Galapagos Rainfall retrieval using multispectral GOES-16 infrared brightness temperature _ Part 1: Rainfall area delineation

The Galápagos archipelago is renowned for its exceptional and diverse flora and fauna, primarily due to its unique location and climate. However, since the archipelago has limited access to permanent freshwater sources, available freshwater depends on rainfall. Unfortunately, information is scarce regarding the spatial and temporal distribution of rainfall in the Galápagos, making it challenging to fully comprehend short- and long-term rainfall patterns and changes.  This is particularly important considering extreme rainfall events caused by climate change, as highlighted in the IPCC report. IPCC emphasizes that the increased frequency and severity of extreme events caused by climate change impact the destruction and harm to both nature and people. 

For such regions, satellite-based rainfall products potentially represent a source of reliable and area-wide data on rainfall. Therefore, the aim of this project is to develop the Galápagos Rainfall Retrieval (GRR) product;, a new satellite-based algorithm for retrieving rainfall in Galapagos archipelago. The GRR has the potential to provide high spatio-temporal resolution rainfall data (2 km, 10 min) in near real-time for the study region. 

 

The GRR algorithm combines physical methods with machine learning techniques using sequences of Geostationary Earth Orbit infrared (GEO-IR) images to retrieve both cold season Garua drizzle and warm season convective rainfall. The algorithm comprises of two main steps. i) Rain area delineation: a threshold technique and spectral spatial analysis are utilized to identify areas with cloud cover and differentiate between low, middle, and high clouds. Following this, a slope test and machine learning algorithm are used to classify cloud-covered areas and identify low stratus/Garua drizzle and potentially convective core regions. The convective cores are then evaluated to determine whether they are decaying or not, and subsequently labeled as stratiform rain or active convective cores, respectively. ii) Rain rate assignment is carried out by training random forest regression models separately for convective and stratiform cells, based on microwave-only IMERG-V06 rainfall data. To train rainfall rates for Garua detected regions, CloudSat and the newly installed automated weather station (AWS) network from DARWIN project is utilized. 

 

The GRR product is set to be developed between 1/1/2022-1/1/2023 and then applied to the entire available GOES-16 dataset. Independent microwave-only IMERG-V06 rainfall data and AWS network with a high temporal resolution of 10 minutes will be used for validation purposes. The AWS network will cover W-E and luff-lee transects over three islands (Isabela, S. Cruz, S. Cristóbal). 

 The poster will present the overall structure of the GRR algorithm and some first results of the rain area delineation.