The research team, led by Marcelo Sacardi Biudes of the Federal University of Mato Grosso, evaluated twelve widely used gridded precipitation products against ground truth from 120 automatic weather stations operated by the Brazilian National Institute of Meteorology. The stations, which passed strict quality filters requiring less than 10 percent missing data, were distributed across four biomes: the Amazon, the Cerrado savanna, the Pantanal wetlands, and fragments of the Atlantic Forest. The study covered the 2000 to 2020 period, and the researchers deliberately avoided any gap-filling or interpolation of the station records, preserving the raw observational signal as their benchmark. Hourly measurements were summed into daily totals, and monthly and annual values were computed only from complete records, ensuring that the comparison was never contaminated by artificial variability.

The twelve datasets represented the full spectrum of modern precipitation estimation. Satellite-based products such as TRMM, CHIRPS v3, GPM IMERG V07B, GSMaP, PERSIANN-CDR, and GPCP derive rainfall from infrared and passive microwave sensors, often corrected with gauge data. Reanalysis products, including ERA5-Land, GLDAS, and MERRA-2, generate physically consistent precipitation fields through numerical modeling and data assimilation. Gauge-based and multi-source products—CPC, the Brazilian Daily Weather Gridded Data known as BR-DWGD, and the Multi-Source Weighted-Ensemble Precipitation dataset MSWEP—lean directly on station observations and merging procedures. The team also constructed an unweighted ensemble mean of all twelve products to test whether averaging the field improves reliability, a common assumption in hydrological practice.

The validation framework was unusually comprehensive. Continuous statistical metrics quantified error magnitude, bias, and agreement: root mean squared error, mean absolute error, percent bias, Willmott’s index of agreement, Spearman’s rank correlation, and Lin’s concordance correlation coefficient, which jointly assesses precision and accuracy against the one-to-one line. These were combined into a Composite Model Ranking that treated every metric equally, producing a single score for each dataset at daily, monthly, and annual scales, for the region as a whole, for each biome, and even for each individual station. On top of that, the researchers ran a categorical event-detection analysis at four rainfall thresholds—1, 10, 20, and 50 millimeters per day—classifying every daily estimate as a hit, a miss, or a false alarm, and computing the probability of detection, the false alarm ratio, and the critical success index.

The headline finding concerns a home-grown product. BR-DWGD, a high-resolution gauge-based dataset developed specifically for Brazil, delivered the best overall performance across all temporal scales. It recorded the lowest root mean squared error at the daily, monthly, and annual scales—6.77, 48.63, and 253.39 millimeters, respectively—and posted the highest agreement coefficients in nearly every case. Its advantage likely stems from its design: because it is built on Brazil’s own dense observational archive, it captures regional precipitation gradients and long-term accumulation patterns that global products often smooth away. The gauge-based CPC dataset was the runner-up at the daily scale, achieving the lowest mean absolute error of 2.64 millimeters and the second-lowest root mean squared error, confirming its strength in tracking short-term rainfall variability.

Among the satellite-derived products, CHIRPS v3 emerged as the strongest performer at monthly and annual aggregations, with relatively low errors and high concordance. The ensemble mean also ranked among the top approaches at these longer scales, with root mean squared errors of 53.74 and 275.97 millimeters at the monthly and annual scales, respectively. Yet in a result that challenges a popular assumption, the ensemble never outperformed BR-DWGD. Its real value, the authors conclude, lies in stability: by averaging out the idiosyncratic errors of individual products, the ensemble delivers consistent estimates across scales and biomes rather than superior accuracy in any single metric. For applications demanding the closest match to station observations, the ensemble is not a substitute for the best individual dataset.

At the other end of the spectrum, MERRA-2, GSMaP, MSWEP, and in some cases GPCP and TRMM showed the weakest results. MERRA-2 posted the highest errors at every scale, reaching 12.88 millimeters daily and 400.85 millimeters annually, alongside the lowest agreement coefficients. GSMaP underestimated mean daily precipitation by roughly 25 percent regionally, and by about 19 and 22 percent in the Amazon and Cerrado, consistent with its negative percent bias of minus 13.20 percent. Most satellite products, including GPCP, GPM IMERG, PERSIANN-CDR, and TRMM, showed negative biases of around minus 11 to minus 12 percent, a systematic tendency to underreport rainfall in this convective tropical environment. In contrast, CPC overestimated, while ERA5-Land, BR-DWGD, and the ensemble kept biases comparatively small.

