CORDIS provides links to public deliverables and publications of HORIZON projects.
Links to deliverables and publications from FP7 projects, as well as links to some specific result types such as dataset and software, are dynamically retrieved from OpenAIRE .
Deliverables
The general idea behind the MICRA statistical integration techniques is to combine the appealing spatial and temporal sampling of IR sensors, mounted on geo-stationary platforms, with the higher accuracy of passive MW methods for rain-rate retrieval. The statistical integration techniques are applied within a procedure, which is supposed to run continuously on global scale. This procedure is based on a background process and a foreground process.
The background process consists first of estimating the surface rain-rate from available LEO-MW measurements by means of either empirical retrieval algorithms or inversion schemes based on parametric cloud radiative models (inversion step). This means that we are considering an estimator F-1 which enables the inversion a set of TBs at frequency nn and polarization pm, generally spanning from 10GHz to 150GHz and two linear orthogonal polarizations for rainfall applications, to provide a rain-rate product spatially integrated within the nominal area A.
The second step of the background process pursues the combination of LEO-MW sensor data with data coming from GEO-IR sensor in space and time on a global scale (collocation step). This step consists of temporally locating the GEO-IR data within the series of past ten minutes of the LEO-MW data time and of re-mapping it into the geographic coordinates available both for GEO- IR and LEO-MW measurements and observations. It is worth noting that, since spatial resolution of MW data is generally worse than IR ones, a MW field-of-view of nominal area A generally includes more than one IR pixel. For DMSP-SSM/I products, for instance, the nominal resolution of 25 km corresponds at mid-latitudes at about 5x5 pixels of MeteoSat-VISSR IR channel.
Thus, for a given MW-based rain-rate R, attributed to a nominal area A, we can compute several spatial moment of IR brightness temperature TIR:
- Average value Ta within A;
- Minimum value Tm within A;
- Standard deviation sT within A.
As a result of the background process, a data set is generated, containing the per-pixel rain-rate retrieved from LEO-MW data, the co-located GEO-IR brightness temperature and the pixel geo-location. This process is continuously ongoing, since new LEO-MW and GEO-IR data are continuously ingested on a global scale depending on available satellite platforms.
A physically based classification of the raindrops size distribution (RDSD) in terms of rain intensity R, rain water content W and median volume drop diameter D0 is done with major processes shaping RDSD, by order of practical importance:
- Cloud microstructure:
-- Microphysically continental- larger D0 for the same W, or smaller R for the sameZ;
-- Microphysically maritime- smaller D0 for the same W, or larger R for the sameZ.
- Cloud dynamics
-- Convective - relatively small drops in maritime conditions;
-- Stratiform - larger D0 for the same W compared to the convective maritime.
- Orography:
-- Extremely small D0 - gross radar underestimate of R.
This can explain large systematic variations in the radar Z-R relations, where R for a given Z is greater by a factor of more than 3 for maritime compared to extremely continental clouds, a factor of 1.5 to 2 greater R for stratiform compared to maritime convective clouds, and up to a factor of 10 greater R for the same Z in orographic precipitation.
A large potential exists for significant improvements in radar rainfall estimates by application of a dynamic Z-R relation, based on the microphysical, dynamical and topographical context of the rain clouds.
EURAINSAT/A 1.0 Geo scale rainfall estimates using histogram matching of IR and SSM/I ancillary data
EURAINSAT/A 1.0 is the simplest rainfall algorithm developed in EURAINSAT. Following the recommendations of the International Precipitation Working Group (IPWG), the idea behind it is to provide the end-user with a straightforward but comprehensive package for rainfal estimation using both received or downloaded images.
The algorithm generates rainfall rates using infrared imagery from geostationary satellites like Meteosat or GOES.
Instantaneous rainfall rates can be cumulated to produce climatologies. The instantaneous rainrates are to be directly used at their 0.5 deg resolution.
The algorithm recommends itself for its extreme simplicity of use, which make it possible its application to a large number of environments, even with very low cost equipment and relatively low initial expertise.
The satellite rainfall estimation algorithm relies upon the regional calibration of co-located IR and MW observations and is based upon initial work by Kidd (1999) over the western Pacific. This work found that the calibration thresholds and rain rates varied significantly over relatively short time scales (<<1 month) and must therefore be reflected in the calibration scheme. It was also recognised that the number of valid (i.e. raining) data points needed to be sufficient to be representative of the raining system being observed. This latter issue imposes a constraint on the time and spatial scale of the calibration data set.
Once the two primary data sets have been pre-processed, the IR data is sub-sampled to 1/10th degree resolution and the MW retrieval used to generate the rainfall retrieval, each coincident 1/10th pixel are added to a database. The database contains the number of observations of cloud-top temperatures and rainfall rates for each 1degree grid cell over a one-day period. Therefore, assuming perfect MW coverage (6overpasses a day) and infrared observations, a total of 600 points are possible.
The daily databases are then amalgamated to ensure a more continuous and contiguous coverage. Daily databases for the preceding four days, together with the current day, are accumulated using a time-weighted function. Spatial accumulation is also performed using a distance weighted function over +/- 2 degrees from the central grid cell. These procedures ensure grid-cell to grid-cell as well as day-to-day continuity in the rainfall estimates.
Once the accumulated database has been generated a cumulative histogram matching technique is applied, relating the cold cloud-top temperature pixels with the heaviest rainfall retrievals. This procedure ensures that the statistical distribution of rainfall intensities identified by the PMW data is replicated in the IR-derived rainfall estimates. The matching technique generates a cloud-top temperature to rain rate look-up-table which can then be used to convert the global IR data set into PMW calibrated IR rainfall retrievals. The combined rainfall product has an output resolution of 1/10th degree every 30min, allowing the user to resample to their requirements.
Neural Networks (NN) have been used in the EURAINSAT project in the development of combined IR/MW estimation methods and the simulation of complex rain estimation models. It is well known that instantaneous rainfall estimation through MW sensors presents many advantages over IR-based techniques and gauge measurements but disadvantages, such as the temporal sampling and the spatial resolution, seriously impede their effectiveness. Data fusion and merging approaches using IR information are capable of avoiding these drawbacks without giving up the physically based rainfall discrimination they provide. Some methods even improve the results of the MW sensors themselves.
Several NN were tested, from MLP to Adaptive Resonance Theory (ART) architectures. NN analytical selection process was explained and half hourly rain gauge data over Andalusia, Spain were used for validation purposes. Several interpolation procedures were also tested to transform point to areal measurements, including the maximum entropy method, but it remains unclear which validation methodology adapts better to the satellite rainfall estimation problem. Rainfall estimations were also compared with GOES Precipitation Index (GPI) and cumulative histogram matching (CHM) estimates. Comparison of the NN and CHM estimates against 30 minute 0.1 degree gauge data produced correlations of 0.20 and 0.23 respectively, although both techniques were significantly poorer than the MW alone estimates that produced a correlation of 0.48. Correlations for coarser resolution estimates, for all products, were in excess of 0.8.
Searching for OpenAIRE data...
There was an error trying to search data from OpenAIRE
No results available