Downscaling Algorithms Research Articles

A Spatial Downscaling Algorithm for Satellite-Based Precipitation over the Tibetan Plateau Based on NDVI, DEM, and Land Surface Temperature

Precipitation is an important controlling parameter for land surface processes, and is crucial to ecological, environmental, and hydrological modeling. In this study, we propose a spatial downscaling approach based on precipitation–land surface characteristics. Land surface temperature features were introduced as new variables in addition to the Normalized Difference Vegetation Index (NDVI) and Digital Elevation Model (DEM) to improve the spatial downscaling algorithm. Two machine learning algorithms, Random Forests (RF) and support vector machine (SVM), were implemented to downscale the yearly Tropical Rainfall Measuring Mission 3B43 V7 (TRMM 3B43 V7) precipitation data from 25 km to 1 km over the Tibetan Plateau area, and the downscaled results were validated on the basis of observations from meteorological stations and comparisons with previous downscaling algorithms. According to the validation results, the RF and SVM-based models produced higher accuracy than the exponential regression (ER) model and multiple linear regression (MLR) model. The downscaled results also had higher accuracy than the original TRMM 3B43 V7 dataset. Moreover, models including land surface temperature variables (LSTs) performed better than those without LSTs, indicating the significance of considering precipitation–land surface temperature when downscaling TRMM 3B43 V7 precipitation data. The RF model with only NDVI and DEM produced much worse accuracy than the SVM model with the same variables. This indicates that the Random Forests algorithm is more sensitive to LSTs than the SVM when downscaling yearly TRMM 3B43 V7 precipitation data over Tibetan Plateau. Moreover, the precipitation–LSTs relationship is more instantaneous, making it more likely to downscale precipitation at a monthly or weekly temporal scale.

Impacts of climate change on precipitation and discharge extremes through the use of statistical downscaling approaches in a Mediterranean basin

Mediterranean region is characterized by high precipitation variability often enhanced by orography, with strong seasonality and large inter-annual fluctuations, and by high heterogeneity of terrain and land surface properties. As a consequence, catchments in this area are often prone to the occurrence of hydrometeorological extremes, including storms, floods and flash-floods. A number of climate studies focused in the Mediterranean region predict that extreme events will occur with higher intensity and frequency, thus requiring further analyses to assess their effect at the land surface, particularly in small- and medium-sized watersheds. In this study, climate and hydrologic simulations produced within the Climate Induced Changes on the Hydrology of Mediterranean Basins (CLIMB) EU FP7 research project were used to analyze how precipitation extremes propagate into discharge extremes in the Rio Mannu basin (472.5km2), located in Sardinia, Italy. The basin hydrologic response to climate forcings in a reference (1971–2000) and a future (2041–2070) period was simulated through the combined use of a set of global and regional climate models, statistical downscaling techniques, and a process based distributed hydrologic model. We analyzed and compared the distribution of annual maxima extracted from hourly and daily precipitation and peak discharge time series, simulated by the hydrologic model under climate forcing. For this aim, yearly maxima were fit by the Generalized Extreme Value (GEV) distribution using a regional approach. Next, we discussed commonality and contrasting behaviors of precipitation and discharge maxima distributions to better understand how hydrological transformations impact propagation of extremes. Finally, we show how rainfall statistical downscaling algorithms produce more reliable forcings for hydrological models than coarse climate model outputs.

Mapping Annual Precipitation across Mainland China in the Period 2001–2010 from TRMM3B43 Product Using Spatial Downscaling Approach

