GloH2O

Menu

SaRa

Saudi Rainfall (SaRa)

SaRa provides hourly 0.1° resolution precip­i­tation data from 1979 to the present with a 2‑hour delay for the Arabian Peninsula. 

Superior accuracy 

SaRa delivers high-resolution rainfall data tailored specif­i­cally for the Arabian Peninsula, one of the most arid and data-sparse regions on Earth. With hourly updates at a 0.1° resolution, SaRa provides unpar­al­leled accuracy for hydro­logical modeling, flood management, and climate research. 

AI-powered 

SaRa integrates 18 advanced machine learning models to fuse satellite and model-based precip­i­tation data. Evaluated against 20 global products, SaRa consis­tently ranks first across all perfor­mance metrics, including bias and corre­lation, ensuring you get the most reliable rainfall estimates. 

Near real-time 

Stay informed with SaRa’s near real-time updates—data is available with less than 2‑hour latency. Plus, access a compre­hensive historical rainfall record spanning back to 1979, enabling long-term analysis for climate trends, water resource management, and disaster risk mitigation. 

Last week’s SaRa rainfall

Precip­i­tation patterns across the Arabian Peninsula from SaRa over the past 7 days, with time expressed in UTC. Each frame repre­sents the hourly precip­i­tation accumu­lation. The video is updated hourly. 

Algorithm

SaRa leverages cutting-edge machine learning to seamlessly merge and bias-correct satellite and model data, making it the most advanced precipitation product on the market.

The algorithm behind SaRa involves 18 distinct machine learning model stacks trained on various combi­na­tions of satellite and (re)analysis precip­i­tation products along with static predictors. Each model stack integrates four sub-models that are applied succes­sively to estimate daily precip­i­tation, disag­gregate it into finer temporal resolu­tions, correct proba­bility distri­b­u­tions for wet days and peak magni­tudes, and refine these estimates to hourly intervals, respec­tively. This approach ensures SaRa delivers accurate, high-resolution precip­i­tation data suitable for a wide range of applications.

Details of the algorithm are available in the following peer-reviewed scien­tific publication:

Wang, X., Alharbi, R. S., Baez-Villanueva, O. M., Green, A., McCabe, M. F., Wada, Y., Van Dijk, A. I. J. M., Abid, M. A., and Beck, H. E. (2025). Saudi Rainfall (SaRa): Hourly 0.1° Gridded Rainfall (1979–Present) for Saudi Arabia via Machine Learning Fusion of Satellite and Model Data. Hydrology and Earth System Sciences, 29, 4983–5003.

Proven performance

We conducted the most extensive evaluation to date of gridded precipitation products in the Arabian Peninsula.

In our compre­hensive peer-reviewed study — freely acces­sible here — SaRa emerged as the top performer in accuracy compared to other precip­i­tation products. The bar chart on the right presents normalized median daily Kling-Gupta Efficiency (KGE) scores from over 12,000 independent, quality-controlled rain gauges worldwide. KGE scores provides a useful summary of how well each product reflects real-world condi­tions by capturing key factors like timing, variability, and bias. 

SaRa versus other products

60%
55%
46%
44%
44%
42%
38%
37%
27%
24%
20%
16%
16%
15%
12%
1%
PERSIANN-CCS -22%

SaRa model_06
SaRa model_01*
GDAS
CMORPH-RT
ERA5*
SM2RAIN+GPM*
JRA-3Q*
PDIR-Now
CPC Unified*
CHIRP*
GSMaP-Std V8
CMORPH-RAW
PERS-CCSCDR*
SM2RAIN-ASCAT
SM2RAIN-CCI*
IMERG-Late V7
PERSIANN-CCS*

Performance comparison between MSWEP and other widely used precipitation datasets, based on independent rain gauge observations in Saudi Arabia. Values are rescaled median Kling Gupta Efficiency (KGE) scores, where a KGE of −0.41 (equivalent to using the average observed value) is set to 0, and a perfect KGE of 1 is set to 1. Datasets marked with * are not available in near real‑time.

SaRa versus ERA5

Our dataset surpasses ERA5 in all evaluated metrics

1 %
Better
correlation
1 %
Lower peak
bias
1 %
Lower wet-
day bias
1 %
Better event detection

Saudi Arabia faces worsening water scarcity and flash floods from climate change and population growth. SaRa provides real-time rainfall data to support response, planning, and policy under Vision 2030.

Dr. Hylke Beck
Founder and Technical Director, GloH2O

Data access

SaRa is free for non-commercial use. Contact us for commercial licensing.

SaRa is released under the Creative Commons Attri­bution-NonCom­mercial (CC BY-NC 4.0) license. Simply request access and we’ll send you an email with download instructions.

Note: SaRa data has recently been integrated into MSWEP. Please request access to MSWEP below.

Who’s Using MSWEP?

Leading organi­za­tions relying on MSWEP in their analyses. 
anu_logo-1
European_Commission.svg-1
exeter_logo-1
ghent_logo-1
iiasa_logo-1
KIT-1
KU_Leuven_logo.svg-1
National-Geographic-Logo-1
princeton-logo_1-1
TH_Koeln_Logo-1
tudelft_logo-1
UC_Davis_Logo-1
unesco_logo-1
university-of-melbourne-logo-1
unsw_logo_-1
wageningen_logo-1
wsl-1
nasa_logo-2

Easily access SaRa using the GloH2O API

The GloH2O API offers seamless access to a wide range of meteo­ro­logical datasets, including SaRa, all updated in near real-time. 
GloH2O API

Frequently asked questions

SaRa stands out among gridded precip­i­tation products for its use of machine learning algorithms to seamlessly blend and bias-correct satellite- and model-based precip­i­tation estimates, resulting in signif­i­cantly improved accuracy. With high hourly resolution at 0.1°, a historical data record starting from 1979, and near real-time updates, SaRa supports a wide range of appli­ca­tions, including water resource management, hydro­logical modeling, agricul­tural planning, disaster risk reduction, and climate studies. Unlike many other products on the market, SaRa is built on trans­parent, openly acces­sible algorithms and rigor­ously validated methods that have undergone thorough scien­tific peer review, ensuring both relia­bility and credi­bility for research and opera­tional use.

The “Past_nogauge” and “NRT” variants of SaRa work together to provide a seamless and continuous dataset. The “Past_nogauge” variant repre­sents historical satellite-model merged data, while the “NRT” variant extends this record into near real-time with a latency of approx­i­mately 2 hours. Together, these two variants offer compre­hensive coverage from 1979 to just 2 hours before the present, ensuring both historical depth and near real-time availability.

Gauge under-catch correc­tions were not applied in SaRa because users generally prefer precip­i­tation estimates that closely reflect raw rain gauge obser­va­tions, and such correc­tions often introduce signif­icant uncer­tainty. This uncer­tainty stems from factors like wind speed, gauge design, precip­i­tation type, and the limited avail­ability of metadata about gauge setups. Additionally, the empirical models used for these correc­tions may not accurately account for local condi­tions, further reducing their reliability.

SaRa leverages a state-of-the-art machine learning algorithm to merge and bias-correct satellite and model data, deliv­ering improved accuracy and relia­bility. In addition to this advanced algorithm, SaRa incor­po­rates additional data sources, effec­tively reduces drizzle overes­ti­mation, and minimizes peak under­es­ti­mation compared to MSWEP V2, ensuring superior perfor­mance across various applications.

Non-commercial users can currently access SaRa exclu­sively through Google Drive, utilizing rclone as outlined in the access instruc­tions provided via email. In contrast, commercial users benefit from additional download options, including FTP and API access, offering greater flexi­bility and convenience.