Patterns of
Gridded monthly rainfall estimates can be used for a number of research applications, including hydrologic modeling and weather forecasting. Automated interpolation algorithms, such as the “autoKrige” function in R, can produce gridded rainfall estimates that validate well but produce unrealistic spatial patterns. In this work, an optimized geostatistical kriging approach is used to interpolate relative rainfall anomalies, which are then combined with long-term means to develop the gridded estimates. The optimization consists of the following: 1) determining the most appropriate offset (constant) to use when log-transforming data; 2) eliminating poor quality data prior to interpolation; 3) detecting erroneous maps using a machine learning algorithm; and 4) selecting the most appropriate parameterization scheme for fitting the model used in the interpolation. Results of this effort include a 30-yr (1990–2019), high-resolution (250-m) gridded monthly rainfall time series for the state of Hawai‘i. Leave-one-out cross validation (LOOCV) is performed using an extensive network of 622 observation stations. LOOCV results are in good agreement with observations (
A new method is developed to map rainfall in Hawai‘i using an optimized geostatistical kriging approach. A machine learning technique is used to detect erroneous rainfall maps and several conditions are implemented to select the optimal parameterization scheme for fitting the model used in the kriging interpolation. A key finding is that optimization of the interpolation approach is necessary because maps may validate well but have unrealistic spatial patterns. This approach demonstrates how, with a moderate amount of data, a low-level machine learning algorithm can be trained to evaluate and classify an unrealistic map output.
- Award ID(s):
- 1920304
- NSF-PAR ID:
- 10366870
- Publisher / Repository:
- American Meteorological Society
- Date Published:
- Journal Name:
- Journal of Hydrometeorology
- Volume:
- 23
- Issue:
- 4
- ISSN:
- 1525-755X
- Page Range / eLocation ID:
- p. 561-572
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
Abstract δ 18O andδ 2H in Earth's precipitation provide essential scientific data for use in hydrological, climatological, ecological and forensic research. Insufficient global spatial data coverage promulgated the use of gridded datasets employing geostatistical techniques (isoscapes) for spatiotemporally coherent isotope predictions. Cluster‐based isoscape regionalization combines the advantages of local or regional prediction calibrations into a global framework. Here we present a revision of a Regionalized Cluster‐Based Water Isotope Prediction model (RCWIP2) incorporating new isotope data having extensive spatial coverage and a wider array of predictor variables combined with high‐resolution gridded climatic data. We introduced coupling ofδ 18O andδ 2H (e.g.,d ‐excess constrained) in the model predictions to prevent runaway isoscapes when each isotope is modelled separately and cross‐checked observed versus modelledd ‐excess values. We improved model error quantification by adopting full uncertainty propagation in all calculations. RCWIP2 improved the RMSE over previous isoscape models by ca. 0.3 ‰ forδ 18O and 2.5 ‰ forδ 2H with an uncertainty <1.0 ‰ forδ 18O and < 8 ‰ forδ 2H for most regions of the world. The determination of the relative importance of each predictor variable in each ecoclimatic zone is a new approach to identify previously unrecognized climatic drivers on mean annual precipitationδ 18O andδ 2H. The improved RCWIP2 isoscape grids and maps (season, monthly, annual, regional) are available for download athttps://isotopehydrologynetwork.iaea.org . -
more » « less
Abstract High temporal and spatial resolution precipitation datasets are essential for hydrological and flood modeling to assist water resource management and emergency responses, particularly for small watersheds, such as those in Hawai‘i in the United States. Unfortunately, fine temporal (subdaily) and spatial (<1 km) resolutions of rainfall datasets are not always readily available for applications. Radar provides indirect measurements of the rain rate over a large spatial extent with a reasonable temporal resolution, while rain gauges provide “ground truth.” There are potential advantages to combining the two, which have not been fully explored in tropical islands. In this study, we applied kriging with external drift (KED) to integrate hourly gauge and radar rainfall into a 250 m × 250 m gridded dataset for the tropical island of O‘ahu. The results were validated with leave-one-out cross validation for 18 severe storm events, including five different storm types (e.g., tropical cyclone, cold front, upper-level trough, kona low, and a mix of upper-level trough and kona low), and different rainfall structures (e.g., stratiform and convective). KED-merged rainfall estimates outperformed both the radar-only and gauge-only datasets by 1) reducing the error from radar rainfall and 2) improving the underestimation issues from gauge rainfall, especially during convective rainfall. We confirmed the KED method can be used to merge radar with gauge data to generate reliable rainfall estimates, particularly for storm events, on mountainous tropical islands. In addition, KED rainfall estimates were consistently more accurate in depicting spatial distribution and maximum rainfall value within various storm types and rainfall structures.
Significance Statement The results of this study show the effectiveness of utilizing kriging with external drift (KED) in merging gauge and radar rainfall data to produce highly accurate, reliable rainfall estimates in mountainous tropical regions, such as O‘ahu. The validated KED dataset, with its high temporal and spatial resolutions, offers a valuable resource for various types of rainfall-related research, particularly for extreme weather response and rainfall intensity analyses in Hawai’i. Our findings improve the accuracy of rainfall estimates and contribute to a deeper understanding of the performance of various rainfall estimation methods under different storm types and rainfall structures in a mountainous tropical setting.