Spatially explicit precipitation data is often responsible for the prediction accuracy of hydrological and ecological models. Several statistical downscaling approaches have been developed to map precipitation at a high spatial resolution, which are mainly based on the valid conjugations between satellite-driven precipitation data and geospatial predictors. Performance of the existing approaches should be first evaluated before applying them to larger spatial extents with a complex terrain across different climate zones. In this paper, we investigate the statistical downscaling algorithms to derive the high spatial resolution maps of precipitation over continental China using satellite datasets, including the Normalized Distribution Vegetation Index (NDVI) from the Moderate Resolution Imaging Spectroradiometer (MODIS), the Global Digital Elevation Model (GDEM) from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), and the rainfall product from the Tropical Rainfall Monitoring Mission (TRMM). We compare three statistical techniques (multiple linear regression, exponential regression, and Random Forest regression trees) for modeling precipitation to better understand how the selected model types affect the prediction accuracy. Then, those models are implemented to downscale the original TRMM product (3B43; 0.25° resolution) onto the finer grids (1 × 1 km2) of precipitation. Finally we validate the downscaled annual precipitation (a wet year 2001 and a dry year 2010) against the ground rainfall observations from 596 rain gauge stations over continental China. The result indicates that the downscaling algorithm based on the Random Forest regression outperforms, when compared to the linear regression and the exponential regression. It also shows that the addition of the residual terms does not significantly improve the accuracy of results for the RF model. The analysis of the variable importance reveals the NDVI related predictors, latitude, and longitude, elevation are key elements for statistical downscaling, and their weights vary across different climate zones. In particular, the NDVI, which is generally considered as a powerful geospatial predictor for precipitation, correlates weakly with precipitation in humid regions.

Bayesian Inference aided analog downscaling for near-surface winds in complex terrain

Assessing atmospheric boundary layer flows in complex terrain for short-range real-time applications demands fast and reliable downscaling from coarser-resolution meteorological data to the relevant scale. An ideal statistical downscaling numerical experiment was performed for surface winds above complex terrain in Israel's northern Negev desert region. Dynamical downscaling have been performed by the WRF model to create a historical database by the following two sets. The first set used 5 nested domains from 40.5km to 0.5km. The second set used 3 nested domains ranging from 40.5km to 4.5km. The 4.5km data (stage 2) was defined as predictors while data on 0.5km (stage 1) served as predictands for statistical downscaling. Two statistical downscaling algorithms: minimal distance analog and a Bayesian inference aided analog (hereafter Bayesian algorithm) were tested by the above data. Unlike most analog algorithms, the Bayesian algorithm refers to the probability to get the best analog instead of the minimal differences between predictands. The comparison of the two algorithms shows that the Bayesian approach yields improved results. The Bayesian algorithm reproduces the 0.5km resolution dynamically downscaled surface winds with an average absolute direction difference of 43 and 20 for calm winds and moderate/strong winds respectively. Its average wind speed error is ~1.1ms−1. ~40days are sufficient to create a representative database. Given the database, the procedure is extremely fast (a few seconds) and easy to operate, which makes it suitable for real-time non-expert fast-response applications.

A new satellite-based monthly precipitation downscaling algorithm with non-stationary relationship between precipitation and land surface characteristics

Remote sensing is an important tool to monitor precipitation over regions with sparse rain gauge networks. To provide high-resolution precipitation estimates over un-gauged areas, great efforts have been taken to downscale low-resolution satellite precipitation datasets using the Normalized Difference Vegetation Index (NDVI) and the Digital Elevation Model (DEM) based on the assumption that precipitation can be simulated by vegetation and topography proxies at various spatial scales. However, the non-stationarity of the relationship between precipitation and vegetation or topography has not been appropriately considered when low-resolution satellite precipitation datasets are downscaled using NDVI and DEM in previous studies. To overcome this limitation, a new downscaling algorithm was proposed in this study by introducing a regional regression model termed as geographically weighted regression (GWR) to explore the spatial heterogeneity of the precipitation–NDVI and precipitation–DEM relationships. The performance of this new downscaling algorithm was assessed by downscaling the latest version of monthly TRMM precipitation datasets (referred to TRMM 3B43 V7) over the eastern Tibetan Plateau and the TianShan Mountains from 0.25° (about 25km) to 1km spatial resolution, and the downscaled precipitation datasets were validated against ground observations measured by rain gauges. The validation results indicate that the high-resolution precipitation datasets obtained through the new algorithm not only performed better than the traditional downscaling algorithms, but also had higher accuracy than the original TRMM 3B43 V7 dataset. Besides, we found that the performance of this new algorithm was largely dependent on the accuracy of the original TRMM 3B43 V7 data. We therefore recommend considering the non-stationarity of the precipitation–NDVI and precipitation–DEM relationships in the downscaling process, and demonstrate the possibility of downscaling satellite precipitation with NDVI and DEM at monthly temporal scale.

