Land surface temperature anomaly
Land surface temperature is how hot or cold the ground feels to the touch. An anomaly is when something is different from average. The above land surface temperature anomaly map for December 2019 shows how that month’s average temperature was different from the average temperature for all Decembers between 2001 and 2010. Highlighting where earth’s surface was warmer or cooler in the daytime than the average from 2001 – 2010. Warmer than average (red), areas near-normal (white) and places cooler than average (blue). (Sources: Image – NASA; Data – MODIS Satellite, https://neo.sci.gsfc.nasa.gov.)
This study focused on integrating climate and thermal satellite remote sensing data to assess the reaction of the grapevine to a changing environment.
Satellites providing additional information
Climate in the Western Cape is dynamic due to the variation of topography, proximity to the ocean and the sea breeze effect. The current meteorological network is not a dependable source in terms of distribution and consistency of data accuracy due to factors such as financial constraints, vandalism of some stations, out-dated equipment and lack of efficient monitoring of stations. The weather station network in the Western Cape has degraded significantly over the past 10 years, with almost half the stations lost around the year 1998. There is an urgent need for more weather stations in the complex terrain of the Western Cape, but also to find an alternative temperature resource to supplement current weather station networks. The application of traditional geospatial interpolation methods in dynamic terrains remain challenging and difficult to optimise and the accuracy will be highly dependent on station network inputs. The use of land surface temperature (LST) maps from satellites have been used as alternative resource supplement weather station (WS) temperatures in areas of complex terrain.
The availability of continuous climate and weather data at an applicable spatial and temporal level is also crucial to support studies on grapevine phenology, growth and ripening models. This can be achieved by integrating existing weather station networks and remote sensing resources. The study aimed to evaluate the use of satellite LST to supplement WS temperature to aid adaption strategies. One of the most important potential applications of the LST retrieved from satellite data is to validate and improve the global meteorological model prediction.
Daily satellite land surface temperatures is a possible solution to overcome our weather station limitations in complex terrains
The daily LST layers can greatly improve the estimation of temperature in spatiotemporal patterns, improving the knowledge of both climate and biological processes in the wine industry on regional and global spatial scales. The weather station spatial layout in the context of the Moderate Resolution Imaging Spectroradiometer (MODIS) LST pixel outline (1 km x 1 km) and proximity to the ocean are displayed in Figure 1. The LST layers can be used as a daily series of values as the satellite take an image of the entire earth four times per day. The reconstructed daily time series from the satellite can potentially be useful in many cases to substitute meteorological temperature observations.
FIGURE 1. Example of a MODIS tile extent, categorised into growing season temperatures. Distribution of weather stations used in the study also highlighted. In the top right corner is a zoomed-in version highlighting the MODIS tile resolution of 1 km x 1 km pixels.
The study objectives were to:
- Estimate air temperature (WS) using remote sensing LST from the MODIS satellite, (LST) over a set three years (1 April 2012 to 30 April 2015).
- Provide temperature estimations with an accuracy which will support future applications.
- Explore factors and processes influencing the temperature estimation errors.
- Highlight the limitations and possible future developments of using LST in agriculture.
- The estimation of air temperature (WS) using remote sensing LST layers, had the best correlation when the daily mean LST was compared to the daily mean WS. The daily mean LST was calculated form the maximum and minimum LST layers for the day, rather than using the mean temperature averaged from all four LST layers for the day. The mean daily LST layers does sufficiently estimated the daily mean WS temperature.
- Simple statistical methods estimated the daily mean WS temperature to have a temperature difference of 2.3 – 3.6°C compared to the daily mean LST temperatures. The model created from the data source comparison, accounts for the error differences between the two instruments (satellite vs weather stations). Thereby providing a more accurate temperature mean map for the day, ensuring continuous temperature maps by supplementing WS with LST.
- Factors and processes influencing the temperature estimation errors. The extreme sites far inland and near the coast tended to have had lower consistency, with an over estimation of warmer temperatures and underestimation of cooler temperatures. The daily LST temperature layers has its limitations in capturing the extremes within a day due to the fly over time and thermal nature of the instrument. Secondly, cloud cover days result in no data for that image/day, which would leave a gap in the time series. We can overcome these limitations, with the use of the regression equations from the entire study period. The regression equation corrects the errors in the LST layer using the actual WS data values as a calibration factor. The integration of LST and WS results in a more accurate and continuous temperature layer, a layer that is already intrinsically spatial in nature taking into account the terrain complexity of the Western Cape. To also avoid the over and under estimation of LST temperatures, the best correlation was using the mean daily temperature from the LST layers (calculated from the minimum and maximum for the day) rather than using the average of all four layers.
- Although the overall difference between the daily average LST temperature and weather station temperature is relatively high (difference of 2.23°C), the addition of it to the temperature interpolations, and ultimately the bioclimatic indices, added basically no error. As the interpolation algorithm uses the LST only as a means to explain the variation in the temperature, the over- and underestimation within LST itself, might not influence the final interpolated surfaces. A growing degree map (GDM) for the season (1 September to 31 March) was calculated for LST (Figure 2a) and WS with standard interpolations using elevation and continentality as covariates (Figure 2b). As a continuous daily set of maps from September to March is needed for indices calculations, interpolations with only elevation and continentality were incorporated for the days where LST had no data due to cloud cover. The same calculation (as for GDM) was done at 14 weather station points and used as reference for the accuracy assessment. The accuracy assessment of the results comparing the three sources was relatively accurate. The LST, standard spatial interpolation had a 2% difference in the average compared to the reference weather station values.
FIGURE 2A AND 2B. Example of a growing degree map (GDM) for the season (1 September to 31 March), calculated for LST (Figure 2a) and WS with standard interpolations using elevation and continentality as covariates (Figure 2b).
Temporal, spatial and thermal resolutions of mean temperature acquired from daily LST products, offers a new and powerful tool for classification of viticulture landscapes and seasonal monitoring. A wide scope of applications can benefit from the improved remote sensing based mean WS temperature estimations presented in this study for the area of Western Cape.
Two main practical aspects stand out in terms of the contribution of this work to research applications. Firstly, to provide improved spatial-temporally distributed estimations of temperature and indices maps which are particularly important in regions with low station density or with highly variable spatial patterns between stations. Secondly, a collection of remote sensing layers should be created to complement and fill weather station gaps and used for integrated grapevine studies for future use in the wine industry as a key tool for decision making. Future work to quantify climate change in the Western Cape and South Africa can be complemented with the use of intrinsically spatialised remote sensing products, such as land surface temperature layers.
The topography of the Western Cape is complex and changes drastically over short distances which is why the climate changes over short distances. For this reason increased resolution of climate data is crucial for effective adaptive strategies in the context of climate change. Reliable climate data can be costly and currently requires intensive data validation. This study aimed to find an alternative resource to quantify the climate over the spatial extent of the Western Cape, for possible semi-real time applications. Land surface temperature maps are intrinsically spatialised, providing daily temperature values that in the past would have only been possible by spatial interpolation of sparse weather station networks, which could only be as accurate as the input data. The daily mean land surface temperature and weather station temperature data exhibited a strong linear relationship with good prediction accuracy in the complex terrain of the Western Cape. The integration of land surface temperature images with field weather station temperatures will improve the mapping of temperature for improved decision making at farm and field level. This will help producers to stay economically sustainable and to make strategic decisions about future production decisions.
– For more information, contact Tara Southey at email@example.com.