Statistical Downscaling of Daily Precipitation from ERA5 Atmospheric Fields to IMERG over West Africa
Authors: David Sagna, Redouane Lguensat, Samo Diatta
Organizations: Laboratoire d’Océanographie et des Sciences de l’Environnement et du Climat (LOSEC), Université Assane Seck de Ziguinchor, BP 523, Ziguinchor, Senegal · Institut Pierre-Simon Laplace (IPSL), Sorbonne Université, IRD, 75005 Paris, France
Global reanalyses are too coarse to resolve the convective systems that produce most West African monsoon rainfall. In this work, we use U-Net based architectures to downscale daily ERA5 fields (0.25°) to IMERG V07 satellite precipitation (0.1°) over West Africa for the June to October seasons of 2000 to 2024, using geopotential, temperature, humidity and wind at three pressure levels together with elevation and land fraction. Adding ERA5 precipitation as input slightly improves performance for this particular challenging region. We find that all the deterministic networks investigated for this work are too dry and lose almost all observed variance below a few hundred kilometers, which motivated our use of a GAN-based alternative. We also investigated permutation importance for the U-Net that ranks 500 hPa specific humidity first, consistent with the sensitivity of deep convection to mid-tropospheric moisture.
Figures & tables
Figure 1: Study domain and mean IMERG V07 precipitation over the June–October seasons of 2000–2024 (mm d -1 ). The black line is the coastline (land fraction 0.5), the brown contours mark elevations of 500 and 1000 m, and the dashed lines at 10°N and 15°N separate the Guinean, Sudanian and Sahelian zones.
RMSE
MAE
Bias
r
P99
FSS20
CSI5
CSI20
CSI40
ratio
Raw outputs
Climatology
9.94
4.82
− 0.13
0.386
0.34
0.064
0.334
0.025
0.000
ERA5 precipitation
10.71
4.17
− 0.38
0.465
0.85
0.410
0.399
0.188
0.125
BCSD
11.18
4.54
+ 0.12
0.477
1.05
0.446
0.386
0.215
0.141
DeepESD
8.88
3.55
− 1.07
0.581
0.67
0.464
0.416
0.223
0.143
Table 1: Scores on the test seasons (2021–2024, all pixel-days of the study domain), without post-processing (top) and after quantile mapping (bottom). RMSE, MAE and bias in mm d -1 . The best value of each column within each block is in bold; bias and P99 ratio have no best value.
Figure 2: Daily precipitation on three test days at the 50th, 90th and 99th percentiles of the domain-mean observed precipitation (4.1, 6.4 and 8.2 mm d -1 ): IMERG, ERA5 precipitation, DeepESD, the U-Net, the U-Net after quantile mapping, and one member of the residual Wasserstein GAN. Values below 1 mm d -1 are white. The number in each panel is the spatial correlation with IMERG on that day.
Figure 3: Left: mean bias over the test seasons (mm d -1 ; brown too dry, green too wet), with its domain mean. Right: daily correlation with IMERG at each pixel, with its domain mean. Rows: ERA5 precipitation, BCSD, DeepESD, the U-Net without and with quantile mapping, and one member of the residual Wasserstein GAN; the U-Net with ERA5 precipitation is not shown. The dashed lines separate the three zones.
Figure 4: FSS20 (a, b) and CSI20 (c, d) on the test seasons, by zone and over land (a, c) and by month (b, d). Bars (a, c) and solid lines (b, d) show the scores after quantile mapping (QM); black ticks (a, c) and dotted lines (b, d) show the raw scores. The GAN is represented by one member.
Figure 5: Radially averaged power spectral density of the daily fields over the test seasons, for IMERG and the models, without post-processing (a) and after quantile mapping (b). Spectra are computed with pysteps on two 190×190 windows, each with its mean removed and a Hann taper, and averaged over the windows and the days. The curve of the GAN is that of one member.
CRPS
SSR
SSR 2°
Outside (%)
RMSE mean
FSS20 mean
GAN
2.56
0.92
0.76
27
8.99
0.494
U-Net + ERA5 precip.
