Department of Geography & Environmental Studies, University of Stellenbosch, Private Bag X1, Matieland (Stellenbosch) 7602, South Africa
* Corresponding author: firstname.lastname@example.org
Climate projections for the future suggest favourable conditions for some wine producing regions, but challenging conditions for others. For instance, temperature increases are likely to shift grapevine phenology, ripening and harvest dates, and potentially affect grape quality and yield. The commercial value of accurate and up-to-date climate emerged from feedback received in response to a series of demonstrations to the wine and fruit industries. TerraClim combines high resolution terrain data with weather station data (sourced from several data providers) to model climatic conditions within an orchard or vineyard. The TerraClim climate database allows for dynamic mapping, statistical interrogation, data mining, machine learning and climate change analyses over time and space. The TerraClim initiative has a strong research and development drive that involves continuously updating and extending the climate and terrain databases, automated data collection, interpolation protocol development, as well as the extension of existing logger and weather station networks. The developed technology is novel and scalable to other regions. As proof of concept, the TerraClim webapp (www.terraclim.co.za) presents high temporal and spatial resolution maps of climatic and geographic datasets as a series of dynamic near-real time map layers. The webapp includes an interactive vineyard profiling tool, query functionality and crop/cultivar suitability analysis. TerraClim allows users to obtain pertinent information about climate, terrain and soils to aid long- and short-term agricultural decision-making.
Article extracted from Tara Southey's presentation in occasion of Enoforum Web Conference (23-25 February 2021)
South Africa lies in one of the regions of the world that is more vulnerable to climate variability and change (IPCC, 2014). With the most recent annual warming being driven by increased temperatures in the first and last four months of the year. This is an important time in the agricultural production of southern hemisphere summer-ripening crops such as wine grapes. Season variability is prominent in driving grapevine response, the variability compelled by extreme out of the ordinary climate events such as extreme wind, rainfall or higher temperatures earlier in growing season and ripening period, confirming the unpredictability of seasons predicted in the context of climate change (Southey, 2017). In recent years, the seasonal adaptations (increase irrigation efficiency, additional disease management etc) required to accommodate these changing conditions have resulted in the average wine industry production cost exceeding producers’ income, often leading to a reduction in new investments, reduced long-term vineyard establishment and, in extreme cases, to farm closures and sales. Climate projections for the future suggest benefits for some regions in the Western Cape, but challenges for others (Midgley et al., 2015 and Midgley et al., 2016). Temperature increases may shift grapevine phenology, ripening and harvest dates, and potentially affect grape quality and yield.
In view of climate change, economic pressures and limited water availability in the agricultural sector, information about the suitability of land for viticulture is paramount to aid long and short-term decision making. A local survey conducted among researchers, consultants and producers highlighted the need for accessible data that can support decisions at farm and field level. The absence of a single, integrated database with a user-friendly interface, where viticulturists can obtain pertinent information about climate (as well as terrain and soils), was cited as one of the main obstacles for preparing for the transition of the South African wine industry, necessitated by climate change. However, continuous monitoring of environmental conditions is hampered in the Western Cape by four factors, namely: 1) the expense and logistical difficulties in performing frequent ground-based surveys over an extensive production area; 2) the sparse and irregular distribution of existing weather stations; 3) the inaccessibility (cost) and poor quality of existing weather station data; and 4) the province’s complex terrain, which complicates the modelling of the dramatic local climatic variations observed in many wine producing areas.
TerraClim emanated from recent research (Southey 2017) in which the importance of accurate climatic data for viticultural decision making was illustrated. In addition, TerraClim builds on previous research (Van Niekerk 2010, Van Niekerk & Joubert 2011) on the development and use of terrain and climate data for providing online spatial decision support to the agricultural sector. The inaccessibility to climate and terrain data faced in previous research (Southey, 2017) and the wine industry underpinned the establishment an accessible and reliable central climate data of high temporal resolution, to aid infield decision making. Spatial and temporal data sources need to be integrated to create new spatial and temporal layers of higher resolution, which would aid more insightful within-season decision making and adaptive strategies for a warmer (or cooler) future.