Perhaps the most sobering result came from the event-detection analysis. Detection skill declined steadily as rainfall intensity increased. For light rain at the 1 millimeter threshold, most datasets detected events reliably with moderate-to-high success ratios. But at 20 and 50 millimeters per day—exactly the intensities that matter for floods, soil erosion, and crop damage—probabilities of detection and critical success indices dropped across the board. This is a fundamental limitation of both satellite retrievals and model-based estimates in the tropics, where rainfall is dominated by short-lived, spatially patchy convective storms that fall between the pixels of a gridded product. BR-DWGD and CPC again proved most robust across thresholds, with CPC showing consistently high success ratios, meaning fewer false alarms, while products such as MERRA-2, GLDAS, TRMM, GPCP, GSMaP, and MSWEP suffered the steepest skill losses for intense events.

The station-level spatial analysis added a final layer of nuance. When the Composite Model Ranking was computed independently for each of the 120 stations, the best-performing dataset varied from place to place and from scale to scale. BR-DWGD and CPC were most frequently the local champions at daily and monthly scales, while annual-scale performance was more spatially heterogeneous. The ensemble maintained moderate-to-high rankings across most of the region, but even it showed lower values at some stations for long-term accumulations. The practical implication is clear: regional or biome-level averages can mask local differences that matter enormously for a specific watershed, farm, or conservation area, so dataset selection should always include a localized check.

The study’s authors acknowledge important caveats. Station density varies dramatically among biomes, from 2.80 stations per 10,000 square kilometers in the Atlantic Forest fragments to just 0.27 in the Pantanal, which is represented by only four stations and therefore demands cautious interpretation. Point-to-pixel comparisons cannot fully capture sub-grid convective variability, and gauge-based products such as BR-DWGD, CPC, CHIRPS v3, and MSWEP are not fully independent of the very observations used to validate them. Even so, the guidance for practitioners is concrete: BR-DWGD is the strongest reference for applications requiring high agreement with station data, CPC excels for daily monitoring and event detection, and CHIRPS v3 or the ensemble offer competitive alternatives at monthly and annual scales. In a region that produces a large share of Brazil’s food, regulates continental water cycles, and faces intensifying climate extremes, choosing the right rainfall map is no longer a technical afterthought—it is a decision with real consequences, and this study provides the evidence to make it well.

Subject of Research: Multi-scale validation of gridded precipitation datasets against weather station observations in Central-West Brazil

Article Title: Multi-scale validation of gridded precipitation datasets in Central-West Brazil

Article References: Biudes, M. S., Machado, N. G., dos Santos, L. O. F., & da Silva Lotufo, J. B. (2026). Multi-scale validation of gridded precipitation datasets in Central-West Brazil. Theoretical and Applied Climatology, 157(10), Article 623. https://doi.org/10.1007/s00704-026-06546-8

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06546-8

Keywords: precipitation datasets, Central-West Brazil, gridded climate data, satellite rainfall estimation, BR-DWGD, CHIRPS, GPM IMERG, reanalysis, rainfall validation, Cerrado, Amazon, hydrological modeling

The study, published in Theoretical and Applied Climatology, targets the IMERG product, the Integrated Multi-satellitE Retrievals for GPM, which is generated by NASA’s Global Precipitation Measurement mission. IMERG stitches together observations from a constellation of satellites to produce near-global rainfall maps at high spatial and temporal resolution. Yet over the interior of the Tibetan Plateau, the product suffers from well-documented biases. It tends to overestimate both the intensity and the spatial extent of precipitation in some areas while underestimating rainfall locally in others, and its errors grow worse as rainfall intensifies. Because the plateau feeds the headwaters of major Asian rivers and drives regional hydrological cycles, these errors ripple directly into flood simulation, drought monitoring, and water resource management.

Lead author Jiaxin Zhao and colleagues, including Aili Liu, Chong Wang, Lin Zhao, and Tingting Dong, tackled the problem with a two-stage machine learning pipeline that pairs a convolutional neural network, or CNN, with Extreme Gradient Boosting, better known as XGBoost. The key innovation is that the framework is explicitly aware of precipitation intensity. Rather than treating all rainy days as a single undifferentiated class, the system first classifies each satellite estimate into rainfall categories and then uses that classification to guide the correction of the rainfall amounts themselves.

The technical architecture works as follows. The researchers trained a CNN-based classification model using IMERG estimates together with nine auxiliary topographic and meteorological variables, such as terrain characteristics and atmospheric conditions, with ground-based gauge observations serving as the reference truth. The CNN learns to recognize the spatial patterns and contextual signatures that distinguish light rain from moderate, heavy, and torrential rainfall in satellite data. Crucially, the model outputs not just a single category label but a set of probabilities across all rainfall classes. These probability vectors are then fed into the XGBoost regressor as additional features, alongside the original IMERG values and environmental variables, allowing the gradient-boosted trees to correct the magnitude of rainfall differently depending on how likely an event is to be a deluge rather than a drizzle.