3.41
—
—
—
8.58
0.516
s/e 300 km
s/e 100 km
s/e 50 km
P 100 km
P 50 km
Areas (size)
GAN
0.96
1.03
1.15
1.09
1.37
96 (3)
U-Net + ERA5 precip.
—
—
—
< 0.01
< 0.01
2 (514)
IMERG
—
—
—
1
1
101 (3)
Table 2: The ensemble of the GAN on the test seasons (ten members). Top: CRPS (mm d -1 ), spread-skill ratio (SSR) at the pixel scale and for 2° boxes, percentage of the pixel-days with IMERG outside the range of the members, among those on which IMERG or a member has at least 1 mm d -1 (18% for a calibrated ensemble), and RMSE (mm d -1 ) and FSS20 of the ensemble mean; for the U-Net with ERA5 precipitation, on which the GAN builds, the CRPS is the MAE and the ensemble mean is the prediction. Bottom: ratio of the power of the differences between members to that of the errors (s/e) at 300, 100 and 50 km, ratio of the power of one member to that of IMERG ( P ) at 100 and 50 km, and number per day and median size (cells) of the areas with at least 20 mm d -1 .
Figure 6: Top: the test day at the 99th percentile of domain-mean precipitation (23 August 2024): IMERG, the prediction of the U-Net with ERA5 precipitation, on which the GAN builds, two members of the GAN, and the mean and the standard deviation of its ten members. (a) Rank of IMERG among the ten members over all test pixel-days (grey) and over the pixel-days on which IMERG or at least one member has at least 1 mm d -1 (red); amounts below 0.1 mm d -1 count as zero and ties are broken at random; the dashed line is the flat histogram of a calibrated ensemble. (b) Power spectra, computed as in Figure 5 , of the difference between a member and IMERG (error, five members) and between two members (spread, five pairs), averaged over the test days; for a calibrated ensemble the two curves coincide.
Figure 7: Increase in the RMSE of the U-Net on the test seasons when one input channel is permuted between days, for the whole domain (bars, with the standard deviation over three permutations) and for each zone (symbols). The RMSE without permutation is 8.61 mm d -1 .
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
Upper tail
Highest level
Pooling
RMSE
Bias
r
P99
FSS20
CSI20
CSI40
No mapping
10.72
− 0.35
0.449
0.81
0.370
0.162
0.110
Ratio
maximum
month, 5×5
19.19
+ 1.38
0.317
1.53
0.386
0.175
0.105
Ratio
0.99
month, 5×5
17.54
+ 1.11
0.331
1.39
0.387
0.176
0.106
Difference
maximum
month, 5×5
13.97
+ 0.82
0.396
1.32
0.390
0.177
0.110
Difference
0.99
month, 5×5
12.90
+ 0.55
0.413
1.18
0.393
0.179
0.115
Cap
maximum
month, 5×5
12.87
+ 0.59
0.406
1.21
0.391
0.178
0.111
Appendix
Table 3: Two-fold cross-validation of quantile-mapping designs within the validation seasons, for the ERA5 precipitation baseline. “Ratio” and “difference” extend the correction above the highest fitted quantile with a constant ratio or difference; “cap” sets those values to the highest fitted observed quantile. The retained design is in bold.
Figure 8: Training (dashed) and validation (solid) loss of the U-Nets (a) and of DeepESD (b). Dots mark the retained epochs; epochs are counted from 1.
RMSE
MAE
Bias
r
P99
FSS20
CSI5
CSI20
CSI40
ratio
Climatology
9.94
4.87
− 0.19
0.383
0.34
0.058
0.340
0.022
0.000
ERA5 precipitation
10.72
4.32
− 0.35
0.449
0.81
0.370
0.396
0.162
0.110
BCSD
11.21
4.73
+ 0.23
0.459
1.01
0.417
0.378
0.194
0.126
DeepESD
8.92
3.66
− 1.03
0.572
0.63
0.427
0.420
0.198
0.123
U-Net
8.67
3.51
− 0.95
0.602
0.65
0.459
0.452
0.222
0.140
Appendix
Table 4: Scores on the validation seasons (2019–2020), without post-processing. Same layout as Table 1 .