Recent advances in geospatial technologies have opened up new opportunities for generating detailed and accurate climate surfaces. High performance computing (HPC) infrastructure and parallel processing allows for the generation of very high resolution terrain data, known as digital elevation models (DEMs). DEM derivatives (e.g. height, solar radiation) are required to interpolate individual weather station records into accurate wall-to-wall climate surfaces. Very high resolution (<30m) DEMs can improve interpolation accuracies to a point where local climatic variations can be modelled. Remotely sensed co-variates have also been shown to improve interpolation accuracies, especially when weather stations are sparsely distributed. This includes land surface temperatures (captured by the MSG satellite at 15-minute intervals) and rainfall (captured by the GPM satellites at 4-hour interval). Although a large body of knowledge exists on the use of these (and other) earth observation satellites for climatic applications, relatively little has been done on integrating these data sources with terrain and weather station data. TerraClim integrates multiple data resources into a central database with a user-friendly interface that allows users to obtain pertinent information about climate, terrain and soils to aid long- and short-term agricultural decision-making, building resilience in the face of climate uncertainty in the Western Cape.
The TerraClim project aims to improve the understanding of climate change in the Western Cape and how the grapevine/plant responds to these changes. Specific objectives include building a comprehensive climate and terrain database, using new research and technologies to spatialise climate, terrain and vineyard information. TerraClim uses automated functionality to collate, improve, spatialise and disseminate climate data, empowering the farmer and researcher to better analyse and mitigate climate change at regional, farm and vineyard level.
As proof of concept, a set of hourly and daily temperature surfaces were generated and a prototype online web application (www.terraclim.co.za) was developed through which the surfaces could be interactively visualized (Figure 1). The feedback overall was extremely positive and highlighted the immense commercial value of accurate and up-to-date climate surfaces emerged from feedback received in response to a series of demonstrations to the wine and fruit industries. In contrast to other climate data providers that provide climate data recorded at weather stations, TerraClim combines terrain data (supplied by Geosmart) with weather station data (obtained from several data providers) to model climatic conditions at any location within a specified region (e.g. Western Cape). The climate surfaces were developed using hourly weather station data provided by a private weather station network and 2m resolution terrain data provided by Geosmart (www.geosmart.space). The continued aim for the future is to integrate the geodatabase with grapevine plantings and seasonal responses for the identification of cultivar distributions compared to more ideal cultivar distribution in the context of a warmer future with limited water resources.
Figure 1 The TerraClim platform presents high-resolution maps of climatic and geographic datasets as a series of dynamic map layers. The maps can be visualised and overlaid in any given area, much like in a geographic information system. Field report (far right) generated at vineyard level contains terrain, temperature, bioclimatic indices and temperature profiles for multiple seasons.
TerraClim is driven by robust central climate database.
Such a central database does not currently exist for the Western Cape, and portions of existing climate data amongst various custodians are highly variable in completeness, accuracy and structure. TerraClim climate layers are based on a standardised, long-term, live climate database of hourly, daily and monthly temporal resolutions. Automated workflows have been developed for ingesting multiple climate datasets into one standardised climate database. The central climate database incorporates, standardises, gap fills ingested data from over 600 weather stations in the Western Cape. Temperature surface generation is near real time using the regionality surface interpolation methods developed within the TerraClim project to best model the climatic dynamics of a complex terrain such as the Western Cape.
The application of regional climate surface interpolation improves the accuracy and processing time.
The regional interpolation method development for surface temperature, was undertaken by delineating and interpolating the temperature surfaces on subsections (wine regions) of the Western Cape (Figure 2a), resulting in an improved understanding of the topographic differences driving temperature changes. In some areas temperature strongly affected by covariates such as distance to coast, elevation, solar radiation. The regional interpolation improves the final merged output by allowing the incorporation of locally-tailored covariate relationships per region, compared to applying the same, generalised covariate relationships to all regions (Figure 2). The regionality results has allowed for study area to be expanded to cover 33% of the Western Cape and 97.9% of wine grape vineyards (Figure 2). The Spatial distribution of accuracy per region is described in Figure 2 (left), areas in red are greater concern than areas in green. Certain regions identified as “error hotspots” as they have higher RMSE values, highlighting the need for more weather stations to improve accuracies. The research results rendered improved accuracies, faster run times, scalable methodology and valuable recommendations for future weather station density and region size for more accurate temperature layers in the future (Figure 3).