This division of labor exploits the complementary strengths of the two algorithms. CNNs excel at extracting spatial features from gridded data, making them well suited to interpreting the two-dimensional structure of satellite precipitation fields and their relationship to the plateau’s complex topography. XGBoost, a scalable tree-boosting system renowned for its performance on tabular regression problems, handles the numerical correction task efficiently and robustly. By passing the CNN’s rainfall-category probabilities into the regression stage, the framework ensures that the intensity information directly shapes the bias correction, rather than being lost in a one-size-fits-all adjustment.

To evaluate the approach, the team benchmarked the corrected IMERG estimates against two independent, long-term, high-resolution meteorological datasets: the Third Pole long-term high-resolution meteorological forcing dataset, known as TPMFD, and the China Meteorological Forcing Dataset, or CMFD. Both are widely trusted reference products for the region, and comparing against two separate benchmarks helps guard against the possibility that improvements simply reflect fitting to one particular dataset’s quirks.

The results are striking. On a daily timescale, the corrected IMERG substantially outperformed the original product, with the correlation coefficient increasing by 0.15 and the root mean square error and mean absolute error decreasing by 0.46 and 0.04 millimeters per day, respectively. The framework also proved seasonally robust, maintaining satisfactory performance in both the dry and wet seasons, with correlation coefficients of 0.35 and 0.65. In a region where monsoon dynamics produce sharp seasonal contrasts in precipitation regime, that consistency matters: many correction methods tuned for the wet season degrade badly during dry months, or vice versa.

The spatial improvements are equally significant. The original IMERG product over the plateau’s interior contains spurious high-value regions where it paints unrealistic pockets of intense rainfall, alongside localized zones of underestimation. The corrected product suppressed these artifacts, producing rainfall maps whose spatial patterns align much more closely with gauge observations. When compared with the CMFD reference, the corrected IMERG reduced the overestimation of both precipitation intensity and spatial coverage, two of the most persistent failure modes of satellite retrieval over high terrain, where orographic effects and complex cloud microphysics confound passive microwave and infrared retrieval algorithms.

Perhaps the most consequential gains came in the extreme rainfall categories. For heavy and torrential rainfall events, the corrected IMERG achieved correlation coefficient increases of 0.14 to 0.31 and reductions in root mean square error of 1.47 to 2.01 millimeters per day relative to the original product, all while maintaining stable performance for light and moderate rainfall. That balance is rare. Conventional bias correction methods often improve average conditions at the cost of smearing out extremes, or they sharpen extremes at the cost of distorting the drizzle that dominates the record. An intensity-aware framework that improves both ends of the spectrum simultaneously represents a meaningful advance for hydrological extreme-event simulation, where the difference between a corrected and uncorrected heavy rainfall estimate can determine whether a flood model captures a real hazard or misses it entirely.

The implications extend well beyond the Tibetan Plateau. Roughly a third of humanity depends on water that originates in High Mountain Asia, and as climate change intensifies both the hydrological cycle and the frequency of extreme precipitation, the demand for accurate, gauge-independent rainfall estimates in ungauged mountains will only grow. The framework’s ingredients are all publicly available: the IMERG product is distributed freely by NASA, and the TPMFD and CMFD datasets are hosted by the National Tibetan Plateau Data Center. That openness, combined with the modular design of the CNN-XGBoost pipeline, suggests the approach could be adapted to other satellite precipitation products and other data-sparse mountain regions worldwide, from the Andes to the Hindu Kush. For a region where every rain gauge covers hundreds of square kilometers of some of the most inaccessible terrain on Earth, teaching satellites to see rainfall more clearly is not just a technical refinement. It is a step toward understanding, and preparing for, the water future of an entire continent.

Subject of Research: Machine learning-based bias correction of GPM IMERG satellite precipitation estimates over the interior of the Tibetan Plateau

Article Title: A precipitation-intensity-aware CNN-XGBoost framework for IMERG bias correction over the interior of the Tibetan Plateau

Article References: A precipitation-intensity-aware CNN-XGBoost framework for IMERG bias correction over the interior of the Tibetan Plateau. (n.d.). https://doi.org/10.1007/s00704-026-06606-z

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06606-z

Keywords: IMERG, Tibetan Plateau, satellite precipitation, bias correction, convolutional neural network, XGBoost, machine learning, extreme rainfall, hydrology, GPM mission, climate data, remote sensing