalysis/pressure.

Surface files

USGS 1 km resolution elevation maps are used for topography initialization. Sea surface temperature data files are available at a 2.5x2.5 degree resolution. Vegetation type data are available at 1 km resolution, but they are not used in the simulations.

Options

The following options are set in the RAMSIN file:

Thesis (pdf/hard copy)

pdf's

Hard Copy

Hard copy's are still available through my email address below.

Feel free to email me at