# Ocean Surface Latent Heat Flux Diagnostic Documentation¶

Last update: 12/10/2021

Tropical intra-seasonal (20-100 day) convection regulates weather patterns globally through extratropical teleconnections. Surface latent heat fluxes help maintain tropical intra-seasonal convection and the Madden-Julian oscillation by replenishing column water vapor lost to precipitation. Latent heat fluxes estimated using surface meteorology from moorings or satellites and the COARE3.0 bulk flux algorithm suggest that latent heat fluxes contribute about 8% of the intra-seasonal precipitation anomaly over the Indian and western tropical Pacific Oceans [Dellaripa and Maloney, 2015, Bui et al., 2020].

For this diagnostic, we use in-situ data from TAO/TRITON/RAMA to create a location-based latent heat flux matrix determined by specific humidity deficiency at the surface layer (dq) and surface wind speed (sfcWind). By comparing the matrix between observation and models/reanalysis, the diagnostic can help revealing where model/reanalysis latent heat flux biases are in the dq-sfcWind space. The latent heat flux biases shown in the diagnostic demonstrate dependence on both sfcWind and dq. An offline latent heat bias correction can be performed on simulations based on the bias latent heat fluxes matrix as a function of dq and sfcWind.

## Version & Contact info¶

Version/revision information: version 2 (12/10/2021)

PI (Charlotte A. DeMott, Colorado State University, charlotte.demott@colostate.edu)

Developer/point of contact (Chia-Wei Hsu, Colorado State University, Chia-Wei.Hsu@colostate.edu)

### Open source copyright agreement¶

The MDTF framework is distributed under the LGPLv3 license (see LICENSE.txt).

## Functionality¶

The main script generates the Latent heat flux matrix and bias matrix.

### Python function used¶

- groupby_variables.bin_2dThe function is written to bin the variable (target_var) in a xr.Dataset
based on two other variables (bin1_var, bin2_var) in the same xr.Dataset. The function calculate the mean, std, and count values of the target_var after binning.

- model_read.regional_varThe function is written to read the model output and required varaibles.
The function also crop the data based on the user set time period and region. Two varibales is calculated in this function 1) saturation specific humidity near surface (determined by surface temperature and surface pressure) and 2) dq which represent the vertical difference of specific humidity near surface.

- obs_data_read.tao_tritonThe function is written to read the observational data and required varaibles
from the TAO/TRITON array. The function also crop the data based on the user set time period and region. Two varibales is calculated in this function 1) saturation specific humidity near surface (determined by surface temperature and surface pressure) and 2) dq which represent the vertical difference of specific humidity near surface.

- obs_data_read.ramaThe function is written to read the observational data and required varaibles
from the RAMA array. The function also crop the data based on the user set time period and region. Two varibales is calculated in this function 1) saturation specific humidity near surface (determined by surface temperature and surface pressure) and 2) dq which represent the vertical difference of specific humidity near surface.

## Required programming language and libraries¶

The programming language is python version 3 or up. The third-party libraries include “matplotlib”, “xarray”, “metpy”,”numpy”,”scipy”.

## Required model output variables¶

With daily frequency from the model output. This diagnostic needs

### input atmosphere model variables¶

‘huss’ : Surface 2m Humidity (kg kg-1)

‘ts’ : Skin Temperature (SST for open ocean; K)

‘sfcWind’ : Near-Surface Wind Speed (10 meter; m s-1)

‘psl’ : Sea Level Pressure (Pa)

‘hfls’ : Surface Upward Latent Heat Flux (W m-2 and positive upward)

‘pr’ : Precipitation (kg m-2 s-1)

The script is written based on the CESM2-CMIP6 daily data download hosted by WCRP.

The dimension of all variable is 3-D with (time,lat,lon) in dimension and 2-D array for lat and lon as coordinate.

## Required observational data¶

With daily frequency from the observational data. This diagnostic needs

### input observational variables¶

‘RH’ : Relative Humidity (%)

‘SST’ : Sea Surface Temperature (for open ocean; K)

‘WindSpeed10m’ : Near-Surface Wind Speed (10 meter; m s-1)

‘SLP’ : Sea Level Pressure (Pa)

‘Latent’ : Surface Upward Latent Heat Flux (W m-2 and positive upward)

‘airT’ : Near Surface Temperature (K)

#### data access :¶

All variables can be downloaded from PMEL NOAA hosted website https://www.pmel.noaa.gov/tao/drupal/flux/index.html

## References¶

1. C.-W. Hsu et al. (2021): Ocean Surface Flux Algorithm Effects on Tropical
Indo-Pacific Intraseasonal Precipitation. *GRL*, under review.

## More about this diagnostic¶

Surface latent heat flux from ocean to the atmosphere is one of the important processes that provides water vapor and energy to the daily tropical rainfall. A visually intuitive latent heat flux diagnostic is proposed to better understand the model shortfall on its latent heat flux representation. This diagnostic allows a simple assessment of model latent heat flux biases arising either from biases in water vapor or surface wind speed as well as other empirical coefficients in the model. Sample POD result shows that, compared to ‘’observed’’ fluxes also estimated from water vapor and surface wind speed measured at tropical moorings, tropical latent heat fluxes in the NCAR CEMS2 models are significantly overestimated when extreme water vapor or surface wind speed happens.