RTX 5090
RTX 2080 Ti
CPU, 8 threads
CPU, 1 thread
DeepESD
0.18 (0.05)
0.64 (0.39)
2.6
4.3
U-Net
2.8 (2.2)
9.7 (6.9)
263
913
U-Net + ERA5 precip.
2.9 (2.2)
9.8 (6.9)
295
860
GAN, ten members
20.0 (11.3)
50.6 (34.3)
1200
3340
Appendix
Table 5: Time to downscale one day of the test seasons with the trained models (ms per day), one day at a time and, on the GPUs, in batches of 16 days (in brackets). The GAN generates ten members. CPU: Intel Core i5-14600KF, with eight threads and with one thread.
Generative machine learning is an increasingly important complement to dynamical downscaling for producing high-resolution precipitation projections, with diffusion models currently the leading approach. Flow matching is a related generative framework that has recently achieved strong results across image, video and other domains, and shown early promise for downscaling. We train a flow matching model to map daily precipitation from 8 km to 2 km over a convective-scale domain centred on Singapore, and benchmark it against CPMGEM, a score-based diffusion model. Flow matching achieves consistently better spatial skill: higher fractions skill score at every precipitation threshold and neighbourhood scale tested, and tighter structure and amplitude components of the SAL score with comparable location skill. However, flow matching underestimates the upper tail of the precipitation distribution, resulting in a dry bias in the climatological mean. These results suggest that flow matching is a competitive generative framework for convective-scale precipitation downscaling, particularly well suited to capturing spatial structure.
Global reanalysis products such as ERA5-Land provide spatially complete weather fields but at resolutions too coarse for local applications, particularly in mountainous regions where temperature can vary by several degrees over short distances. This project investigates Convolutional Conditional Neural Processes (ConvCNPs) for statistical downscaling of daily maximum temperature from the ~11km resolution ERA5-Land grid to ~1km resolution over Switzerland, building upon the architecture of Vaughan et al. (2022) and adapting it to the topographically complex Swiss domain with high-resolution elevation features from the swisstopo DHM25. The best model, trained on ten years of data (2014-2023) with five-fold temporal cross-validation, achieves a mean absolute error of 1.31 Celsius and a CRPS-based skill score of 0.524 relative to bilinear interpolation, reducing the expected prediction error by more than half. An ablation study reveals that the elevation MLP is the indispensable component - without it, the model diverges entirely - while explicit seasonal features and Topographic Position Index provide secondary benefits. Under sparse on-grid input the model degrades gracefully, maintaining positive skill down to approximately 10% of the input grid; however, zero-shot deployment on off-grid station observations does not achieve positive skill at any density tested. All configurations exhibit severely overconfident uncertainty estimates, a structural limitation of the Gaussian likelihood training objective. These results demonstrate that ConvCNPs are a viable and effective approach to climate downscaling in complex terrain, and identify uncertainty calibration and native support for non-gridded input as the key challenges for operational deployment.
High-resolution precipitation information is essential for climate impact assessment, yet global climate models remain too coarse to resolve key small-scale processes. Existing machine learning downscaling methods often require paired low- and high-resolution data for supervised learning, are tied to fixed regions or scale factors during inference, and can be computationally expensive to train and run in physical space. Here we introduce Longwang, a zero-shot latent generative framework for global spatiotemporal precipitation downscaling. Longwang learns a context-conditioned latent generative prior and combines it with a physically informed observation operator through posterior sampling, enabling daily O(10 km) precipitation fields to be generated from monthly O(100 km) inputs. On ERA5 reanalysis, Longwang outperforms standard posterior sampling with an unconditional generative prior in reconstructing fine-scale spatial patterns, preserving temporal coherence, and recovering extreme precipitation intensities. The framework further generalizes to historical climate simulations and future climate projections under substantial distribution shift.
Yue Wang, Daniele Visioni
1*Department of Earth and Atmospheric Sciences, Cornell University, 112 Hollister Drive, Ithaca, 14853, NY, USA.