Setting Up Optimal Meteorological Networks: An Example From Italy

Rita Aromolo*, Valerio Moretti and Tiziano Sorgi

Council for Agricultural Research and Economics (CREA), Rome, Italy

E-mail: rita.aromolo@crea.gov.it

*Corresponding Author

Received 18 July 2021; Accepted 21 July 2021; Publication 05 October 2021

Abstract

A permanent assessment of climate regime in forest sites has a key role in forest resource conservation and preservation of ecosystem services, biodiversity and landscape multi-functionality, informing sustainable forest management. In this view, time-series of meteorological data relative to several monitoring sites from the ICP-Forest network in Italy, were analyzed with the aim to define the number of site-specific observations, which can be considered adequate for further analysis on forest resource management. The relative importance of each factor accounted in our analysis (season, year, variable, plot, sampling proportion) was investigated comparing results through the use of descriptive statistics.

Keywords: Climate, gauging stations, descriptive statistics, Mediterranean Europe.

Introduction

A permanent assessment of climate regime in forest sites has a key role in forest resource conservation and preservation of ecosystem services, biodiversity and landscape multi-functionality, informing sustainable forest management (Marchetti et al. 2014, Salvati et al. 2016). In this view, time-series of meteorological data relative to several monitoring sites from the ICP-Forest network in Italy, were analyzed with the aim to define the minimum number of site-specific observations, which can be considered adequate for further analysis on forest resource management (Marchi et al. 2017). for each recorded variable, missing values were filled by using a multivariate Singular Value Decomposition (SVD) imputation. A decreasing number of records were progressively removed using a bootstrap repetition following evaluation of the representativeness of the remained records in terms of error of estimation and explained variance on the whole climatic period. Moreover, the relative importance of each factor accounted in our analysis (season, year, variable, plot, sampling proportion) was investigated fitting a model on the results of the overall procedure.

images

Figure 1 Spatial distribution of the monitoring sites.

Methodology

The climatic data derives form the data logger installed at 13 selected areas: ABR 1 Selva Piana; CAL 1 Piano Limina; EMI 1 Carrega; LAZ 1 Monte Rufeno; PIE 1 Valsessera; VEN 1 Pian Cansiglio; EMI 2 Brasimonte; FRI 2 Tarvisio; LOM 1 Val Masino; PUG 1 Foresta Umbra; TRE 1 Passo Lavaz ; LAZ 2 Monte Circeo; BOL 1 Renon (Figure 1). The detected variables, monitored over multiple annual cycles, are air temperature for the profile of 0.1 m and 2 m (AT01 and AT2), relative humidity for the profile of 0.1 m and 2 m (RH01 and RH2), and precipitation (PR). The data time lapse is 1th January 1998–31th December 2013, with 1-day sampling interval. The daily data has been obtained from the average of each performed scan (sensors scan frequency is 10 seconds) for each examined variable, except for the precipitation, which derives from the measurements sum.

Results and Discussion

Results are reported in the following tables and figures.

Table 1 Description of the analyzed time period, including the available years number of monitoring (N years), the starting and ending date, the valid and missing values number over a total of 5844 observations, calculated for each variable, relative to 13 selected meteorological stations