Content-adaptive image downscaling

This paper introduces a novel content-adaptive image downscaling method. The key idea is to optimize the shape and locations of the downsampling kernels to better align with local image features. Our content-adaptive kernels are formed as a bilateral combination of two Gaussian kernels defined over space and color, respectively. This yields a continuum ranging from smoothing to edge/detail preserving kernels driven by image content. We optimize these kernels to represent the input image well, by finding an output image from which the input can be well reconstructed. This is technically realized as an iterative maximum-likelihood optimization using a constrained variation of the Expectation-Maximization algorithm. In comparison to previous downscaling algorithms, our results remain crisper without suffering from ringing artifacts. Besides natural images, our algorithm is also effective for creating pixel art images from vector graphics inputs, due to its ability to keep linear features sharp and connected.

Soil moisture downscaling across climate regions and its emergent properties

[1] Land surface models of water and energy fluxes can benefit from the characterization of soil moisture variability provided by robust downscaling algorithms over a wide range of climatic settings. In this study, we present the application of a multifractal-based statistical downscaling scheme using 800 m aircraft-derived soil moisture data collected during three field campaigns in contrasting climatic regimes. The disaggregation scheme was tested in a previous work using data of the Southern Great Plains experiment in 1997 (SGP97) in a temperate region in Oklahoma. Here, we explore its capability on different climates by using data from two other campaigns: Soil Moisture Experiment in 2002 (SMEX02), in an agricultural region with subhumid climate in Iowa, and Soil Moisture Experiment in 2004 (SMEX04), conducted in two semiarid areas in Arizona and Sonora (Mexico). We first demonstrate the presence of multifractality in soil moisture fields over the scale range from 0.8 km (aircraft footprint) to 25.6 km (satellite footprint) over most wetness conditions. Next, we identify an empirical regional calibration relation linking model parameters with the spatial mean soil moisture and coarse-scale predictors that account for topography, soil texture, and land cover in each site. The downscaling model shows good performance in a broad range of conditions, except for a few cases where specific physiographic features introduce relevant spatial nonhomogeneity in the soil moisture field. The calibrated downscaling model is then applied to study the relation between spatial variability and mean soil moisture across the different climate settings. In such a way, we explain the diverse shapes presented in previous studies and suggest possible physical explanations for intraregional and interregional differences.

Towards deterministic downscaling of SMOS soil moisture using MODIS derived soil evaporative efficiency

A deterministic approach for downscaling ∼ 40 km resolution Soil Moisture and Ocean Salinity (SMOS) observations is developed from 1 km resolution MODerate resolution Imaging Spectroradiometer (MODIS) data. To account for the lower soil moisture sensitivity of MODIS surface temperature compared to that of L-band brightness temperature, the disaggregation scale is fixed to 10 times the spatial resolution of MODIS thermal data (10 km). Four different analytic downscaling relationships are derived from MODIS and physically-based model predictions of soil evaporative efficiency. The four downscaling algorithms differ with regards to i) the assumed relationship (linear or nonlinear) between soil evaporative efficiency and near-surface soil moisture, and ii) the scale at which soil parameters are available (40 km or 10 km). The 1 km resolution airborne L-band brightness temperature from the National Airborne Field Experiment 2006 (NAFE'06) are used to generate a time series of eleven clear sky 40 km by 60 km near-surface soil moisture observations to represent SMOS pixels across the three-week experiment. The overall root mean square difference between downscaled and observed soil moisture varies between 1.4% v/v and 1.8% v/v depending on the downscaling algorithm used, with soil moisture values ranging from 0 to 15% v/v. The accuracy and robustness of the downscaling algorithms are discussed in terms of their assumptions and applicability to SMOS.