Studying cloud properties from space using sounder data: A preparatory study for INSAT-3D Munn V. Shukla and P. K. Thapliyal Atmospheric Sciences Division Atmospheric and Oceanic Sciences Group Space Applications Centre (ISRO), Ahmedabad-380015,India. ( munnvinayak@sac.isro.gov.in, munnvinayak@gmail.com ) Abstract Satellite derived parameters like temperature and humidity profiles, cloud top pressure, cloud top temperature, total precipitable water etc. are very crucial for the improvement in weather forecasting. India is scheduled to launch INSAT-3D in a geostationary orbit with 18 channel infrared (plus a visible channel) sounder. The sounder on board INSAT-3D will play a major role in providing observations over extended regions particularly measurements over data sparse Indian Ocean region. In addition to it INSAT-3D sounder will also give the information regarding the cloud cover and cloud top pressure and temperature, which in turn will be very helpful for nowcasting and for severe weather prediction. This working paper is aimed at cloud detection and computation of cloud top pressure using INSAT-3D Sounder observations. For this purpose, various threshold tests have been formulated for the cloud detection and CO 2 -slicing technique has been used for the computation of cloud top pressure. The algorithm has been tested with GOES sounder dataset over Continental US region. For cloud detection we have tried minimal use of the ancillary dataset to avoid the effect of ancillary data bias on the cloud flag. The cloud flag tests were validated with MODIS cloud flag and cloud top pressure has been validated with CALIPSO data. 1. Introduction: The satellite observations had become the essential tool for weather forecasting. India has launched INSAT series of satellites for better weather monitoring and its predictions. India is planning to launch the first Sounder onboard INSAT-3D to provide the temperature and humidity profiles over data sparse region on continuous temporal scale. The INSAT-3D sounder instrument will have 18 IR band and one visible band for such measurements. Many earlier studies have developed algorithms for temperature and humidity measurements over regions that are not contaminated by clouds. In this direction various studies have been made to detect clouds from the sounder instruments. The accurate cloud flagging is a crucial precursor to the success of
any retrieval methodology. This working paper presents a methodology for the cloud flagging using sounder data and its validation with high spatial resolution standard cloud flags available from MODIS. To make a show case for the utility of INSAT-3D sounder data for the cloud flagging, GOES-13 sounder data is used and compared with MODIS cloud flag. 2. Methodology: The INSAT-3D sounder instrument has similarity with the GOES-13 sounder instrument. Therefore, this working paper presents implementation and validation of the cloud detection algorithm for INSAT-3D based on GOES-13 cloud flagging. The GOES-13 sounder has multispectral threshold based algorithm for cloud flagging. The following channels are being used: Band number and Central Band width Use in Cloud flagging 3-13 µm CO2 absorption band, used for detecting thin cirrus cloud 6-12.6 µm Dirty Window used in detecting some low level inversions following cold polar outbreaks 7-12.0 µm Dirty Window used in detecting low clouds and fog 8-11.0 µm Long Wave Window used in detecting various level of clouds 17-3.9 µm Short Wave Window used to detect low clouds and fog at night 18-3.7 µm Short Wave Window used to detect low clouds after sunrise 19-0.6 µm Visible band used in detecting various level of clouds during daylight The schematic diagram of cloud detection technique is given in Fig.1. For the cloud detection scheme only surface observations are taken as ancillary information. For using regression equation surface skin temperature is computed and used for first gross flag for cloud. The first flag is called primary array flag and is based on visible and window channel brightness temperature and computed skin temperature. Those FOVs which pass the primary array flags are tested for other checks. There are cold sea test, regressed skin temperature test and reflected sunglint test in secondary array flag. The secondary array flag is further tested by individual FOV flag. The tests of individual FOV flag are the final test for the given FOV.
Fig.1: schematic diagram of cloud detection technique 3. Results: The algorithm has been implemented on GOES-13 sounder data. The example is shown and tested on data for Julian day#227 at 17 GMT. The 17 GMT data is taken for the collocation with MODIS pass available over US CONUS region. The cloud flag is shown in fig.2. The GOES-13 cloud mask is compared with MODIS cloud flag. The spatial resolution of GOES- 13 and MODIS are different. The MODIS spatial resolution for cloud flag is ~ 1 km at nadir, whereas, the size of one GOES-13 FOV at nadir point is ~8km. As the zenith angle increases the size of FOV increases in both the cases. For the comparison the MODIS FOVs are remapped over GOES FOVs. The distribution of MODIS FOV in one GOES-13 FOV is given in fig.3.
Fig.2: GOES-13 cloud mask for 227 Julian day 17 GMT Fig.3: Distribution of MODIS FOV in one GOES FOV
In MODIS cloud flag there are four kinds of flags are given: 1) Clear with 99% confidence 2) Clear with 95% confidence 3) Clear with 66% confidence 4) Cloudy The distribution of various categories are given in fig.4(a-d). Fig.4a:distribution of 99% confidence clear Fig.4b: distribution of 95% confidence clear Fig.4c: distribution of 66% confidence clear Fig.4d: distribution of cloudy pixels
Fig.5: MODIS cloud mask over GOES-13 CONUS region. To validate GOES-13 mask, the 99% confidence and 95% confidence clear pixels are assumed to be clear. The cloudy pixels are taken as cloudy and rest of the pixels are assumed to be unknown. In fig.5 the MODIS cloud mask over US CONUS region is shown. 4. Future Work: The validation results shows a good match of GOES cloud mask, still it leads a wide scope for the improvement in the cloud mask algorithm. The cloud flagging based on principal component analysis can be used for the further improvement. The CO 2 slicing method will also be applied to compute cloud top pressure and effective cloud amount over the cloudy pixels.