Stations Statistics AT01 AT2 RH01 RH2 PR
01-ABR1 N years 4 16 4 16 16
Starting date 01/01/1998 01/01/1998 01/01/1998 01/01/1998 01/01/1998
Ending date 26/11/2001 31/12/2013 26/10/2001 31/12/2013 31/12/2013
N valid data 1215 4473 4242 981 4178
N missing data 4629 1371 4863 1666 1602
03-CAL1 N years 12 12 12 12 12
Starting date 01/05/1999 01/05/1999 01/05/1999 01/05/1999 01/05/1999
Ending date 26/07/2010 26/07/2010 26/07/2010 26/07/2010 26/07/2010
N valid data 2997 2949 2977 2997 2991
N missing data 2847 2895 2847 2853 2867
05-EMI1 N years 16 16 16 16 16
Starting date 01/01/1998 01/01/1998 01/01/1998 01/01/1998 01/01/1998
Ending date 14/10/2013 14/10/2013 14/10/2013 14/10/2013 14/10/2013
N valid data 5698 5698 5726 5722 5722
N missing data 146 146 122 122 118
09-LAZ1 N years 16 16 16 16 16
Starting date 01/01/1998 01/01/1998 01/01/1998 01/01/1998 01/01/1998
Ending date 29/12/2013 29/12/2013 29/12/2013 29/12/2013 29/12/2013
N valid data 5532 5531 5582 5535 5567
N missing data 312 313 309 277 262
12-PIE1 N years 14 14 14 14 14
Starting date 01/11/1999 01/11/1999 01/11/1999 01/11/1999 01/11/1999
Ending date 12/11/2012 12/11/2012 12/11/2012 12/11/2012 12/11/2012
N valid data 4130 4130 4188 4130 4130
N missing data 1714 1714 1714 1714 1656
20-VEN1 N years 15 15 15 15 15
Starting date 01/08/1999 01/08/1999 01/08/1999 01/08/1999 01/08/1999
Ending date 31/12/2013 31/12/2013 31/12/2013 31/12/2013 31/12/2013
N valid data 4799 4982 4933 4812 4916
N missing data 1045 862 1032 928 911
06-EMI2 N years 11 11 10 11 11
Starting date 20/10/1998 20/10/1998 09/06/1999 20/10/1998 20/10/1998
Ending date 31/12/2008 31/12/2008 31/12/2008 31/12/2008 31/12/2008
N valid data 3172 3205 3479 2840 3082
N missing data 2672 2639 3004 2762 2365
08-FRI2 N years 15 15 15 15 15
Starting date 05/06/1998 05/06/1998 05/06/1998 05/06/1998 05/06/1998
Ending date 04/04/2012 04/04/2012 04/04/2012 04/04/2012 04/04/2012
N valid data 4801 4801 4773 4801 4558
N missing data 1043 1043 1043 1286 1071
10-LOM1 N years 10 10 10
Starting date 16/04/2003 16/04/2003 16/04/2003
Ending date 04/11/2013 04/11/2013 04/11/2013
N valid data 2383 2052 2383
N missing data 5844 3461 5844 3461 3792
13-PUG1 N years 5 5 5
Starting date 28/05/2009 28/05/2009 28/05/2009
Ending date - 31/12/2013 31/12/2013 31/12/2013
N valid data 1608 1610 1608
N missing data 5844 4236 5844 4236 4234
17-TRE1 N years 16 16 16 16 16
Starting date 17/09/1998 17/09/1998 17/09/1998 17/09/1998 17/09/1998
Ending date 07/04/2013 26/10/2001 07/04/2013 07/04/2013 26/10/2001
N valid data 5205 5205 5082 5205 5205
N missing data 639 639 639 639 762
22-LAZ2 N years 8 8 8
Starting date 27/07/2005 27/07/2005 27/07/2005
Ending date 27/06/2012 27/06/2012 27/06/2012
N valid data 2458 2458 2457
N missing data 5844 3386 5844 3387 3386
27-BOL1 N years 9 4 9
Starting date 11/02/2004 17/09/2004 07/01/2004
Ending date 07/11/2012 31/12/2007 31/12/2007
N valid data 2658 1144 2798
N missing data 5844 3186 5844 3046 4700

Table 2 Summary statistics, including the mean and median, standard deviation and coefficient of variation values, calculated for each variable, relative to 13 selected sites during the entire analyzed time period. Note that the variation coefficient (CV) is a dimensionless quantity

Stations Statistics AT01 (C) AT2 (C) RH01 (%) RH2 (%) PR (mm)
01-ABR1 Mean 6.63 7.21 79.62 83.12 2.06
Median 7.1 7.6 82.1 86.4 0
Std Dev 6.66 6.74 16.49 15.87 6.61
CV 100.5 93.48 20.71 19.1 320.65
03-CAL1 Mean 10.83 11 89.17 86.4 3.91
Median 10.8 11 95.5 93 0
Std Dev 6.23 6.24 13.87 15.64 10.22
CV 57.53 56.71 15.56 18.1 261.12
05-EMI1 Mean 12.22 12.02 81.16 80.64 1.92
Median 12.5 12.3 84.75 85.2 0
Std Dev 8.17 8.06 17.28 18.37 6.1
CV 66.89 67.03 21.29 22.78 318.06
09-LAZ1 Mean 11.25 11 79.63 75.62 2.51
Median 11.1 10.6 82 78.2 0
Std Dev 7.3 6.92 15.18 17.51 6.99
CV 64.86 62.93 19.06 23.16 279.01
12-PIE1 Mean 6.09 7.31 85.94 79.27 4.54
Median 6.1 7.3 93.9 84.9 0
Std Dev 6.96 6.82 17.23 20.1 15.52
CV 114.35 93.19 20.05 25.35 341.81
20-VEN1 Media 4.5 5.72 91.34 85.75 4.72
Median 4.8 6 93.4 90.1 0
Dev std 7.28 7.2 8.98 14.2 13.85
CV 161.67 125.89 9.83 16.56 293.21
06-EMI2 Mean 9.8 9.54 82.07 80.06 3.64
Median 10 9.3 85.9 84.2 0
Std Dev 7 7.03 16.36 17.36 9.95
CV 71.45 73.67 19.94 21.69 273.66
08-FRI2 Mean 6.16 5.91 92.63 91.36 3.62
Median 6.6 6.3 96.3 95.5 0
Std Dev 7.56 7.64 8.79 10.89 11.05
CV 122.7 129.15 9.49 11.92 305.05
10-LOM1 Mean 7.55 71.1 2.96
Median 8.2 73 0
Std Dev 6.71 17.34 8.13
CV 88.82 24.38 274.13
13-PUG1 Mean 11.91 78.7 2.09
Median 11.7 83 0
Std Dev 6.87 13.71 6.99
CV 57.71 17.41 334.7
17-TRE1 Mean 1.8 2.21 94.8 86.39 2.19
Median 1.1 2 99.7 91.2 0
Std Dev 7.24 7.29 9 14.44 5.89
CV 401.24 330.18 9.49 16.71 268.88
22-LAZ2 Mean 15.72 77.42 1.3
Median 15.3 79 0
Std Dev 5.96 12.34 4.43
CV 37.9 15.94 340.25
27-BOL1 Mean 4.45 72.75 3.01
Median 4.7 75 0
Std Dev 6.91 18.04 8.91
CV 155.35 - 24.79 295.97

