Contents lists available at ScienceDirect Agricultural and Forest Meteorology journal homepage: www.elsevier.com/locate/agrformet A new spatial modeling and interpolation approach for high-resolution temperature maps combining reanalysis data and ground measurements Mariassunta Viggiano a, ⁎ , Lorenzo Busetto b , Domenico Cimini a,c , Francesco Di Paola a , Edoardo Geraldi a , Luigi Ranghetti b , Elisabetta Ricciardelli a , Filomena Romano a a National Research Council of Italy, Institute of Methodologies for Environmental Analysis, (CNR-IMAA), 85050, Tito Scalo (PZ), Italy b National Research Council of Italy, Institute on Remote Sensing of Environment, (CNR-IREA), 20133, Milano, Italy c CETEMPS Center of Excellence, Università dell'Aquila, 67100, L’Aquila, Italy ARTICLE INFO Keywords: Spatial statistics Geostatistics Meteorological downscaling Agro-meteorology Kriging Rice cropping ABSTRACT Despite of their increasing importance as inputs to models for a wide range of scientific fields, high-resolution meteorological variables are not recorded very often on spatially regular grids. This problem is usually overcome by using data from reanalysis models, although they are less accurate. This paper discusses the development of a new spatial downscaling methodology to provide high-resolution maps of daily maximum and minimum air temperature. The application of this approach provides thorough observations in sparsely sampled areas by combining the accuracy of measurements from ground-based stations with the high availability and uniformity of model-based data. The dataset includes more than a decade (2003–2013) of data collected at 113 stations, about 30% of which constituted an independent set for the validation procedure. The efficacy of this approach is evaluated using statistical scores that are regularly employed in model evaluation studies and the improvements over the classical approach are remarkable. The results show that overall the our "hybrid" method provides fair estimates of temperature values. Particularly, MBE is less than 0.29 °C and 0.60 °C for the daily maximum and minimum air temperature respectively; RMSE is less than 1.24 °C for the maximum temperature and 1.86 °C for the minimum temperature, the analysis on MAE assures that there is not contribution of the errors in the spatial variability (MAE ≈ RMSE). The correlation coefficient, close to 1 (ρ ≈ 0.97), indicates a strong positive linear relationship. 1. Introduction Due to practical constrains, such as installation and maintenance costs, networks of weather sensor stations for ground observations of meteorological parameters are necessarily limited and not deployed on a regular grid. Consequently, many areas have only a sparse coverage. Therefore the availability of meteorological measurements demon- strates to be spatially inadequate for most applications (Goovaerts, 1999; Miller and Franklin, 2002; Robertson, 1987; Running et al., 1987). Besides this, data from these networks may exhibit temporal drop out and spatial gaps due to sensor failure (Folken, 2017; Liu and Tang, 2014). When working on areas larger than an intensive research site, as it happens when crop models are applied on a regional scale, it is often the case that the research points are located substantially far away from a meteorological station and therefore the available climatic data are not spatially dense nor representative enough to ensure a correct output of the model. For these reasons, meteorological data on a virtually continuous spatial scale are required, precluding that the collection of data from individual stations and the use of values re- corded by the nearest station provide feasible solutions. In order to reconstruct a model of the field of interest for the entire surface, in- terpolation techniques are mandatory (Goovaerts, 1999; Hudson and Wackernagel, 1994; Hungerford et al., 1989; Thornton et al., 1997). In this scenario, statistical downscaling has become a powerful tool. An increasing number of downscaling studies based on different statistical techniques applied to different variables and domains have been de- scribed in several textbooks and scientific publications (e.g. Benestad et al., 2008; Hewitson and Crane, 1996; Wilby and Wigley, 1997; Wilks, 2011) in the last years. But, it should also be noted that, in a process of interpolation, a defective station could strongly affect the resulting https://doi.org/10.1016/j.agrformet.2019.05.021 Received 10 May 2018; Received in revised form 4 April 2019; Accepted 21 May 2019 ⁎ Corresponding author at: Institute of Methodologies for Environmental Analysis (IMAA), C.da S. Loja, 85050, Tito Scalo (PZ), Italy. E-mail addresses: mariassunta.viggiano@imaa.cnr.it (M. Viggiano), busetto.l@irea.cnr.it (L. Busetto), domenico.cimini@imaa.cnr.it (D. Cimini), francesco.dipaola@imaa.cnr.it (F. Di Paola), edoardo.geraldi@imaa.cnr.it (E. Geraldi), ranghetti.l@irea.cnr.it (L. Ranghetti), elisabetta.ricciardelli@imaa.cnr.it (E. Ricciardelli), filomena.romano@imaa.cnr.it (F. Romano). Agricultural and Forest Meteorology 276-277 (2019) 107590 0168-1923/ © 2019 Elsevier B.V. All rights reserved. T