Figure 2 Shows the regional delineation selected for TerraClim processing, area extent covers approximately 97.9% of vineyards in the Western Cape. The image of the LEFT shows the RMSE per region for Study Area, areas highlighted in red are greater concern than areas in green. On the RIGHT is an example of merged daily average temperature (14 February 2019) symbolised using a natural breaks stretch (left) and quantile classification (right).
TerraClim’s climate surfaces are unique in that they are at very high spatial (up to 2m) and temporal (hourly) resolutions. The figures below illustrate the value of very high resolution terrain and climate data .Integrated plant response threshold maps are a value tool to aid in season and long term decision making, as understanding the environment that drives the grapevine physiological responses provides insights for improved decision making. The maps within TerraClim maps provide information about the number of hours the grapevine is at a specific temperature profile. Recent climate change analysis has shown the most significant changes the be happening within the hourly profile of a day. Recent climate change and grapevine response studies (Southey, 2017) show the traditional methods and reviewing the season using daily means, maximum or minimums are no longer sufficient to accurately quantify a season or changes over seasons. Figure 4 is a series of maps for a few hours of the day, highlighting 15:00 to be hottest time of the day, the maps provide temperature profiles where previously there was no data. These maps provide valuable information for infield and farm management when understanding the impact of the temperature profile on the grapevine’s physiological responses. Recent studies (Southey, 2017) on grapevine phenology and climate highlighted that more hours between 35-40°C for the current season, results in early flowering in the next season, mapping this provides additional information spatially of where the changes are happening (Figure5). Figure6 is a series of maps considering hours for a season that affect the grapevines physiological functioning for optimal photosynthesis (Hunter & Bonnardot, 2011; Soltanzadeh et al., 2016). Comparing maps of over season provide even more insights to in field management or management over larger areas comparing regions.
Figure 3 A Visual comparison of final temperature surface layers using the regional interpolation (A) compared to standard interpolation without regionalisation (B). The image on the left (A) captures the regional variation with greater accuracy (more very red and very blue areas), compared to the over-generalised output of the standard interpolation (B).
Figure 4 Hourly maps transitioned into an index map for hours observed between 30-35°C from November to February. The map highlights darker areas where earlier flowering and harvest can be expected compared to lighter areas.
Figure 5 HEAT MAPS for November 2018 to March 2019 : (A) Hours Experienced below 20°C, the areas less favourable for optimal photosynthetic activity (green). The greener areas could require more management practices for more optimal grapevine functioning. (B) Hours Experienced between 25-30°C. More hours at 25-30°C (Greener) are areas of optimal photosynthetic areas, the red areas have less hours at 25-30°C, hence would require more management inputs to ensure optimal photosynthesis. (C) Hours Experienced >35°C . More hours above 35°C, red areas have more hours at warmer conditions, less favourable for optimal photosynthesis, more orange to red areas would require management practices to help with the heat.
The TerraClim platform has devoloped a comprehensive infield viewing and reporting functionality driven by industry user feedback.
The TerraClim platform has devoloped a comprehensive infield viewing and reporting functionality driven by industry user feedback (Figure 1). The modular dashboard allows the user to select, view and interact with their climatic or terrain variables of choice at field level (Figure 6). Selecting a field of interest populates the online dashboard with selected information for that specific field and generates a comprehensive report viewable online or downloadable as a pdf. The report generated has been populated with climate and terrain information at field level, including context for the interpretation of maps and graphs represented within the report. The continued aim of the report is to develop as a guide the aid user interpretation of the graphs and maps within the context of climate change and industry standards. The website www.terraclim.co.za is currently in the final beta version, improving continuously as we acquire new user feedback.