The presence of missing values in a data set can affect the conclusions made using the data. Missing data values must not only be identified, they must also be understood before further analysis can be conducted.

The data filling analysis can be useful for prevention and treatment of missing data. A multivariate Singular Value Decomposition (SVD) imputation has been used to simulate missing values. The algorithm commonly known as Principal Component Analysis (PCA), for instance, is just a simple application of the SVD. In this method, the SVD is used to obtain a set of orthogonal vectors that are linearly combined to estimate missing data values; this procedure is repeated until it converges and the convergence depends on the configuration of the missing entries. The fact that SVD gives us an optimal low-rank representation guarantees that this sort of simulation preserves most of the detail in the data matrix. The time series data analyses relevant to each analyzed variable in the 13 selected sites have been explored to cross-validate the SVD imputation. Figure 3 depicts both the source and implemented time series data analyses relevant to each analyzed variable in all the stations, as a representative example for all stations.

The time series trend analysis has also been performed to verify if there is a linear long-term increase or decrease in both the source and filled data (Table 3).

images

Figure 2 Number of valid and missing values, calculated for each variables, relative to 13 selected sites during the entire analyzed time period.

images

Figure 3 Meteorological time series of air temperature (0.1 and 2 m), relative humidity (0.1 and 2 m), and precipitation, as recorded at ABR1 station during the considered time period. The source data are represented by blue points, the simulated data by red ones.

Table 3 The source/filled time series linear trend parameters, calculated for each analyzed variable in all the stations

Source Data
AT01 AT2 RH01 RH2 PR

Slope Intercept Slope Intercept Slope Intercept Slope Intercept Slope Intercept
Station (C year-1) (C) (C year-1) (C) (C year-1) (C) (C year-1) (C) (C year-1) (C)