-
Constructing retrospective gridded daily weather data for agro‐hydrological applications in Oklahomamore » « less
Abstract Regional, automated meteorological networks, such as the Oklahoma Mesonet can potentially provide high quality forcing data for generating gridded surfaces, but proven methods of interpolating weather variables between the station locations are needed. We compared two interpolation methods, ordinary kriging (OK) and empirical Bayesian kriging (EBK), with and without using long‐term climate imprints (CI), for creating spatially continuous, daily weather datasets. Daily meteorological variables (maximum and minimum temperature, solar radiation, and precipitation) from the Oklahoma Mesonet for the period 1997–2014 were interpolated using geoprocessing tools in ArcGIS. Cross‐validation was used for evaluation of interpolation methods, with 90% of sites chosen randomly for the training set and the remaining 10% left for validation. For all interpolation approaches, cross‐validation showed coefficient of determination (
R 2) values of .99 and .98 for daily maximum and minimum air temperatures, with mean absolute error (MAE) ranging from ±0.45–0.50 °C for maximum temperature and ±0.77–0.80 °C for minimum temperature. Likewise, for daily solar radiation,R2 values of .94 and .93 showed overall good prediction accuracy with MAE values 1.00 and 1.01 MJ m–2 d–1for EBK and OK, respectively. However, for rainfall, all methods yieldedR2 values ≤.67, suggesting a need for more effective interpolation method. Based on its lower computational time and lower input data requirement, OK appears preferable to the other approaches tested here to provide the daily weather data for gridded models in Oklahoma and other regions with similar monitoring networks. -
more » « less
Abstract Wetlands are responsible for 20%–31% of global methane (CH4) emissions and account for a large source of uncertainty in the global CH4budget. Data‐driven upscaling of CH4fluxes from eddy covariance measurements can provide new and independent bottom‐up estimates of wetland CH4emissions. Here, we develop a six‐predictor random forest upscaling model (UpCH4), trained on 119 site‐years of eddy covariance CH4flux data from 43 freshwater wetland sites in the FLUXNET‐CH4 Community Product. Network patterns in site‐level annual means and mean seasonal cycles of CH4fluxes were reproduced accurately in tundra, boreal, and temperate regions (Nash‐Sutcliffe Efficiency ∼0.52–0.63 and 0.53). UpCH4 estimated annual global wetland CH4emissions of 146 ± 43 TgCH4 y−1for 2001–2018 which agrees closely with current bottom‐up land surface models (102–181 TgCH4 y−1) and overlaps with top‐down atmospheric inversion models (155–200 TgCH4 y−1). However, UpCH4 diverged from both types of models in the spatial pattern and seasonal dynamics of tropical wetland emissions. We conclude that upscaling of eddy covariance CH4fluxes has the potential to produce realistic extra‐tropical wetland CH4emissions estimates which will improve with more flux data. To reduce uncertainty in upscaled estimates, researchers could prioritize new wetland flux sites along humid‐to‐arid tropical climate gradients, from major rainforest basins (Congo, Amazon, and SE Asia), into monsoon (Bangladesh and India) and savannah regions (African Sahel) and be paired with improved knowledge of wetland extent seasonal dynamics in these regions. The monthly wetland methane products gridded at 0.25° from UpCH4 are available via ORNL DAAC (
https://doi.org/10.3334/ORNLDAAC/2253 ). -
more » « less
Abstract Soil moisture spatial patterns with length scales of 1‐100 km influence hydrological, ecological, and agricultural processes, but the footprint or support volume of existing monitoring systems, for example, satellite‐based radiometers and sparse in situ monitoring networks, is often either too large or too small to effectively observe these mesoscale patterns. This measurement scale gap hinders our understanding of soil water processes and complicates calibration and validation of hydrologic models and soil moisture satellites. One possible solution is to utilize geostatistical techniques that have proven effective for mapping static patterns in soil properties. The objective of this study was to determine how effectively dynamic, mesoscale soil moisture patterns can be mapped by applying regression kriging to the data from a sparse, large‐scale in situ network. The fully automated system developed here uses several data sets: daily soil moisture measurements from the Oklahoma Mesonet, sand content estimates from the Natural Resource Conservation Service Soil Survey Geographic Database, and an antecedent precipitation index computed from National Weather Service multisensor precipitation estimates. A multiple linear regression model is fitted daily to the observed data, and the residuals of that model are used in a semivariogram estimation and kriging routine to produce daily statewide maps of soil moisture at 5‐, 25‐, and 60‐cm depths at 800‐m resolution. During over 3 years of operation, this mapping system has revealed complex, dynamic, and depth‐specific mesoscale patterns, reflecting the shifting influences of both soil texture and precipitation, with a mean absolute error of ≤0.0576 cm3/cm3across all three depths.