Figure 6 The TerraClim platform presents high-resolution temperature maps, the unique adaptive colours tab when enabled, adjusts the colour ramp (and legend) to the area being viewed within the extend of the viewing window. This provide a more detailed spatial description of temperature changes at field level for a specific day and/or hour.
TerraClim project has developed a grapevine suitability tool that can recommend future plantings, utilising the integrated geodatabase.
The grapevine suitability tool within TerraClim, provides recommend future plantings based on the data within the geodatabase. This new and novel suitability tool developed within TerraClim, can be tailored to any crop type based on the input data for analysis. Extensive consultation with viticulturists, consultants and researchers rendered the best source for suitability analysis to be an existing database of currently planted vineyards in the Western Cape. The suitability analysis is based on a data mining approach, using a test database of actual vineyards (seven wine grape cultivars namely Cabernet Sauvignon, Merlot, Chardonnay, Chenin Blanc, Shiraz, Sauvignon Blanc and Pinotage) planted in the three regions (Figure 7) as an initial analysis for the development of a robust tool. A random forest machine learning classification performed on 42 data sources/layers within the geodatabase resulting in a feature importance list. The analysis showed digital elevation model, solar radiation, slope, distance from coast, aspect, wind speed, growing degree days, growing season temperature, rootstock, trellis system, soil depth respectively to be the top eleven layers (each with a feature importance weighting) that drive cultivar selection/recommendation. The suitability tool has three main functions that returns a suitability recommendation, the user can draw a new polygon/field, select a previously drawn polygon/field or compare two previously drawn polygons/fields with each other (Figure 7).
Figure 7 Shows an example of the new suitability tool developed, the main window of the website displays the suitability study extent (LEFT), the tool has three main functions that returns a suitability recommendation, the user can draw a new polygon/field, select a previously drawn polygon/field (MIDDLE) or compare two previously drawn polygons/fields with each other (RIGHT) which provides a list of recommended cultivars with a summary table (quick summary values of the most important variables driving the cultivar recommendation).
TerraClim project has provided further solutions for improving climate change analysis and accuracies in the future.
Firstly, TerraClim, has identified priority locations in the Western Cape for the establishment of additional weather stations in the future based on a regional approach for identifying most suitable weather station locations that efficiently accounts for local variation in key temperature drivers, namely elevation and solar radiation. The weighting and ranking of quaternary drainage regions illustrate what can be done to prioritize regions for new weather station positioning and can be adapted to include more spatial factors (https://arcg.is/491Ki). The current results of the analysis present a practical solution that can initiate the systematic improvement of the current weather station network to build towards a more densely distributed and representative weather station network that would sufficiently account for regional variations in topography. Figure 8, shows the minimum network that would ensures a more homogenous spatial network of weather stations, overcoming the problem of sparsely and irregular distribution of stations.
Secondly, TerraClim aims to model high-resolution spatial variation, with and without loggers for the improvement of spatial interpolation methodologies (Banghoek Area, high topographic variation). A high density network of newly developed wireless temperature and relative humidity loggers have been installed to better understand the relationship of environmental factors and terrain elements (Figure 9).Experiments will be conducted to assess a) the spatial and temporal variation of temperature dynamics at high resolution and b) the impact of logger data inclusion on the accuracy of interpolations. The objective aims to pursue a more in-depth analysis of variability of temperature at various temporal scales (hourly, daily, monthly, seasonal) and spatial scales (very high to low resolution), as well as its relationship with relevant covariates i.e. elevation, distance to coast and solar radiation. Tests will thus include high resolution covariate regression analysis combining weather station and logger data to establish an improved understanding of the temporal and spatial relationship between temperature and elevation, distance to coast and specifically solar radiation. Further analysis will assess the potential of including logger data in spatial interpolation methodologies to improve regional interpolation accuracy.
Figure 8 Map of some of the newly identified priority locations in the Western Cape for the establishment of additional weather stations, which would ensure a more homogenous spatial network of weather stations for improved accuracy of temperature maps in the future. For more detailed information please review the ESRI story Map at https://arcg.is/491Ki
Figure 9 Map of the recently installed high density, low cost wireless temperature and relative humidity logger network for further research and development within the TerraClim platform.