‘01-ABR1’ 7,35E-01 5,16 2,53E-02 7,00 -2,22E+00 84,11 8,25E-01 76,38 -3,37E-02 2,34
‘03-CAL1’ -9,26E-02 11,44 -9,94E-02 11,65 3,86E-01 86,62 6,75E-01 81,95 8,13E-02 3,37
‘05-EMI1’ 2,50E-02 12,02 -9,05E-03 12,09 7,30E-01 75,39 5,15E-03 80,60 -1,96E-03 1,93
‘09-LAZ1’ 1,99E-01 9,72 1,04E-02 10,92 -4,63E-01 83,23 -3,70E-01 78,49 6,16E-02 2,03
‘12-PIE1’ 9,42E-03 6,01 1,20E-01 6,34 1,01E+00 77,70 1,09E+00 70,35 -9,70E-02 5,33
‘20-VEN1’ -1,72E-02 4,65 9,56E-02 4,89 -4,29E-01 95,05 -1,42E+00 97,90 -1,64E-02 4,87
‘06-EMI2’ 4,49E-01 7,22 2,18E-01 8,30 1,30E-01 81,26 -1,67E-01 81,02 -4,15E-02 3,88
‘08-FRI2’ -6,18E-02 6,61 -1,78E-01 7,21 4,20E-01 89,57 6,96E-01 86,24 -2,10E-01 5,15
‘10-LOM1’ - - -2,36E-01 9,80 - - 1,44E+00 57,36 -1,18E-01 4,14
‘13-PUG1’ - - -3,32E-02 12,37 - - 1,76E-02 78,46 -1,80E-01 4,56
‘17-TRE1’ -1,19E-01 2,74 -7,87E-02 2,83 8,49E-01 88,12 2,35E-01 84,54 -4,62E-02 2,56
‘22-LAZ2’ - - -5,53E-02 16,33 - - -4,10E-01 81,95 -1,86E-01 3,35
‘27-BOL1’ - - 1,92E-01 2,46 - - -5,05E-01 78,32 -1,47E+00 15,34
‘01-ABR1’ -7,93E-03 3,68 1,60E-01 4,64 -1,98E-01 82,06 8,55E-01 75,51 4,07E-02 0,46
‘03-CAL1’ 1,55E-02 11,63 4,56E-01 10,43 -2,22E-02 89,81 7,91E-01 81,14 2,97E-01 5,72
‘05-EMI1’ 2,32E-02 11,90 1,79E-02 11,84 7,13E-01 75,49 5,25E-02 80,25 2,29E-02 1,83
‘09-LAZ1’ 2,53E-01 9,46 7,99E-02 10,65 -4,47E-01 83,05 -3,14E-01 78,13 8,75E-02 1,92
‘12-PIE1’ -2,26E-01 7,52 5,82E-02 7,08 9,50E-01 77,35 8,23E-01 72,12 -3,78E-02 4,99
‘20-VEN1’ -2,65E-02 4,93 7,55E-02 5,15 -6,01E-02 91,52 -7,74E-01 91,54 7,56E-02 3,75
‘06-EMI2’ 9,60E-02 8,84 1,64E-01 8,94 2,11E-02 80,99 -4,82E-02 79,04 -1,51E-01 3,54
‘08-FRI2’ 3,92E-02 6,61 4,44E-02 6,10 3,00E-01 90,34 2,64E-01 88,33 -9,54E-02 4,41
‘10-LOM1’ -1,96E-01 3,33 1,29E-01 3,84 7,91E-01 64,37 8,70E-01 60,67 -1,15E-02 1,77
‘13-PUG1’ 7,39E-02 5,47 4,31E-01 4,49 -6,03E-02 74,93 4,46E-01 72,89 1,89E-01 -1,50
‘17-TRE1’ -3,45E-02 2,78 4,68E-02 2,69 6,75E-01 88,78 2,92E-01 84,07 1,23E-02 1,93
‘22-LAZ2’ 9,04E-02 7,76 4,97E-01 7,92 -1,63E-01 74,66 6,45E-01 69,23 3,15E-02 1,22
‘27-BOL1’ -1,06E-01 5,08 -2,37E-02 5,23 4,11E-01 75,18 1,32E-01 70,93 -6,88E-02 4,91

Conclusions

The climatic data, collected in the 16-year period (1998–2013), has been evaluated as a part of the SMART4Action Project. The integration of climate impact assessment into the forest response is fundamental to achieve sustainable forest management. The climatic data derives form a monitoring network composed by 13 test sites. In particular, the analyzed variables, detected in each site, are air temperature and relative humidity for the profile of 0.1 m and 2 m and precipitation. Time series represent the time-evolution of the meteorological dynamic process and are used to evaluate patterns and behavior in data over time and to examine daily, weekly, seasonal or annual variations, or before-and-after effects of a process change. This data has been used to provide an overview of the climatic conditions of the tested areas, including a summary statistics of the measured variables in each provided sites and their time-evolution. Moreover, once defined the number of valid and missing values, a data filling analysis has been performed to obtain useful information for prevention and treatment of missing data. To check whether a linear long-term increase or decrease occurs in both source and filled data, a time series trend analysis has also been performed. The source and simulated results comparison has revealed that a proper data filling procedure can be used when valid data are not available.

References

Bertini G., Amoriello T., Fabbio G., Piovosi M., 2011. Forest growth and climate change. Evidences from the ICP-Forests intensive monitoring in Italy. iForest 4: 262–267.

Breiman L (2001). Random forests. Machine learning 5–32. doi: 10.1023/A:1010933404324

Fabbio G, Merlo M, Tosi V., 2003. Silvicultural management in maintaining biodiversity and resistance of forests in Europe—the Mediterranean region. Journal of Environmental Management 67: 67–76.

Fares S., Mugnozza G. S., Corona P., Palah M., 2015. Sustainability: Five steps for managing Europe’s forests. Nature 519 (7544): 407–409.

Ferrara C., Marchi M., Fares S., Salvati, L., 2017. Sampling strategies for high quality time-series of climatic variables in forest resource assessment. iForest, underlineaccepted.

Ferretti M., Fischer R., editors, 2013. Forest Monitoring, Vol 12, DENS, UK: Elsevier, 2013, 507 ps.

Golub G. and Kahan W., 1965. Calculating the singular values and pseudo-inverse of a matrix, J. Soc. Indust. Appl. Math. Ser. B Numer. Anal., 2: 205–224.

Jump S.A., Penuelas J., 2005. Running to stand still: adaptation and the response of plants to rapid climate change. Ecology Letters 8: 1010–1020.

Lindner M., Maroschek M., Netherer S., Kremer A., Barbati A., Garcia-Gonzalo J., Seidl R., Delzon S., Corona P., Kolstroma M., Lexer MJ., Marchetti M., 2010. Climate change impacts, adaptive capacity, and vulnerability of European forest ecosystems. Forest Ecology and Management 259: 698–709.

Marchetti M, Vizzarri M, Lasserre B, Sallustio L, Tavone A (2014). Natural capital and bioeconomy?: challenges and opportunities for forestry. Annals of Silvicultural Research 38:62–73. doi: 10.12899/ASR-1013.

Marchi M, Ferrara C, Bertini G, Fares S, Salvati L (2017b). A sampling design strategy to reduce survey costs in forest monitoring. Ecological Indicators 81:182–191. doi: 10.1016/j.ecolind.2017.05.011

R Core Team (2017). R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Salvati L, Becagli C, Bertini G, Cantiani P, Ferrara C, Fabbio G (2016). Toward sustainable forest management indicators? A data mining approach to evaluate the impact of silvicultural practices on stand structure. International Journal of Sustainable Development & World Ecology 0:1–11. doi: 10.1080/13504509.2016.1239138

Scheick J.T., Linear algebra with applications, McGraw-Hill, New York, 1996.

Travaglini D., Chirici G., Bottalico F., Ferretti M., Corona P., Barbati A., Fattorini F., 2013. Large-Scale Pan-European Forest Monitoring Network: A Statistical Perspective for Designing and Combining Country Estimates. Example for Defoliation. In Marco Ferretti, Richard Fischer, editors: Forest Monitoring, Vol 12, DENS, UK: Elsevier, 2013, pp. 105–136.

Biographies

images

Rita Aromolo – Senior Technologist, degree in Biological Sciences from the La Sapienza University of Rome and specialization in General Pathology. Since 1988 she has been working for the Council for Agricultural Research and Analysis of the Agricultural Economy at Center of Agriculture and Environment. She carries out research in the study and analysis of heavy metals in soil and plants, chemical indicators of soil fertility in assessing the environmental impact, the influence of cultivation practices on product quality and chemical characteristics physics of soils, air pollution and environmental quality. Responsible for various research also in the context of Mipaaf projects, the most recent “Valorbio”, Agroener, and thesis co-supervisor in the degree course in Forestry Sciences of the University of Tuscia of Viterbo. Since 1996 she has been responsible for various conventions in the environmental monitoring of the Presidential Estate of Castelporziano. She cooperates in various research projects such as projects aimed at optimizing fertilization through the use of biomass of various kinds on crops for industrial use, for the agricultural enhancement of waste biomass, and for the effects of the dynamics of unwanted elements on the soil. Since 2014 She has been working as an expert in the scientific technical group set up by interministerial decree as part of the investigations on the Terra dei Fuochi. She is the author of over 140 scientific publications and two patents.

images

Valerio Moretti, born in Rome on 16/12/1980. In 1999 he obtained a scientific high school diploma and subsequently began working at CREA in the context of national projects for the environmental protection of the Presidential Estate of Castelporziano. Since 2016 he has been dealing with the management of European projects both for the technical and administrative aspects. The most relevant technical tasks are the installation of weather stations in Italy and the implementation of climate databases.

images

Tiziano Sorgi, born in Rome on 24/07/1974. In 1993 he graduated as a technician of the electrical and electronic industries with a score of 60/60 and subsequently obtained a diploma of hardware and software technician with a score of 30/30. In 1998 he was hired at CREA where he is currently employed as a computer technician and electronic technician; among the most important tasks are the installation of meteorological stations on the Italian territory, the management of climate databases and the wiring of electronic instruments to the relative acquisition devices. The passion for electronics and for the computer led him to implement the design of interfaces and devices through the use of Arduino and the knowledge of the Visual C++ programming language.

Abstract

Introduction

images

Methodology

Results and Discussion

images

images

Conclusions

References

Biographies