Accurate climate data and weather forecasts are critical in agriculture as it supports intra-season (e.g. irrigation scheduling, pest control, fertilization) and inter-season (e.g. crop and variety selection) decision-making. High temporal (frequent updates) and spatial (locally relevant) resolution interpolated climate surfaces enable biophysical modelling of plant (crop) responses to irrigation, pesticides, and fertilizer application. It also enables accurate predictions of expected (locally relevant) weather conditions, pest outbreaks, crop quality, and yields. TerraClim is an integrated resource database with a user-friendly interface that allows users to obtain pertinent information about climate, terrain and soils to aid long- and short-term agricultural decision-making. The platform presents high-resolution maps of climatic and geographic datasets as a series of dynamic map layers. The maps can be visualised and overlaid in any given area, much like in a geographic information system.
The TerraClim project has a strong research and development component that involves frequently updating and extending the climate and terrain databases, automated data collection, interpolation protocol development, as well as the extension of existing logger and weather station networks. This tool allows the wine industry – and the agricultural sector in general – to better understand the complexity of the Western Cape’s climate and terrain at a higher spatial (geographical) resolution for improved adaptation to climate change.
This work is funded by Winetech, supported by Stellenbosch University and the Centre for Geographical Analysis (CGA).
Acknowledgement of co-workers in the study namely Prof Van Niekerk, G. Stephenson, C. Theron, A. Prins, T. Pauw, Z. Mounton, L Vermeulen.
Hunter, J.J. & Bonnardot, V., 2011. Suitability of some climatic parameters for grapevine cultivation in South Africa, with focus on key physiological processes. S. Afr. J. Enol. Vitic 32, 137-154.
IPCC, 2014. Climate Change 2014: synthesis report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Core Writing Team, Pachauri RK, Meyer LA (eds.), IPCC, Geneva, Switzerland
Midgley, G.F., New, M., Johnston, P., Methner, N., Cole, M., Cullis, J., Drimie, S., Dzama, K., Guillot, B.,Harper, J., Jack, C., Knowles, T., Louw, D., Mapiye, C., Oosthuizen, H.J. & Smit, J., 2015. A status quo review of climate change and the agriculture sector of the Western Cape Province. CSIR Environmentek, Stellenbosch CSIR. CSIR, P.O. Box 395; Pretoria 0001; South Africa
Midgley, G.F., New, M., Methner, N., Cole, M., Cullis, J., Drimie, S., Dzama, K., Guillot, B., Harper, J., Jack, C., Johnston, P., Knowles, T., Louw, D., Mapiye, C., Oosthuizen, H.J., Smit, J. & Van dern Broeck, D., 2016. A status quo review of climate change and the agriculture sector of the Western Cape Province. CSIR Environmentek, Stellenbosch CSIR Report No. ENV-S-C 2005-073. CSIR, P.O. Box 395; Pretoria 0001; South Africa.
Soltanzadeh, I., Bonnardot, V., Sturman, A., Quénol, H. & Zawar-Reza, P., 2016. Assessment of the ARW-WRF model over complex terrain: the case of the Stellenbosch Wine of Origin district of South Africa. Theor. Appl. Climatol., 1-21.
Southey, T.O., 2017. Integrating climate and satellite remote sensing to assess the reaction of Vitis vinifera L. cv. Cabernet Sauvignon to a changing environment. PhD Thesis, Stellenbosch University, Private Bag X1, 7602 Matieland (Stellenbosch), South Africa. http://hdl.handle.net/10019.1/101124
Van Niekerk, A., 2010. A comparison of land unit delineation techniques for land evaluation in the Western Cape, South Africa. Land Use Policy 27, 3, July 2010, 937-945. https://doi.org/10.1016/j.landusepol.2009.12.007
Van Niekerk, A., & Joubert, S.J., 2011. Input variable selection for interpolating high-resolution climate surfaces for the Western Cape. Water SA 37, 271-280. https://doi.org/10.4314/wsa.v37i3.68475